diff --git a/skills/add-orbitinfer-model/SKILL.md b/.agents/skills/add-orbitinfer-model/SKILL.md similarity index 91% rename from skills/add-orbitinfer-model/SKILL.md rename to .agents/skills/add-orbitinfer-model/SKILL.md index f6fccd8..22b2a99 100644 --- a/skills/add-orbitinfer-model/SKILL.md +++ b/.agents/skills/add-orbitinfer-model/SKILL.md @@ -30,19 +30,19 @@ encoding, MoE routing, recurrent/sliding state, vocabulary, and multimodal detai Reuse an existing recipe when its formatting, parsing and media semantics match. When working in the OrbitInfer repository or a fork, implement a new native Rust recipe -in [`orbitinfer-recipe`](../../crates/orbitinfer-recipe/README.md) and its bindings. When +in [`orbitinfer-recipe`](../../../crates/orbitinfer-recipe/README.md) and its bindings. When importing OrbitInfer as a library without its source, implement the -[`orbitinfer.recipe.Recipe`](../../orbitinfer/recipe.py) protocol in Python instead; a Rust +[`orbitinfer.recipe.Recipe`](../../../orbitinfer/recipe.py) protocol in Python instead; a Rust change is not required. Keep formatting and parsing in the recipe rather than duplicating them inside a deployment script. Before implementation, read the relevant parts of: -- [Runner authoring](../../docs/model-forward-pass.md): model preparation, +- [Runner authoring](../../../docs/model-forward-pass.md): model preparation, shapes, graphs, cache ownership, speculation, and validation. -- [CUDA package boundaries](../../models/cuda/ARCHITECTURE.md): where model and +- [CUDA package boundaries](../../../models/cuda/ARCHITECTURE.md): where model and hardware code belongs. -- [Distributed execution](../../docs/distributed.md), when using several GPUs. +- [Distributed execution](../../../docs/distributed.md), when using several GPUs. ## 2. Implement and qualify the forward-pass Runner @@ -50,7 +50,7 @@ Implement the reference-correct forward path and checkpoint loading in the ownin model package, importing or implementing the best suitable kernels. Keep cache allocation, scheduling, prefix ownership and request lifetime in the Engine. Use existing runtime staging, graph, sampling and distributed machinery. Extend the -shared [model registry](../../deploy/models.py) for hardware presets and runtime +shared [model registry](../../../deploy/models.py) for hardware presets and runtime construction; do not create a per-model deployment file or a new Modal probe suite. Compare logits and greedy tokens against an independent trusted reference before @@ -87,14 +87,14 @@ requirements, including imported upstream kernels: is not proof. If tooling lacks this evidence, add reusable instrumentation or leave that qualification explicitly unresolved. -Read [CUDA graph contracts](../../docs/cuda-graphs.md) and -[kernel artifact caching](../../docs/kernel-cache.md) for the relevant invariants. +Read [CUDA graph contracts](../../../docs/cuda-graphs.md) and +[kernel artifact caching](../../../docs/kernel-cache.md) for the relevant invariants. ## 3. Implement fast, graphed and overlapped speculative decoding Use the research to select the supported speculator and implement draft generation, target verification, acceptance, commit and rollback inside the runner. Reuse -[`orbitinfer.backend.speculative`](../../orbitinfer/backend/speculative.py) and the shared +[`orbitinfer.backend.speculative`](../../../orbitinfer/backend/speculative.py) and the shared CUDA completion/overlap machinery; do not introduce a separate scheduler. Keep the GPU draft/verify path graphable and prepare its buffers before capture. @@ -127,7 +127,7 @@ claiming the full integration is qualified. Place the runner in the Engine and start its native HTTP server on the agreed GPU host. This can be a temporary test server; a public Modal deployment is not needed -for validation. Use [`scripts/check_server.py`](../../scripts/check_server.py) +for validation. Use [`scripts/check_server.py`](../../../scripts/check_server.py) and the normal runtime tests, rather than adding model-specific diagnostic CLIs. Run ordinary JSON/streaming requests, token usage/logprobs, concurrency, context @@ -163,9 +163,9 @@ runner-level evidence from step 2 to close gaps in HTTP observability. Use the [performance optimization skill](../optimize-orbitinfer-performance/SKILL.md) for the profiling and iteration loop, including lessons from earlier OrbitInfer work. -Use [`scripts/bench.py`](../../scripts/bench.py), backed by `orbitinfer.bench`, and -[`scripts/plot_bench.py`](../../scripts/plot_bench.py). Read the -[benchmark guide](../../docs/benchmarks.md) for timing and topology semantics. +Use [`scripts/bench.py`](../../../scripts/bench.py), backed by `orbitinfer.bench`, and +[`scripts/plot_bench.py`](../../../scripts/plot_bench.py). Read the +[benchmark guide](../../../docs/benchmarks.md) for timing and topology semantics. Start with an 8K curve and traces on the target GPU configuration: ```bash @@ -213,7 +213,7 @@ commands, natural acceptance, baseline comparisons, and unresolved limitations. ## 6. Deploy to Modal if requested -Use [`deploy/modal_serve.py`](../../deploy/modal_serve.py) with `ORBITINFER_MODEL` +Use [`deploy/modal_serve.py`](../../../deploy/modal_serve.py) with `ORBITINFER_MODEL` selecting the shared registry entry. Keep checkpoint download/publication separate from GPU startup; add configuration to the registry instead of a new deployment script. When deployment is already authorized, proceed after qualification without diff --git a/skills/add-orbitinfer-model/agents/openai.yaml b/.agents/skills/add-orbitinfer-model/agents/openai.yaml similarity index 100% rename from skills/add-orbitinfer-model/agents/openai.yaml rename to .agents/skills/add-orbitinfer-model/agents/openai.yaml diff --git a/skills/optimize-orbitinfer-performance/SKILL.md b/.agents/skills/optimize-orbitinfer-performance/SKILL.md similarity index 96% rename from skills/optimize-orbitinfer-performance/SKILL.md rename to .agents/skills/optimize-orbitinfer-performance/SKILL.md index 2abb791..70976fb 100644 --- a/skills/optimize-orbitinfer-performance/SKILL.md +++ b/.agents/skills/optimize-orbitinfer-performance/SKILL.md @@ -28,10 +28,10 @@ Include CPU quota/affinity, NUMA placement and NCCL settings when investigating host or communication behavior. Prefer paired A/B on the same allocation; if the machine or environment changes, label the comparison accordingly. -Use [`scripts/bench.py`](../../scripts/bench.py) and -[`orbitinfer.bench`](../../orbitinfer/bench.py); extend shared tooling when necessary instead +Use [`scripts/bench.py`](../../../scripts/bench.py) and +[`orbitinfer.bench`](../../../orbitinfer/bench.py); extend shared tooling when necessary instead of adding a model-specific Modal diagnostic file. Consult `--help` and -[the timing protocol](../../docs/benchmarks.md). For example, on an existing GPU +[the timing protocol](../../../docs/benchmarks.md). For example, on an existing GPU host with `MODEL_KEY` and `CHECKPOINT` set: ```bash @@ -128,7 +128,7 @@ instead of lists, scalar boxing and repeated validation. Send immutable request media/sampling once and typed mutable step data thereafter. A metadata cache needs explicit invalidation on changed inputs, layout or lifecycle. Keep Engine-owned allocation, leases and commit in the Engine; model-specific attention layout and -replay policy belong in the runner. See [the boundary](../../docs/kernel-boundary.md). +replay policy belong in the runner. See [the boundary](../../../docs/kernel-boundary.md). **Keep context and sampling out of JIT keys.** Logical history lengths, page-table widths, positions, temperature, top-k/top-p and seeds are runtime data. Audit @@ -143,7 +143,7 @@ which rank compiled. If artifacts do not expose compilation reliably, instrument the compiler/cache-miss entry point or leave that claim unresolved. Warm repeat TTFT, mtime-only comparisons, and an exhaustive warmup grid do not establish this. Also verify numerical parity on unseen lengths; see -[kernel caching](../../docs/kernel-cache.md). +[kernel caching](../../../docs/kernel-cache.md). **Graph useful regions without changing the computation.** GPU kernel paths must be graphable, but the selected full/breakable policy determines actual capture @@ -157,7 +157,7 @@ parity and padding cost across the supported buckets. Keep graph inputs/outputs at stable addresses, update mutable metadata, and consume or copy shared output views before replay overwrites them. Shared prefill-region buffers execute serially; overlapping streams cannot replay them concurrently. -See [graph policy and ownership](../../docs/cuda-graphs.md). +See [graph policy and ownership](../../../docs/cuda-graphs.md). **Separate matched-batch parity from cross-batch variation.** Compare target-only and speculative outputs with matched physical batch buckets, prompts, sampling, @@ -178,7 +178,7 @@ across boundaries. Preserve collective participation, rank agreement on readines execution/commit failures, and quarantine after uncertain mutation. Stable decode may reduce transport work, not erase those agreements. Some persistent kernels need the whole GPU: obey measured stream dependencies instead of overlapping every -NCCL operation. See [distributed execution](../../docs/distributed.md). +NCCL operation. See [distributed execution](../../../docs/distributed.md). **Fuse with the real numerical contract.** Preserve accumulator precision, packed scale/weight strides, sentinel ordering, masks, recurrent state and rollback. @@ -200,7 +200,7 @@ counters during timing; retain detailed diagnostics outside it. GPU event storag must survive until its corresponding completion, rather than being overwritten by the next replay. Metrics collection/scraping must not synchronize or introduce collectives. Enable capture-time instrumentation before capture and quantify its -overhead when interpreting small differences. See [metrics](../../docs/metrics.md). +overhead when interpreting small differences. See [metrics](../../../docs/metrics.md). ## Qualify the change and close the loop @@ -221,7 +221,7 @@ Do not loosen tolerances to hide graph padding or parallel reduction drift. Approximate algorithms need an explicit quality comparison and must remain labeled as approximate; they cannot establish exact full-compute parity. -For serving changes, use [`scripts/check_server.py`](../../scripts/check_server.py) +For serving changes, use [`scripts/check_server.py`](../../../scripts/check_server.py) against the test endpoint for prefix reuse, varying-context TTFT, concurrent requests, media/reasoning where supported, stress/recovery and the relevant MMLU subset. The checker's prefix test verifies telemetry; separately compare diff --git a/skills/optimize-orbitinfer-performance/agents/openai.yaml b/.agents/skills/optimize-orbitinfer-performance/agents/openai.yaml similarity index 100% rename from skills/optimize-orbitinfer-performance/agents/openai.yaml rename to .agents/skills/optimize-orbitinfer-performance/agents/openai.yaml diff --git a/skills/optimize-orbitinfer-performance/references/lessons.md b/.agents/skills/optimize-orbitinfer-performance/references/lessons.md similarity index 94% rename from skills/optimize-orbitinfer-performance/references/lessons.md rename to .agents/skills/optimize-orbitinfer-performance/references/lessons.md index 880d0af..c9accff 100644 --- a/skills/optimize-orbitinfer-performance/references/lessons.md +++ b/.agents/skills/optimize-orbitinfer-performance/references/lessons.md @@ -46,8 +46,8 @@ Engine overlap; `4c38de5` changes control transport. Detailed graph statistics cost about 2.4 ms/step in a local diagnostic versus about 0.004 ms for aggregate counters; that is harness overhead, not faster model math. -Read [recorded V4.1 measurements](../../../docs/benchmarks.md#recorded-v41-measurements) -and [distributed contracts](../../../docs/distributed.md). First inspect repeated +Read [recorded V4.1 measurements](../../../../docs/benchmarks.md#recorded-v41-measurements) +and [distributed contracts](../../../../docs/distributed.md). First inspect repeated work and transport between graph bodies when device work is already close to a reference. Preserve exact catalog revalidation and readiness/commit agreement. @@ -65,9 +65,9 @@ macOS host against `4775fd3`: These are CPU microbenchmarks, not measured decode improvements. Preserve ownership, invalidation, eviction and retirement checks while changing their complexity. -Current anchors: [paged resources](../../../crates/orbitinfer-core/src/resources/paged.rs), -[prefix resources](../../../crates/orbitinfer-core/src/resources/prefix.rs) and -[scheduler](../../../docs/scheduler.md). +Current anchors: [paged resources](../../../../crates/orbitinfer-core/src/resources/paged.rs), +[prefix resources](../../../../crates/orbitinfer-core/src/resources/prefix.rs) and +[scheduler](../../../../docs/scheduler.md). ## 3. Repeated preparation and object conversion accumulate @@ -86,7 +86,7 @@ Avoid lists, per-token Python scalars, repeated hashing and copies at a boundary that already validated the data. Retain ownership until asynchronous copies and receives finish. `40a44fd` corrected reusable buffers that became inference tensors; `9ffe6f3` retains pinned storage through asynchronous copies. See -[the runner boundary](../../../docs/kernel-boundary.md). +[the runner boundary](../../../../docs/kernel-boundary.md). ## 4. Runtime geometry prevents first-use compilation stalls @@ -100,7 +100,7 @@ compiled-kernel counts remained unchanged. The full-model EP8 128K prefill experiments measured 9.774 s with 1K chunks, 4.009 s with 4K and 3.691 s with 8K. This combines execution/configuration changes; it is not an isolated speedup from disabling compilation. See -[benchmark evidence](../../../docs/benchmarks.md#recorded-v41-measurements). +[benchmark evidence](../../../../docs/benchmarks.md#recorded-v41-measurements). Older precedents: `31ae8d8` keeps only model-fixed trailing cuTile dimensions static; `76f8665` separates variable prefill from tiny decode convolution; @@ -122,7 +122,7 @@ GEMM geometry. GPT-OSS's separate output-projection capture cost 24–30 µs ver operations. DeepSeek exact buckets avoid a demonstrated 17 → 32 row parity failure. These are region timings, not whole-model speedups. Read -[graph policy](../../../docs/cuda-graphs.md), including serial shared-workspace use. +[graph policy](../../../../docs/cuda-graphs.md), including serial shared-workspace use. Maintain graphability, report actual capture scope, and distinguish intentional prefill partial-chunk fallback from a missing capture. Do not loosen numerical checks simply to improve a replay counter. @@ -147,7 +147,7 @@ does not validate packing, scale axes, strides or accumulation precision. Recent positive primitive cases (`41ad813`, `5bd97f0`, `d2a4fc3`) include GPT-OSS router 14.39 → 4.12 µs at eight rows and Inkling 4K convolution 496.74 → 122.11 µs. They remain primitive claims until a matched end-to-end run demonstrates impact. -Read [primitive checks](../../../docs/benchmarks.md#cuda-model-primitive-checks). +Read [primitive checks](../../../../docs/benchmarks.md#cuda-model-primitive-checks). ## 7. Overlap requires ownership and device feedback, not just an async API @@ -174,8 +174,8 @@ read window plus one. `8335f67` adds speculative pre-anchor history. Recompute t complete read/write envelope when changing chunk size, draft width or capacity. Graph padding needs reserved scratch storage; masks alone did not prevent the unscheduled-row corruption fixed by `c79fa23`. See -[the scheduler](../../../docs/scheduler.md) and -[model correctness contract](../../../docs/model-forward-pass.md). +[the scheduler](../../../../docs/scheduler.md) and +[model correctness contract](../../../../docs/model-forward-pass.md). ## 8. Boundary tests expose bugs that ordinary generation misses @@ -204,7 +204,7 @@ This does not establish equality to full prefill. Approximate consumer state sta out of prefix caching and prompt scoring takes the full path. Separate deterministic qualification compared compact/masked replay and prefix restoration, including unequal/idle cohorts; that does not turn the algorithm into exact full computation. -See [the model's validation notes](../../../models/cuda/orbitinfer_models_cuda/deepseek_v41/README.md#validation). +See [the model's validation notes](../../../../models/cuda/orbitinfer_models_cuda/deepseek_v41/README.md#validation). Inspect a reference's actual work before tuning toward its published curve. Larger chunks alone cannot reproduce this algorithm, and consumer-sized graph @@ -226,7 +226,7 @@ cover actual head/channel/state geometry and collective order, not only flags. Historical GPT-OSS 120B H100 sweeps at `b2f7c3c` measured best tokens/s/GPU of 574.750 for TP1, 488.915 TP2, 286.481 TP4, 988.479 EP2 and 1,013.632 EP4. These workload-specific curves argue for measuring the mesh; they are not a general -ranking. See [recorded decode sweeps](../../../docs/benchmarks.md#recorded-decode-sweeps). +ranking. See [recorded decode sweeps](../../../../docs/benchmarks.md#recorded-decode-sweeps). ## 11. Communication overlap and NUMA need controlled evidence @@ -246,7 +246,7 @@ topology/configuration, not a complete causal diagnosis. Local evidence is under `artifacts/benchmarks/2026-10-24-deepseek-v41-b300-ep8-numa/`. -The API commit `3f2893b` and [NUMA guide](../../../docs/numa.md) explain placement +The API commit `3f2893b` and [NUMA guide](../../../../docs/numa.md) explain placement before child startup, container CPU masks/quotas and fallback reporting. Affinity does not reserve CPUs or migrate existing pages. Do not claim that API presence establishes a measured performance improvement. @@ -269,7 +269,7 @@ was a candidate opportunity, not a completed speedup. Those sources used top-k-before-top-p (`580dde5`, `b2f7c3c`). The later fused sampler deliberately changed to top-p on the full scaled distribution followed by the top-k cap, retaining cutoff ties, and CDF sampling changed seeded stochastic choices. -See the current [sampling contract](../../../docs/benchmarks.md#shared-sampler-checks) +See the current [sampling contract](../../../../docs/benchmarks.md#shared-sampler-checks) before choosing an oracle; a faster upstream sampler with another contract is not a drop-in replacement. diff --git a/AGENTS.md b/AGENTS.md index cdd95aa..b93861e 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,9 +1,10 @@ # OrbitInfer agent guide Read docs/architecture.md and docs/completion-plan.md before making changes. -This repository is an independent full-source adaptation of the snapshot in -third_party/upstream/import.json. Product names, modules and tools use OrbitInfer. -Preserve third-party copyright, licenses, notices and the source manifest. +Product names, modules and tools use OrbitInfer. Preserve third-party copyright, +licenses and notices; source attribution lives in NOTICE and package notices. +External projects are reviewed manually as needed, with no vendored source +manifest or automatic synchronization. Project skills live in .agents/skills/. ## Owners @@ -37,8 +38,8 @@ Use the pinned Rust toolchain and Python 3.12 development environment. Run `uv run --no-sync python scripts/check.py`. On Apple Silicon, `uv sync --locked` also builds the MLX extension and installs all model workspace packages. The script runs formatting, Clippy, Rust tests, Python tests, the installed CPU -example and source/package boundary controls. GPU skips are reported and never -counted as hardware evidence. Write installed gates before claiming GPU support. +runtime check and source/package boundary controls. GPU skips are reported and +never counted as hardware evidence. Write installed gates before claiming GPU support. Keep changes with their owner. Record source-import adaptations separately from behavior changes. Tests should force lifecycle interleavings with events diff --git a/NOTICE b/NOTICE index 3eccd8c..a6764ca 100644 --- a/NOTICE +++ b/NOTICE @@ -6,7 +6,8 @@ https://github.com/ekzhang/resin Files are renamed and adapted as described in docs/source-import.md. Original attribution follows unchanged; its paths refer to the upstream tree. -The local paths are mapped in third_party/upstream/import.json. +Local Python and crate prefixes use orbitinfer and orbitinfer_models_*; +the original attribution text below is preserved. Resin diff --git a/README.md b/README.md index d30ab03..4472590 100644 --- a/README.md +++ b/README.md @@ -20,11 +20,11 @@ OrbitInfer also has optional HTTP serving and Prometheus metrics. > inference engine, please use SGLang or vLLM. Read on if you want to hack on > LLM inference systems, new models or hardware. -The complete source import includes CUDA model runners, MLX/Metal runners, +The repository includes CUDA model runners, MLX/Metal runners, Torch reference models, serving tools, benchmarks and tests. The first planned installed target is DeepSeek-V4.1-Flash on SM100. No complete GPU model profile has been qualified for OrbitInfer. See the [plan](docs/completion-plan.md) and -[source import record](docs/source-import.md). Interface changes are recorded in +[source and license notes](docs/source-import.md). Interface changes are recorded in the [change notes](docs/releases.md). ## Features @@ -50,9 +50,9 @@ Start with the two-page [architecture overview](docs/architecture.md). Either fork the repository or import `orbitinfer` and `orbitinfer_models_cuda` as standalone packages. Then use the following skills: -- [**add-orbitinfer-model**](skills/add-orbitinfer-model/SKILL.md) - +- [**add-orbitinfer-model**](.agents/skills/add-orbitinfer-model/SKILL.md) - Create an inference engine for a new model from scratch, using OrbitInfer. -- [**optimize-orbitinfer-performance**](skills/optimize-orbitinfer-performance/SKILL.md) - +- [**optimize-orbitinfer-performance**](.agents/skills/optimize-orbitinfer-performance/SKILL.md) - Iterate on your inference engine to improve performance with an "autoresearch" workflow, diagnose bottlenecks and validate improvements, drawing on past learnings. diff --git a/docs/architecture-plan.md b/docs/architecture-plan.md deleted file mode 100644 index e8a3a91..0000000 --- a/docs/architecture-plan.md +++ /dev/null @@ -1,5 +0,0 @@ -# Architecture plan - -See [architecture](architecture.md) and the [completion plan](completion-plan.md) -for the current full-source scope. The earlier NVIDIA-only evaluation is retained -as a [historical source audit](../third_party/upstream/source-audit.md). diff --git a/docs/architecture.md b/docs/architecture.md index 872fa4e..a620795 100644 --- a/docs/architecture.md +++ b/docs/architecture.md @@ -68,7 +68,7 @@ execute straightforwardly. Separately installable model packages already use this boundary to add models and select kernels. The [model authoring reference](model-forward-pass.md) documents preparation and execution requirements. The -[add-orbitinfer-model skill](../skills/add-orbitinfer-model/SKILL.md) provides the workflow +[add-orbitinfer-model skill](../.agents/skills/add-orbitinfer-model/SKILL.md) provides the workflow for implementing and validating a runner. ## Lifecycle diff --git a/docs/build.md b/docs/build.md index 1a0e13c..6aa409f 100644 --- a/docs/build.md +++ b/docs/build.md @@ -1,6 +1,6 @@ # Build and CPU checks -The source import uses Rust 1.96 and Python 3.12. The core has no Python or GPU +Development uses Rust 1.96 and Python 3.12. The core has no Python or GPU runtime dependency; the binding is abi3 for Python 3.11 and newer. That ABI choice is inherited, not a claim that all Python/GPU combinations have been tested. @@ -11,7 +11,7 @@ uv run --no-sync python scripts/check.py The test group installs CPU-capable Torch for tensor/layout controls. It does not resolve or install the separate CUDA model package. Importing `orbitinfer` must not -import Torch or load an accelerator library. `examples/control_plane.py` runs +import Torch or load an accelerator library. `tests/installed/check_runtime.py` runs through the compiled binding with a deterministic test runner and checked tokens. Individual commands: diff --git a/docs/completion-plan.md b/docs/completion-plan.md index fcee61b..fff827e 100644 --- a/docs/completion-plan.md +++ b/docs/completion-plan.md @@ -2,11 +2,12 @@ ## Current focus -The complete pinned source tree is imported under OrbitInfer names: all 616 -files, including CUDA, MLX/Metal and Torch model packages, tests, benchmarks, -deployment tools and authoring skills. Local validation is recorded in the -[import evidence](source-import.md). A complete source import does not establish -GPU model correctness, performance or production readiness. +OrbitInfer maintains its runtime, CUDA, MLX/Metal and Torch model packages, +tests, benchmarks and deployment tools independently. Agent skills live in +`.agents/skills/`; external updates are reviewed manually as needed. Attribution +and the initial validation record remain in [source notes](source-import.md). +Source availability does not establish GPU model correctness, performance or +production readiness. Follow the [hierarchical external KV design](external-kv-design.md): one Engine/page allocator, one selected provider and separate outcome/retirement @@ -31,7 +32,7 @@ SM100. Other imported runners are available for separate validation. | Phase | Scope | State | | --- | --- | --- | -| Foundation | Complete runtime, all model packages, tooling, tests and provenance | Source imported; validation in source-import.md; clean CPU gate repairs included | +| Foundation | Runtime, model packages, tooling, tests and attribution | Maintained independently; installed-wheel CPU check retained; no source-mirror manifest | | P0 | Final-reference/failed-completion lifetime audit, existing storage integration, real backing, identity and byte bounds | Implemented at source level; real backing, identity, byte and metadata bounds verified on CPU; installed CUDA gate unrun | | P1 | Full-checkpoint offline DeepSeek numerical comparison on SM100 | Unrun; hardware required | | P2 | CUDA graphs, overlap, persistent-state faults and cancellation | Source imported; installed profile unqualified | diff --git a/docs/releases.md b/docs/releases.md index fe70141..5f1bd66 100644 --- a/docs/releases.md +++ b/docs/releases.md @@ -2,6 +2,11 @@ ## Unreleased +- Project skills move to `.agents/skills/`. The counting example becomes an + installed-wheel test. The source-mirror manifest and superseded audit are + removed; licenses and attribution remain in NOTICE and package notices. + External changes are reviewed manually, not synchronized automatically. + - External storage now requires `CacheIdentity` and `prefix_cache_max_bytes` at Engine construction. Keys include weights, format, dtype, semantics, shard and cache geometry; older keys intentionally miss. Invalid/overlapping exports and diff --git a/docs/source-import.md b/docs/source-import.md index ac8f92e..a071445 100644 --- a/docs/source-import.md +++ b/docs/source-import.md @@ -1,59 +1,25 @@ -# Source import +# Source and license notes -OrbitInfer adapts the complete source snapshot of -[Resin at a1989594](https://github.com/ekzhang/resin/tree/a1989594d81921be3bffe43804eb0550b216ca64), -checked on 2026-11-14 and imported on 2026-11-15. The source archive SHA-256 is -`92464bd55a1b7c5f49291742b4dbaa365a52e78c6c4ca8000c638f9d4191733d`. -The [manifest](../third_party/upstream/import.json) maps all 616 upstream files -to their local paths and records original and namespace-transformed hashes. -There are no excluded source paths. The previous subset selected 342 files; -the complete import restores the other 274 files and previously trimmed tests. +OrbitInfer is maintained independently. The original source revision and +third-party attribution are recorded in [NOTICE](../NOTICE); source and model +packages retain their applicable licenses and notices. The original file +inventory remains available in Git history and is not a maintenance contract. -## Scope and naming +Product namespaces and package imports use OrbitInfer. Renamed and adapted +files retain their change notices. The core, CUDA, MLX/Metal and Torch packages +remain separate; no external inference runtime is required at import time. -All four Rust crates, the Python binding, all shared runtime helpers, CUDA, -MLX/Metal and Torch model packages, tests, fixtures, docs, deployment scripts, -benchmarks, workflows and authoring skills are present. No model weights or -upstream Git history are included. The source snapshot, rather than a moving -branch, defines this import. +External changes are inspected manually when relevant. Record the exact +revision and tests in the change review; there is no automatic synchronization. +Review ownership, cache identity and resource bounds before adopting a change. +Existing GPU benchmark tables from the initial source are historical reference +measurements, not OrbitInfer qualification. -Product names, Python imports, Rust crates, environment names, metrics, -protocol fields, kernel symbols, cache paths and tool names use OrbitInfer. -The native Python module is `orbitinfer._native`; there is no upstream runtime -package or compatibility alias. Original project names remain only in legal -notices, provenance records and checks which reject old namespaces. +## Initial validation record (2026-11-15) -Original licenses, copyright and NOTICE text are retained. Source and model -packages include their applicable notices. Upstream benchmark numbers in -imported guides are historical reference measurements, not OrbitInfer results. -See the [historical audit](../third_party/upstream/source-audit.md) for the -original NVIDIA-only evaluation; the full-source scope supersedes its exclusions. - -## Adaptations - -- Rename project identifiers consistently in Python, Rust, C++, CUDA, Metal, - build metadata, templates and fixtures; retain algorithms and ownership. -- Keep the existing equivalent Clippy boolean simplifications, Rust formatting, - and removal of an obsolete PyO3 lint expectation for the pinned Rust toolchain. -- Keep explicit Python initialization in the Rust NumPy-borrow control, and the - CPython-version-specific dataclass exception check. No assertion is removed. -- Restore the complete upstream Python workspace and all model/test entrypoints. - Retain a CPU-only test dependency group for machines without native GPU SDKs. -- Use pytest's importlib mode so CUDA and MLX tests with the same basename can - run together. The MLX ordering control explicitly skips when MLX is absent, - matching the other MLX controls; the Apple workflow requires Metal and cannot - silently accept missing MLX coverage. -- Keep local development versions and `publish = false` for Rust crates. There - is no remote repository or package publication in this step. -- Preserve the deterministic CPU example and isolated installed-wheel check. - CPU distributed controls select the local loopback interface. -- Replace the upstream README's performance promotion with the current - qualification limits. Existing owner documentation remains in place. - -## Validation and limits - -Local host: macOS arm64, Python 3.12.6, Rust 1.96.0. All 616 original file -hashes match the pinned archive and every mapped local file is present. +At the initial import: macOS arm64, Python 3.12.6, Rust 1.96.0. The complete +source snapshot and its original file hashes were verified at that time. This +historical evidence does not require retaining a source mirror or manifest. | Check | Result | | --- | --- | @@ -72,7 +38,6 @@ two Linux checks and 25 checks requiring uncached reference processors or checkpoints. No skip is counted as hardware evidence. The missing local Metal compiler component was installed to compile and execute the MLX controls. -No complete GPU model has been run, no deployment tool has been executed -against a cloud account, and no package has been published. Linux and remote CI -are unrun. The [completion plan](completion-plan.md) separates this source -import from installed model validation and future OrbitKV/OrbiTE integration. +At that checkpoint, no complete GPU model or cloud deployment had run, no +package had been published, and Linux/remote CI were unrun. Current status is +in the [completion plan](completion-plan.md); this table is historical evidence. diff --git a/models/cuda/NOTICE b/models/cuda/NOTICE index 3eccd8c..a6764ca 100644 --- a/models/cuda/NOTICE +++ b/models/cuda/NOTICE @@ -6,7 +6,8 @@ https://github.com/ekzhang/resin Files are renamed and adapted as described in docs/source-import.md. Original attribution follows unchanged; its paths refer to the upstream tree. -The local paths are mapped in third_party/upstream/import.json. +Local Python and crate prefixes use orbitinfer and orbitinfer_models_*; +the original attribution text below is preserved. Resin diff --git a/models/mlx/NOTICE b/models/mlx/NOTICE index 3eccd8c..a6764ca 100644 --- a/models/mlx/NOTICE +++ b/models/mlx/NOTICE @@ -6,7 +6,8 @@ https://github.com/ekzhang/resin Files are renamed and adapted as described in docs/source-import.md. Original attribution follows unchanged; its paths refer to the upstream tree. -The local paths are mapped in third_party/upstream/import.json. +Local Python and crate prefixes use orbitinfer and orbitinfer_models_*; +the original attribution text below is preserved. Resin diff --git a/models/torch/NOTICE b/models/torch/NOTICE index 3eccd8c..a6764ca 100644 --- a/models/torch/NOTICE +++ b/models/torch/NOTICE @@ -6,7 +6,8 @@ https://github.com/ekzhang/resin Files are renamed and adapted as described in docs/source-import.md. Original attribution follows unchanged; its paths refer to the upstream tree. -The local paths are mapped in third_party/upstream/import.json. +Local Python and crate prefixes use orbitinfer and orbitinfer_models_*; +the original attribution text below is preserved. Resin diff --git a/pyproject.toml b/pyproject.toml index 543ac96..b0306f8 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -10,7 +10,7 @@ description = "A small Rust substrate for model-specific inference engines" readme = "README.md" requires-python = ">=3.11" license = "Apache-2.0" -license-files = ["LICENSE", "NOTICE", "third_party/upstream/NOTICE", "crates/orbitinfer-recipe/licenses/*.txt"] +license-files = ["LICENSE", "NOTICE", "crates/orbitinfer-recipe/licenses/*.txt"] classifiers = [ "Development Status :: 3 - Alpha", "Programming Language :: Python :: 3", @@ -74,7 +74,6 @@ python-packages = ["orbitinfer"] features = ["pyo3/extension-module"] include = [ { path = "orbitinfer/backend/cuda/sampling_topk.cu", format = ["sdist", "wheel"] }, - { path = "third_party/upstream/import.json", format = "sdist" }, { path = "docs/source-import.md", format = "sdist" }, ] diff --git a/scripts/check.py b/scripts/check.py index e2b8988..135baf3 100644 --- a/scripts/check.py +++ b/scripts/check.py @@ -1,4 +1,4 @@ -"""Run the source-import gates. No GPU model or weights are loaded.""" +"""Run the repository gates. No GPU model or weights are loaded.""" import os import subprocess @@ -54,14 +54,14 @@ def main(): run(sys.executable, "-m", "venv", out / "venv") python = out / "venv/bin/python" run("uv", "pip", "install", "--python", python, wheels[0]) - run(python, "-I", ROOT / "examples/control_plane.py", cwd=out) + run(python, "-I", ROOT / "tests/installed/check_runtime.py", cwd=out) run( python, "-I", "-c", "import sys, orbitinfer; " "from orbitinfer.integrations.orbitkv import OrbitKVCacheStorage; " - "imported = {'torch', 'orbitkv', 'resin', 'flashinfer', 'mlx'} & sys.modules.keys(); " + "imported = {'torch', 'orbitkv', 'flashinfer', 'mlx'} & sys.modules.keys(); " "wrong = orbitinfer.Engine.__module__ != 'orbitinfer._native'; " "sys.exit(f'unexpected import: {imported}, binding={wrong}' if imported or wrong else 0)", cwd=out, diff --git a/scripts/check_source.py b/scripts/check_source.py index fbbd798..05e936b 100644 --- a/scripts/check_source.py +++ b/scripts/check_source.py @@ -1,9 +1,6 @@ -"""Check complete source coverage, attribution and project namespaces.""" +"""Check first-party package boundaries and retained license notices.""" import ast -import hashlib -import json -import re import subprocess import tomllib from pathlib import Path @@ -17,39 +14,15 @@ def require(condition, message): def main(): - manifest = json.loads((ROOT / "third_party/upstream/import.json").read_text()) + project = tomllib.loads((ROOT / "pyproject.toml").read_text())["project"] + require(project["name"] == "orbitinfer", "unexpected core package name") + for pattern in project["license-files"]: + require(any(ROOT.glob(pattern)), f"missing packaged notice: {pattern}") + license_text = (ROOT / "LICENSE").read_bytes() + notice = (ROOT / "NOTICE").read_bytes() require( - manifest["revision"] == "a1989594d81921be3bffe43804eb0550b216ca64", - "unexpected source revision", + bool(license_text.strip()) and bool(notice.strip()), "empty license or notice" ) - require( - not manifest["excluded_source_paths"], - "the full snapshot must not exclude files", - ) - paths = [entry["path"] for entry in manifest["files"]] - require( - len(paths) == len(set(paths)) == 616, "incomplete or duplicate source inventory" - ) - for entry in manifest["files"]: - require((ROOT / entry["path"]).is_file(), f"missing import: {entry['path']}") - require( - re.fullmatch("[0-9a-f]{64}", entry["source_sha256"]), "invalid source hash" - ) - license_hash = hashlib.sha256((ROOT / "LICENSE").read_bytes()).digest() - require( - license_hash - == hashlib.sha256( - (ROOT / "third_party/upstream/LICENSE").read_bytes() - ).digest(), - "upstream license changed", - ) - forbidden = {"resin", "resin_models_cuda", "resin_models_mlx", "resin_models_torch"} - allowed = { - "NOTICE", - "docs/source-import.md", - "scripts/check_source.py", - "scripts/check.py", - } names = ( subprocess.check_output( ["git", "ls-files", "-z", "--cached", "--others", "--exclude-standard"], @@ -60,15 +33,8 @@ def main(): ) for name in sorted(set(names) - {""}): path = ROOT / name - if not path.is_file() or name.startswith("third_party/"): + if not path.is_file(): continue - require("resin" not in name.lower(), f"old project path: {name}") - if name not in allowed and "LICENSE" not in path.name and path.name != "NOTICE": - try: - text = path.read_text() - except UnicodeDecodeError: - continue - require("resin" not in text.lower(), f"old project namespace in {name}") if path.suffix == ".py": for node in ast.walk(ast.parse(path.read_text(), filename=name)): modules = [] @@ -77,10 +43,6 @@ def main(): elif isinstance(node, ast.ImportFrom) and node.module: modules = [node.module] for module in modules: - require( - module.split(".")[0] not in forbidden, - f"upstream runtime import {module} in {name}", - ) for backend in ("cuda", "mlx", "torch"): prefix = f"orbitinfer_models_{backend}" if module == prefix or module.startswith(prefix + "."): @@ -93,18 +55,28 @@ def main(): ) for path in (ROOT / "crates").glob("*/Cargo.toml"): config = tomllib.loads(path.read_text()) + require( + config["package"]["name"].startswith("orbitinfer-"), + f"unexpected crate name in {path}", + ) require( config["package"].get("publish") is False, f"publication is not disabled in {path}", ) for backend in ("cuda", "mlx", "torch"): + package = ROOT / "models" / backend + config = tomllib.loads((package / "pyproject.toml").read_text()) + require( + config["project"]["name"] == f"orbitinfer-models-{backend}", + f"unexpected {backend} package name", + ) + require((package / "LICENSE").read_bytes() == license_text, "license changed") + require((package / "NOTICE").read_bytes() == notice, "package notice differs") require( (ROOT / "models" / backend / f"orbitinfer_models_{backend}").is_dir(), f"missing {backend} runner package", ) - print( - f"Source boundary passed: all {len(paths)} files; no exclusions or old runtime namespaces; licenses retained." - ) + print("Source boundary passed: package names, model modules and license notices.") if __name__ == "__main__": diff --git a/examples/control_plane.py b/tests/installed/check_runtime.py similarity index 93% rename from examples/control_plane.py rename to tests/installed/check_runtime.py index b82a209..d73374e 100644 --- a/examples/control_plane.py +++ b/tests/installed/check_runtime.py @@ -1,4 +1,4 @@ -"""CPU control-plane example with checked tokens; this is not a language model.""" +"""Installed-wheel CPU runtime check; this is not a language model.""" import numpy as np diff --git a/third_party/upstream/LICENSE b/third_party/upstream/LICENSE deleted file mode 100644 index d645695..0000000 --- a/third_party/upstream/LICENSE +++ /dev/null @@ -1,202 +0,0 @@ - - Apache License - Version 2.0, January 2004 - http://www.apache.org/licenses/ - - TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION - - 1. Definitions. - - "License" shall mean the terms and conditions for use, reproduction, - and distribution as defined by Sections 1 through 9 of this document. - - "Licensor" shall mean the copyright owner or entity authorized by - the copyright owner that is granting the License. - - "Legal Entity" shall mean the union of the acting entity and all - other entities that control, are controlled by, or are under common - control with that entity. For the purposes of this definition, - "control" means (i) the power, direct or indirect, to cause the - direction or management of such entity, whether by contract or - otherwise, or (ii) ownership of fifty percent (50%) or more of the - outstanding shares, or (iii) beneficial ownership of such entity. - - "You" (or "Your") shall mean an individual or Legal Entity - exercising permissions granted by this License. - - "Source" form shall mean the preferred form for making modifications, - including but not limited to software source code, documentation - source, and configuration files. - - "Object" form shall mean any form resulting from mechanical - transformation or translation of a Source form, including but - not limited to compiled object code, generated documentation, - and conversions to other media types. - - "Work" shall mean the work of authorship, whether in Source or - Object form, made available under the License, as indicated by a - copyright notice that is included in or attached to the work - (an example is provided in the Appendix below). - - "Derivative Works" shall mean any work, whether in Source or Object - form, that is based on (or derived from) the Work and for which the - editorial revisions, annotations, elaborations, or other modifications - represent, as a whole, an original work of authorship. For the purposes - of this License, Derivative Works shall not include works that remain - separable from, or merely link (or bind by name) to the interfaces of, - the Work and Derivative Works thereof. - - "Contribution" shall mean any work of authorship, including - the original version of the Work and any modifications or additions - to that Work or Derivative Works thereof, that is intentionally - submitted to Licensor for inclusion in the Work by the copyright owner - or by an individual or Legal Entity authorized to submit on behalf of - the copyright owner. For the purposes of this definition, "submitted" - means any form of electronic, verbal, or written communication sent - to the Licensor or its representatives, including but not limited to - communication on electronic mailing lists, source code control systems, - and issue tracking systems that are managed by, or on behalf of, the - Licensor for the purpose of discussing and improving the Work, but - excluding communication that is conspicuously marked or otherwise - designated in writing by the copyright owner as "Not a Contribution." - - "Contributor" shall mean Licensor and any individual or Legal Entity - on behalf of whom a Contribution has been received by Licensor and - subsequently incorporated within the Work. - - 2. Grant of Copyright License. Subject to the terms and conditions of - this License, each Contributor hereby grants to You a perpetual, - worldwide, non-exclusive, no-charge, royalty-free, irrevocable - copyright license to reproduce, prepare Derivative Works of, - publicly display, publicly perform, sublicense, and distribute the - Work and such Derivative Works in Source or Object form. - - 3. Grant of Patent License. Subject to the terms and conditions of - this License, each Contributor hereby grants to You a perpetual, - worldwide, non-exclusive, no-charge, royalty-free, irrevocable - (except as stated in this section) patent license to make, have made, - use, offer to sell, sell, import, and otherwise transfer the Work, - where such license applies only to those patent claims licensable - by such Contributor that are necessarily infringed by their - Contribution(s) alone or by combination of their Contribution(s) - with the Work to which such Contribution(s) was submitted. If You - institute patent litigation against any entity (including a - cross-claim or counterclaim in a lawsuit) alleging that the Work - or a Contribution incorporated within the Work constitutes direct - or contributory patent infringement, then any patent licenses - granted to You under this License for that Work shall terminate - as of the date such litigation is filed. - - 4. Redistribution. You may reproduce and distribute copies of the - Work or Derivative Works thereof in any medium, with or without - modifications, and in Source or Object form, provided that You - meet the following conditions: - - (a) You must give any other recipients of the Work or - Derivative Works a copy of this License; and - - (b) You must cause any modified files to carry prominent notices - stating that You changed the files; and - - (c) You must retain, in the Source form of any Derivative Works - that You distribute, all copyright, patent, trademark, and - attribution notices from the Source form of the Work, - excluding those notices that do not pertain to any part of - the Derivative Works; and - - (d) If the Work includes a "NOTICE" text file as part of its - distribution, then any Derivative Works that You distribute must - include a readable copy of the attribution notices contained - within such NOTICE file, excluding those notices that do not - pertain to any part of the Derivative Works, in at least one - of the following places: within a NOTICE text file distributed - as part of the Derivative Works; within the Source form or - documentation, if provided along with the Derivative Works; or, - within a display generated by the Derivative Works, if and - wherever such third-party notices normally appear. The contents - of the NOTICE file are for informational purposes only and - do not modify the License. You may add Your own attribution - notices within Derivative Works that You distribute, alongside - or as an addendum to the NOTICE text from the Work, provided - that such additional attribution notices cannot be construed - as modifying the License. - - You may add Your own copyright statement to Your modifications and - may provide additional or different license terms and conditions - for use, reproduction, or distribution of Your modifications, or - for any such Derivative Works as a whole, provided Your use, - reproduction, and distribution of the Work otherwise complies with - the conditions stated in this License. - - 5. Submission of Contributions. Unless You explicitly state otherwise, - any Contribution intentionally submitted for inclusion in the Work - by You to the Licensor shall be under the terms and conditions of - this License, without any additional terms or conditions. - Notwithstanding the above, nothing herein shall supersede or modify - the terms of any separate license agreement you may have executed - with Licensor regarding such Contributions. - - 6. Trademarks. This License does not grant permission to use the trade - names, trademarks, service marks, or product names of the Licensor, - except as required for reasonable and customary use in describing the - origin of the Work and reproducing the content of the NOTICE file. - - 7. Disclaimer of Warranty. Unless required by applicable law or - agreed to in writing, Licensor provides the Work (and each - Contributor provides its Contributions) on an "AS IS" BASIS, - WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or - implied, including, without limitation, any warranties or conditions - of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A - PARTICULAR PURPOSE. You are solely responsible for determining the - appropriateness of using or redistributing the Work and assume any - risks associated with Your exercise of permissions under this License. - - 8. Limitation of Liability. In no event and under no legal theory, - whether in tort (including negligence), contract, or otherwise, - unless required by applicable law (such as deliberate and grossly - negligent acts) or agreed to in writing, shall any Contributor be - liable to You for damages, including any direct, indirect, special, - incidental, or consequential damages of any character arising as a - result of this License or out of the use or inability to use the - Work (including but not limited to damages for loss of goodwill, - work stoppage, computer failure or malfunction, or any and all - other commercial damages or losses), even if such Contributor - has been advised of the possibility of such damages. - - 9. Accepting Warranty or Additional Liability. While redistributing - the Work or Derivative Works thereof, You may choose to offer, - and charge a fee for, acceptance of support, warranty, indemnity, - or other liability obligations and/or rights consistent with this - License. However, in accepting such obligations, You may act only - on Your own behalf and on Your sole responsibility, not on behalf - of any other Contributor, and only if You agree to indemnify, - defend, and hold each Contributor harmless for any liability - incurred by, or claims asserted against, such Contributor by reason - of your accepting any such warranty or additional liability. - - END OF TERMS AND CONDITIONS - - APPENDIX: How to apply the Apache License to your work. - - To apply the Apache License to your work, attach the following - boilerplate notice, with the fields enclosed by brackets "[]" - replaced with your own identifying information. (Don't include - the brackets!) The text should be enclosed in the appropriate - comment syntax for the file format. We also recommend that a - file or class name and description of purpose be included on the - same "printed page" as the copyright notice for easier - identification within third-party archives. - - Copyright [yyyy] [name of copyright owner] - - Licensed under the Apache License, Version 2.0 (the "License"); - you may not use this file except in compliance with the License. - You may obtain a copy of the License at - - http://www.apache.org/licenses/LICENSE-2.0 - - Unless required by applicable law or agreed to in writing, software - distributed under the License is distributed on an "AS IS" BASIS, - WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. - See the License for the specific language governing permissions and - limitations under the License. diff --git a/third_party/upstream/NOTICE b/third_party/upstream/NOTICE deleted file mode 100644 index 5e8bd01..0000000 --- a/third_party/upstream/NOTICE +++ /dev/null @@ -1,208 +0,0 @@ -Resin - -The LFM2 PyTorch model implementation in -models/torch/resin_models_torch/lfm2/model.py contains portions adapted from Hugging Face -Transformers: - - Copyright 2025 The HuggingFace Team - https://github.com/huggingface/transformers - -Hugging Face Transformers is licensed under the Apache License, Version 2.0. - -The GPT-OSS PyTorch model implementation in -models/cuda/resin_models_cuda/gpt_oss/model.py contains portions adapted from OpenAI gpt-oss and -Hugging Face Transformers: - - Copyright 2025 OpenAI - https://github.com/openai/gpt-oss - - Copyright 2025 The HuggingFace Team - https://github.com/huggingface/transformers - -OpenAI gpt-oss and Hugging Face Transformers are licensed under the Apache -License, Version 2.0. - -The retained model primitives and expert kernels in -models/cuda/resin_models_cuda/deepseek_v41 contain portions adapted from DeepSeek-V4: - - Copyright (c) 2023 DeepSeek - https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731 - -DeepSeek-V4 is licensed under the MIT License. - -The DeepSeek-V4.1 architecture, quantization kernels, and Engram implementation in -models/cuda/resin_models_cuda/deepseek_v41 contain portions adapted from: - - Copyright (c) 2023 DeepSeek - https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017 - revision 1bae0e1088993b9b82de08f59de5be5a34e42170 - -The MIT notice is retained in that model directory's LICENSE. - -The prepared DeepSeek-V4.1 forward pass in -models/cuda/resin_models_cuda/deepseek_v41 (attention, linear, HC, compressor, -indexer, Engram prefetch, shared batch metadata and the upstream CUDA operators) -contains portions adapted from DeepInfer: - - https://github.com/deepseek-ai/DeepInfer - revision 774ce244b87cc0eb8d4b94ab9f5de2ab22049f01 - -DeepInfer is licensed under the Apache License, Version 2.0. Its license is -retained in deepseek_v41/upstream/LICENSE. Copied operators identify their -source files and revision; adaptations bind them to Resin's paged storage. -The native batch metadata, paged index expansion and selected-index conversion -kernels adapt prepare_decode_metadata.py, get_kv_index.py and flash_mla_utils.py -from deepinfer/model_executor/layers/tilelang_ops/cuda at that revision. -The DeepSelect dependency is pinned to the matching DeepInfer submodule -revision bfa4507d935f17ebfc3d0f00ff7d3c9a4d0e5c18. - -Model weights are not distributed with Resin. Users who download Liquid AI or -OpenAI or DeepSeek checkpoints are responsible for the checkpoints' separate license terms. - -The native conversation formats in crates/resin-recipe follow the checkpoint -chat templates for thinkingmachines/Inkling-Small-NVFP4, Qwen/Qwen3.5, -RadixArk/Qwen3.8-27B-NVFP4, nvidia/Gemma-4-31B-IT-NVFP4, and -LiquidAI/LFM2.5-230M. Source revisions and retained reference licenses are -recorded in tests/recipe_fuzz/references/README.md and templates/README.md. -The LFM2.5 template is provided under the Liquid AI LFM Open License. - -The GPT-OSS tool-schema formatting in crates/resin-recipe/src/gpt_oss/schema.rs -is adapted from OpenAI Harmony (Copyright 2025 OpenAI), Apache License 2.0. -https://github.com/openai/harmony - -The DeepSeek conversation format and tools prompt in crates/resin-recipe -retain portions adapted from encoding/encoding_dsv4.py in DeepSeek-V4-Flash-0731 and -DeepSeek-V4-Pro-0813, Copyright (c) 2023 DeepSeek, MIT License. -The supported V4.1 format follows encoding/encoding.py in -DeepSeek-V4.1-Flash-1017, revision 1bae0e1088993b9b82de08f59de5be5a34e42170, -under the same copyright and MIT License. -The image preprocessing in crates/resin-recipe/src/media/deepseek.rs follows -inference/image_processor.py from that same V4.1 checkpoint revision. -The vision tower in models/cuda/resin_models_cuda/deepseek_v41/vision.py -follows its inference/vision.py, under the same copyright and MIT License. -The MIT notice is retained at crates/resin-recipe/licenses/deepseek-MIT.txt. - -The image/video geometry, normalization, patch layouts and rotary positions in -crates/resin-recipe/src/media/{gemma,inkling,qwen}.rs follow Hugging Face -Transformers' model processors (Copyright The HuggingFace Team), Apache License 2.0. -https://github.com/huggingface/transformers -The Rust implementations use the image crate's resize filters. - -The shared-expert scale layout in resin/backend/cuda/mega_moe.py and the -validation reference code in tests/test_mega_moe.py and -models/cuda/tests/test_deepseek_v41_kernels_gpu.py are adapted -from DeepGEMM at revision 1fc9c66975ee322d628b3288534ca0e9ce12a64d: -https://github.com/deepseek-ai/DeepGEMM - -MIT License - -Copyright (c) 2025 DeepSeek - -Permission is hereby granted, free of charge, to any person obtaining a copy -of this software and associated documentation files (the "Software"), to deal -in the Software without restriction, including without limitation the rights -to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -copies of the Software, and to permit persons to whom the Software is -furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice shall be included in all -copies or substantial portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE -SOFTWARE. - -The native GLM-5.3-Flash conversation format and tools instructions in -crates/resin-recipe/src/glm follow chat_template.jinja from -nvidia/GLM-5.3-Flash-NVFP4, revision -da920bb0b9f4a06727223a349e55468e38352348, based on zai-org/GLM-5.3-Flash. -Copyright (c) 2026 Z.AI Co., Ltd, MIT License. -The notice is retained at crates/resin-recipe/licenses/glm-MIT.txt. - -GLM image preprocessing in crates/resin-recipe/src/media/glm.rs and the vision -tower in models/cuda/resin_models_cuda/glm5_3_flash/vision.py are adapted from -Hugging Face Transformers, revision 50e35cc0ca18ec19b80653690de553e6decf74f5, -src/transformers/models/glm5_next/{image_processing_glm5_next,modeling_glm5_next}.py. -Copyright 2026 the HuggingFace Team, Apache License 2.0. -https://github.com/huggingface/transformers/tree/50e35cc0ca18ec19b80653690de553e6decf74f5 -The native implementation uses its own float bicubic antialias filter with the -upstream half-pixel convention and GLM canvas padding. Offline processor and -small vision-tower reference fixtures were generated from this same revision. - -The fused mHC readout and RMS normalization kernel in -models/cuda/resin_models_cuda/glm5_3_flash/mhc_cuda.py is adapted from SGLang, -revision baa9befa381511cd4d56f83980951e5a45fffc6d, -python/sglang/kernels/ops/layernorm/mhc.py. -https://github.com/sgl-project/sglang/blob/baa9befa381511cd4d56f83980951e5a45fffc6d/python/sglang/kernels/ops/layernorm/mhc.py -SGLang is licensed under the Apache License, Version 2.0. The adaptation keeps -token count dynamic, uses a fixed reduction split, and disables PDL scheduling. - -The accepted-prefix KDA state kernel in -models/cuda/resin_models_cuda/glm5_3_flash/kda_state_cuda.py is adapted from -flash-linear-attention 0.5.2, fla/ops/kda/fused_recurrent.py. -https://github.com/fla-org/flash-linear-attention/blob/v0.5.2/fla/ops/kda/fused_recurrent.py -The MIT license and copyright notice are retained at -models/cuda/resin_models_cuda/glm5_3_flash/LICENSE.fla. The adaptation computes -only accepted recurrent state, with Resin's FP32 beta sigmoid and explicit -preservation of empty rows; it does not produce attention outputs. - -The Qwen3.5 MLX model implementation in -models/mlx/resin_models_mlx/qwen3_5/model.py contains portions adapted from -Apple MLX LM (mlx_lm/models/qwen3_5.py and qwen3_next.py), which it also imports -for model constructors, activations, gated delta-rule and switched-expert kernels: - - Copyright © 2023 Apple Inc. - Copyright © 2025 Apple Inc. - Copyright © 2026 Apple Inc. - https://github.com/ml-explore/mlx-lm - -MLX LM is licensed under the MIT License. The MIT notice is retained in that -model directory's LICENSE. - -The fused decode convolution in -models/mlx/resin_models_mlx/qwen3_5/fused_conv.py adapts MLX v0.32.4's sigmoid -formula from mlx/backend/metal/kernels/unary_ops.h: - - Copyright 2024 Apple Inc. - https://github.com/ml-explore/mlx/tree/v0.32.4/mlx/backend/metal/kernels - -MLX is licensed under the MIT License. The notice is retained in the model -directory's LICENSE. The adaptation preserves BF16 convolution and sigmoid -rounding while returning Q/K/V and the next carry in the same dispatch. - -The mixed-precision RMS and rounded residual-RMS kernels in -models/mlx/resin_models_mlx/qwen3_5/fused_norm.py adapt the four-values-per-thread -reduction from MLX v0.32.4's mlx/backend/metal/kernels/rms_norm.metal: - - Copyright 2024 Apple Inc. - https://github.com/ml-explore/mlx/tree/v0.32.4/mlx/backend/metal/kernels - -MLX is licensed under the MIT License. The notice is retained in the model -directory's LICENSE. These kernels preserve FP32 effective scales and explicit -BF16 residual and output rounding. - -The small-partition attention merge in -models/mlx/resin_models_mlx/paged_attention.py adapts MLX v0.32.4's FP32 -column/row reduction order and exponential implementation from -mlx/backend/metal/kernels/reduction/{reduce_col,reduce_row,ops}.h and unary_ops.h: - - Copyright 2023-2024 Apple Inc. - https://github.com/ml-explore/mlx/tree/v0.32.4/mlx/backend/metal/kernels - -MLX is licensed under the MIT License. The notice is retained in the model -directory's LICENSE. The adaptation preserves FP32 product/reduction boundaries -and BF16 output rounding. - -The Qwen3.5 MLX vision adapter in -models/mlx/resin_models_mlx/qwen3_5/vision.py reuses MLX-VLM 0.7.11's Qwen3.5 -and Qwen3-VL vision modules, including their geometry and encoder blocks: - - Copyright © 2025 Prince Canuma - https://github.com/Blaizzy/mlx-vlm - -MLX-VLM is licensed under the MIT License. Its notice is retained in the model -directory's LICENSE and in the dependency distribution. diff --git a/third_party/upstream/import.json b/third_party/upstream/import.json deleted file mode 100644 index 3ea7d69..0000000 --- a/third_party/upstream/import.json +++ /dev/null @@ -1,3705 +0,0 @@ -{ - "repository": "https://github.com/ekzhang/resin", - "revision": "a1989594d81921be3bffe43804eb0550b216ca64", - "archive_sha256": "92464bd55a1b7c5f49291742b4dbaa365a52e78c6c4ca8000c638f9d4191733d", - "transform": "Complete source snapshot. Rename project namespaces to OrbitInfer and native module to orbitinfer._native. Preserve original licenses and attribution. Build, portability and documentation adaptations are described in docs/source-import.md.", - "files": [ - { - "source": ".editorconfig", - "path": ".editorconfig", - "source_sha256": "336c0b52be0ffbf18f0683934455045d1913e1fdfea651b7f1966ac8b096c993", - "import_sha256": "336c0b52be0ffbf18f0683934455045d1913e1fdfea651b7f1966ac8b096c993" - }, - { - "source": ".github/workflows/apple-tests.yml", - "path": ".github/workflows/apple-tests.yml", - "source_sha256": "9cfb8de25131c8d5b698ed089e47f5c69e41b7a35c753a543e7d6c967d1f0787", - "import_sha256": "9cfb8de25131c8d5b698ed089e47f5c69e41b7a35c753a543e7d6c967d1f0787" - }, - { - "source": ".github/workflows/python-checks.yml", - "path": ".github/workflows/python-checks.yml", - "source_sha256": "d2902b5b43242fb0b83051ba397ed9dab58d0f8a9cc2a095006d90d8f7839d46", - "import_sha256": "d2902b5b43242fb0b83051ba397ed9dab58d0f8a9cc2a095006d90d8f7839d46" - }, - { - "source": ".gitignore", - "path": ".gitignore", - "source_sha256": "14c10d0626bb101d505438e52951f8bb8b618b168f7359ec7a1173a5020979d1", - "import_sha256": "14c10d0626bb101d505438e52951f8bb8b618b168f7359ec7a1173a5020979d1" - }, - { - "source": ".type-libs/.gitignore", - "path": ".type-libs/.gitignore", - "source_sha256": "581a5e0a7660f37df648f645383fe969aab9465b5135fb23ff7204481352418c", - "import_sha256": "581a5e0a7660f37df648f645383fe969aab9465b5135fb23ff7204481352418c" - }, - { - "source": ".type-libs/patches/flashinfer/api_logging.pyi", - "path": ".type-libs/patches/flashinfer/api_logging.pyi", - "source_sha256": "16a4f7a4465ae33d2e5fd7220ee5f6c5b7a5ba6a66718676aa1b693ea7847552", - "import_sha256": "16a4f7a4465ae33d2e5fd7220ee5f6c5b7a5ba6a66718676aa1b693ea7847552" - }, - { - "source": ".type-libs/sources/.gitignore", - "path": ".type-libs/sources/.gitignore", - "source_sha256": "240a3e0d37d2e86b614063f5347eb02d4f99ca6c254de6b82871ff8d95532a7d", - "import_sha256": "240a3e0d37d2e86b614063f5347eb02d4f99ca6c254de6b82871ff8d95532a7d" - }, - { - "source": "Cargo.lock", - "path": "Cargo.lock", - "source_sha256": "bfa496823d25cf6e88e35b2b9ad465e4a73a404a41d59bbd2aa45c396a1aeb10", - "import_sha256": "095ce5c97f367249f521e3df3ce683a16ecc474051fe9662ae9ab0d6253228a6" - }, - { - "source": "Cargo.toml", - "path": "Cargo.toml", - "source_sha256": "fc836936ba946e92680115880a031fd0d6bd16a29cf51c07092c986bbaf795e2", - "import_sha256": "b0c9c6c9d4aef2be6fbbc129f5b29292eb651ff7fbafca4306a896301806bbb7" - }, - { - "source": "LICENSE", - "path": "LICENSE", - "source_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", - "import_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30" - }, - { - "source": "NOTICE", - "path": "NOTICE", - "source_sha256": "15b6ee5e6c8285b4bac27cbe401a39d6bd5b56eb18be068d5aaf5ce99ab90fec", - "import_sha256": "15b6ee5e6c8285b4bac27cbe401a39d6bd5b56eb18be068d5aaf5ce99ab90fec" - }, - { - "source": "README.md", - "path": "README.md", - "source_sha256": "9b3647fdbfb7cfe7799c4c84117861bc068a972c8f261dcc8d9bbe8e365bd26d", - "import_sha256": "e57a7080c4e6d82027d0349d8de2e8e748b6b618821c4d731042f827ed555cdb" - }, - { - "source": "conftest.py", - "path": "conftest.py", - "source_sha256": "42e17152982c4eb89a09c14d3ca02b516f13d8845a9c99d944d9cf47594973e0", - "import_sha256": "42e17152982c4eb89a09c14d3ca02b516f13d8845a9c99d944d9cf47594973e0" - }, - { - "source": "crates/resin-core/Cargo.toml", - "path": "crates/orbitinfer-core/Cargo.toml", - "source_sha256": "8e36bd5d6ac0d8e9220271e7f817f5734d20aaf129be019dd886ea7d37bb8ffa", - "import_sha256": "a6d2e646bed8768f68fd5e81438c4b7f183d9baefa36b1d801ae9bcc4176d7ec" - }, - { - "source": "crates/resin-core/src/batch.rs", - "path": "crates/orbitinfer-core/src/batch.rs", - "source_sha256": "16305f8c500966bb7f30b6fd9461a4007015af6385b42d315d6550f04b1ed918", - "import_sha256": "16305f8c500966bb7f30b6fd9461a4007015af6385b42d315d6550f04b1ed918" - }, - { - "source": "crates/resin-core/src/engine.rs", - "path": "crates/orbitinfer-core/src/engine.rs", - "source_sha256": "a2bbd70fe9fcbb1db74821c471a30faea3a2a2aeb03dd8f34709bf1449bc4816", - "import_sha256": "a2bbd70fe9fcbb1db74821c471a30faea3a2a2aeb03dd8f34709bf1449bc4816" - }, - { - "source": "crates/resin-core/src/ids.rs", - "path": "crates/orbitinfer-core/src/ids.rs", - "source_sha256": "53a515aec8a6ba99b65ff6b6c47b72d82e124c52c347e88db602c674788fcc79", - "import_sha256": "53a515aec8a6ba99b65ff6b6c47b72d82e124c52c347e88db602c674788fcc79" - }, - { - "source": "crates/resin-core/src/lib.rs", - "path": "crates/orbitinfer-core/src/lib.rs", - "source_sha256": "64c72621ad0d130c71eddf85b4b5c934245636db2ea918a065f8e1d08dd5a3f6", - "import_sha256": "1b6900f8b5418e65b086eebca24d8d4d3bbdea1a96e2374e98a4c3dbf6cb3e8f" - }, - { - "source": "crates/resin-core/src/logprobs.rs", - "path": "crates/orbitinfer-core/src/logprobs.rs", - "source_sha256": "908bace032b1bafceac049b8d947c9fdba171d36429f59c994fb70dff594f250", - "import_sha256": "908bace032b1bafceac049b8d947c9fdba171d36429f59c994fb70dff594f250" - }, - { - "source": "crates/resin-core/src/metrics/histogram.rs", - "path": "crates/orbitinfer-core/src/metrics/histogram.rs", - "source_sha256": "e02785810208205d337ec2eaf4f3931815dd5f32981a3b956cdc9ab0c52ab2e3", - "import_sha256": "e02785810208205d337ec2eaf4f3931815dd5f32981a3b956cdc9ab0c52ab2e3" - }, - { - "source": "crates/resin-core/src/metrics.rs", - "path": "crates/orbitinfer-core/src/metrics.rs", - "source_sha256": "3ba892a50f54ae81c1d5de097fddeab9b378afd13b3a0c8e1d49ba5c274bf052", - "import_sha256": "3ba892a50f54ae81c1d5de097fddeab9b378afd13b3a0c8e1d49ba5c274bf052" - }, - { - "source": "crates/resin-core/src/output.rs", - "path": "crates/orbitinfer-core/src/output.rs", - "source_sha256": "9de37420209f7838352b524feeffdc3264b63abff945989b9daccb7562b3cb83", - "import_sha256": "9de37420209f7838352b524feeffdc3264b63abff945989b9daccb7562b3cb83" - }, - { - "source": "crates/resin-core/src/parallelism.rs", - "path": "crates/orbitinfer-core/src/parallelism.rs", - "source_sha256": "832f73546b587b7d73c5fc19b78ef1a73eebdabfef565fd65a7789c3c190fa9c", - "import_sha256": "832f73546b587b7d73c5fc19b78ef1a73eebdabfef565fd65a7789c3c190fa9c" - }, - { - "source": "crates/resin-core/src/planning.rs", - "path": "crates/orbitinfer-core/src/planning.rs", - "source_sha256": "13c4b4cd65ecb896b5f87d5101d54ffd042585f7ede0a792afbadb5ba0ae6bcb", - "import_sha256": "13c4b4cd65ecb896b5f87d5101d54ffd042585f7ede0a792afbadb5ba0ae6bcb" - }, - { - "source": "crates/resin-core/src/resources/kv.rs", - "path": "crates/orbitinfer-core/src/resources/kv.rs", - "source_sha256": "c50836d9c638f9c201c66a22dda1dae107961bd91e393abd6fb4c31facc30858", - "import_sha256": "c50836d9c638f9c201c66a22dda1dae107961bd91e393abd6fb4c31facc30858" - }, - { - "source": "crates/resin-core/src/resources/mod.rs", - "path": "crates/orbitinfer-core/src/resources/mod.rs", - "source_sha256": "c21d21eb255089bc717ea8e7afed6fdf2f33f01db43620aabf9e042401b4a704", - "import_sha256": "c21d21eb255089bc717ea8e7afed6fdf2f33f01db43620aabf9e042401b4a704" - }, - { - "source": "crates/resin-core/src/resources/paged.rs", - "path": "crates/orbitinfer-core/src/resources/paged.rs", - "source_sha256": "b656832435c4f3aca64408cd023fb61c494b05b6ed170a7a11b479eb69360241", - "import_sha256": "b656832435c4f3aca64408cd023fb61c494b05b6ed170a7a11b479eb69360241" - }, - { - "source": "crates/resin-core/src/resources/paged_kv.rs", - "path": "crates/orbitinfer-core/src/resources/paged_kv.rs", - "source_sha256": "0d4b1b86926d05b3afa516385b30f54707aaadfd838d49ee5f29e43f6f11253b", - "import_sha256": "0d4b1b86926d05b3afa516385b30f54707aaadfd838d49ee5f29e43f6f11253b" - }, - { - "source": "crates/resin-core/src/resources/prefix.rs", - "path": "crates/orbitinfer-core/src/resources/prefix.rs", - "source_sha256": "1825444940b5371779fa4161c3e7fd89aaf2df3ead7659e8c6a5ead1853fa1c1", - "import_sha256": "1825444940b5371779fa4161c3e7fd89aaf2df3ead7659e8c6a5ead1853fa1c1" - }, - { - "source": "crates/resin-core/src/scheduler.rs", - "path": "crates/orbitinfer-core/src/scheduler.rs", - "source_sha256": "ac7896f31256d277288eb5541bd923a19542f22c388b4819c3e4b94fe390e406", - "import_sha256": "ac7896f31256d277288eb5541bd923a19542f22c388b4819c3e4b94fe390e406" - }, - { - "source": "crates/resin-core/tests/engine.rs", - "path": "crates/orbitinfer-core/tests/engine.rs", - "source_sha256": "65a66e7c2273ea39bf2bc75e72270feffa06be160aad3546442c3d890632eba1", - "import_sha256": "0f957d6f9f6a39bffba4f01b6dd28c26f5cbcf43a48455578bcca4681821a0eb" - }, - { - "source": "crates/resin-core/tests/preparation.rs", - "path": "crates/orbitinfer-core/tests/preparation.rs", - "source_sha256": "75d749ccd606702a3b61b33d5fe5fd03e3b9d067b0e6e532dbc413735eebcaab", - "import_sha256": "eb6779b14de8acb5f11e5affe8f0e032405ffef36cf282a862480f674217026b" - }, - { - "source": "crates/resin-core/tests/prompt_scores.rs", - "path": "crates/orbitinfer-core/tests/prompt_scores.rs", - "source_sha256": "5331d3f3dc390772983f6d789fafb1afeba7b12f2d4a7648d72e9fa443606992", - "import_sha256": "01389212a145abe41ace3c6fd33dae75f76667cebb5d73dc7b1f4206cd977b79" - }, - { - "source": "crates/resin-http/Cargo.toml", - "path": "crates/orbitinfer-http/Cargo.toml", - "source_sha256": "c35e4d8497951d711c50a4e51eb051e2586b2aafad3c980f5c2f0f2db7b14b67", - "import_sha256": "9750ff308dfb6becbb6dd4d670ae0987052e5069246e955f9c0b042ccce5fa34" - }, - { - "source": "crates/resin-http/src/anthropic.rs", - "path": "crates/orbitinfer-http/src/anthropic.rs", - "source_sha256": "98d05bb3082bfd6c1187d0be31457785a433fe169a1ead1f692a40101677d81e", - "import_sha256": "98d05bb3082bfd6c1187d0be31457785a433fe169a1ead1f692a40101677d81e" - }, - { - "source": "crates/resin-http/src/index.html", - "path": "crates/orbitinfer-http/src/index.html", - "source_sha256": "9b87a1043abc07466ddd66dca7fd38eef7622a916b6403fb9bba319e965a0b44", - "import_sha256": "4cd4a9dec2927221d50047e415b17927a9ada6c365523c2f8a68981284e29533" - }, - { - "source": "crates/resin-http/src/lib.rs", - "path": "crates/orbitinfer-http/src/lib.rs", - "source_sha256": "3933b2f029cd85b8697282aba608cbec74b4a146b5e66165248c9e3b53b4dc62", - "import_sha256": "2933ee2b787b2fbe00f7da94c61cf3a4e34e21508906902a95e5addc53f08446" - }, - { - "source": "crates/resin-http/src/metrics.rs", - "path": "crates/orbitinfer-http/src/metrics.rs", - "source_sha256": "2c2edb13fdf476d8da2eda34ba5fe8f73c4cd6278c66e5f339211ad78c5cab2b", - "import_sha256": "ed68c46ebbad151841b4e384b29523640358f54b7b10f40e4da051907e98bf9b" - }, - { - "source": "crates/resin-http/src/metrics_encoder.rs", - "path": "crates/orbitinfer-http/src/metrics_encoder.rs", - "source_sha256": "3e9525fd66618923a1dbb2973edf3c934cd8be334867c8bfbd8afb022cce7fa5", - "import_sha256": "c5c4a82ea9fb7127826e52713baf91185944f74f8013c8fbd0342da118884eea" - }, - { - "source": "crates/resin-http/src/openai.rs", - "path": "crates/orbitinfer-http/src/openai.rs", - "source_sha256": "26a606ba024a26a63d7e68a141587b6ba0b24809765144b8fce07a6521dc07f3", - "import_sha256": "740e6dfbbc669a9d64d843d5a5a02c5572db06baebae5dd9933827aa5b489e1a" - }, - { - "source": "crates/resin-http/src/protocol.rs", - "path": "crates/orbitinfer-http/src/protocol.rs", - "source_sha256": "36a3cdf7eb856eef0751ca82f36df7d15b961f14e30ad870cea23329165e4de2", - "import_sha256": "36a3cdf7eb856eef0751ca82f36df7d15b961f14e30ad870cea23329165e4de2" - }, - { - "source": "crates/resin-http/src/ui.rs", - "path": "crates/orbitinfer-http/src/ui.rs", - "source_sha256": "83107d8a084293153c25be28c9818cd59e15733e65db7ed2b047caaef0fcdba0", - "import_sha256": "5e455cd8ba3e58c4d7b0aff592e9978c05621778f192b196550649c9c138d156" - }, - { - "source": "crates/resin-http/tests/metrics.rs", - "path": "crates/orbitinfer-http/tests/metrics.rs", - "source_sha256": "f441a9214f582682f530078bb8edc2bbabe252463a4826381399407964a19eb6", - "import_sha256": "c619f4e82b2b40150bba75ac822478f968b482d7b6805bea38d5253c9ab0db75" - }, - { - "source": "crates/resin-http/tests/protocols.rs", - "path": "crates/orbitinfer-http/tests/protocols.rs", - "source_sha256": "4b370cea4f9dd67e78f52fe598485a959247e83b6f117ff07916e05b52ff7577", - "import_sha256": "113468d02511b33baf253238ce1bd8520f5198b0d46c3037b56caaa9e7626a0b" - }, - { - "source": "crates/resin-py/Cargo.toml", - "path": "crates/orbitinfer-py/Cargo.toml", - "source_sha256": "035b23c0c4f1b35b686523d566c0090b6b7a5764dbd40d30afbc2d02099e7ae2", - "import_sha256": "f5c764d72b9f9dd7309a9df1916b963fafc0251fc468b109d808c2d2c1560169" - }, - { - "source": "crates/resin-py/src/arrays.rs", - "path": "crates/orbitinfer-py/src/arrays.rs", - "source_sha256": "45955fcf420d090acd1b5705c72a50462a22d58db63a1d9337eb8d963c756cf8", - "import_sha256": "45955fcf420d090acd1b5705c72a50462a22d58db63a1d9337eb8d963c756cf8" - }, - { - "source": "crates/resin-py/src/cache.rs", - "path": "crates/orbitinfer-py/src/cache.rs", - "source_sha256": "3ece07249a8eff9e5d1a1d033d541c5eee120c63772c29c538c59c60eb8d0a94", - "import_sha256": "ad2596ef62f158b1c2e3a27a7b96b223852e90ccee91958e8b3f34d0de8ea113" - }, - { - "source": "crates/resin-py/src/cache_storage.rs", - "path": "crates/orbitinfer-py/src/cache_storage.rs", - "source_sha256": "95acd109e3d9c40cd7e1d6df6cc627d0e364b22d65e4ee0e0fb323fc2124e050", - "import_sha256": "28481b398c1ae59834f12c2dc10777376b7a2b542868dc97afa9abb5633b46fe" - }, - { - "source": "crates/resin-py/src/http.rs", - "path": "crates/orbitinfer-py/src/http.rs", - "source_sha256": "1b972468d75ff2513e08f49cc5295014255769af3452c3fb91640ebac5b17191", - "import_sha256": "527732e4acf35846745ccd3dc329a0549b2dc26c78e5254abb4b228a11e4224e" - }, - { - "source": "crates/resin-py/src/http_distributed.rs", - "path": "crates/orbitinfer-py/src/http_distributed.rs", - "source_sha256": "c2cfac1e5dc8881effaac71a179bea279d213cfd85424831dcd3c591c86a6512", - "import_sha256": "9e1df4ff48b418f55b8c06fda7de1e9599fb0d3871be4f8a8a8b5f93d4625923" - }, - { - "source": "crates/resin-py/src/http_logprobs.rs", - "path": "crates/orbitinfer-py/src/http_logprobs.rs", - "source_sha256": "e1e6e9abc201d7899454236a82a2941825b7f58efbe19fd2660e52cf37bcffe3", - "import_sha256": "e2ae0b4427ac8ad664d632989bf2ca83855504c6867c0d4cd2189ad8311f611f" - }, - { - "source": "crates/resin-py/src/http_output.rs", - "path": "crates/orbitinfer-py/src/http_output.rs", - "source_sha256": "e58d22ff557458c26ba5df1afb178d47b182a7550a18b84b312dfaabee746f21", - "import_sha256": "8da8b1ced275b3878b991e5c7837904b6738c24236c0e51ffaf265d53069bc22" - }, - { - "source": "crates/resin-py/src/http_session.rs", - "path": "crates/orbitinfer-py/src/http_session.rs", - "source_sha256": "9e3fd47cf86c226a9bff5ec93ade137c45ccc92850e2619e450c7099e82c0dbf", - "import_sha256": "b44449a33313e15f0c385fdeb9789084c4893723721dd7d9f3ea8ebc69db34c3" - }, - { - "source": "crates/resin-py/src/http_stop.rs", - "path": "crates/orbitinfer-py/src/http_stop.rs", - "source_sha256": "b9d1058b914c55543268e4841dfd04578545ef7c8ad04280c106067c2d22dbf3", - "import_sha256": "e2c591c36efca01ae21199f1419c59132f94c2fd0cc31d672a62ff009615a6bb" - }, - { - "source": "crates/resin-py/src/lib.rs", - "path": "crates/orbitinfer-py/src/lib.rs", - "source_sha256": "c09dc12da742ae337fd762d54a395853f1dd144e61751b078df0fa3d28a83651", - "import_sha256": "30c515387e8ae157e59599d10e166dc34a199e4388efc2e2c3e7136bb56b2c4d" - }, - { - "source": "crates/resin-py/src/media_payload.rs", - "path": "crates/orbitinfer-py/src/media_payload.rs", - "source_sha256": "527006862ffeaa8d1f1d02d5d306ac52f8590b77484a376a19de9d77937a9c36", - "import_sha256": "f38914607f7bf085b326e4421f50c9dfdfb16ce00ab9fd40fc7fbd28b8161678" - }, - { - "source": "crates/resin-py/src/media_preprocess.rs", - "path": "crates/orbitinfer-py/src/media_preprocess.rs", - "source_sha256": "d2d5d322e71a34684574e0fe7589599f0423dd19bd6b015756d546a55ef3cce5", - "import_sha256": "e101092a8e2d357800289b8d2251c296b51be055a3d885ec5d94446582236e9c" - }, - { - "source": "crates/resin-py/src/metrics.rs", - "path": "crates/orbitinfer-py/src/metrics.rs", - "source_sha256": "a214c25b79d05d446ec44817f5c7baadae09161d783522f7112bc33b11ce8e2a", - "import_sha256": "413d010e22074bcabc770225322c19b76d39453d57d717ef805607f305cd5525" - }, - { - "source": "crates/resin-py/src/numa.rs", - "path": "crates/orbitinfer-py/src/numa.rs", - "source_sha256": "489c1384bc10d047a502649b48d1375dbb1b8341a3b56561a2efe86fcb9c86a2", - "import_sha256": "489c1384bc10d047a502649b48d1375dbb1b8341a3b56561a2efe86fcb9c86a2" - }, - { - "source": "crates/resin-py/src/output.rs", - "path": "crates/orbitinfer-py/src/output.rs", - "source_sha256": "6d1b9e1054fe0e26d034bd9db4c7cdb05644e77c10f33ab877b95ad6c51333c4", - "import_sha256": "fa45c7be09a63609f1b960f0dea41a9ffb8e26db879d057883a486047e25bb0d" - }, - { - "source": "crates/resin-py/src/parallelism.rs", - "path": "crates/orbitinfer-py/src/parallelism.rs", - "source_sha256": "dcd79f09b73337026d3c186d46a0c1ecf8fadb18a9e5fe8361ec409baeba0346", - "import_sha256": "3db1a4f37de520853138a6e56ab28e02a35b66d0332b5ada27aa27632934b7c3" - }, - { - "source": "crates/resin-py/src/planning.rs", - "path": "crates/orbitinfer-py/src/planning.rs", - "source_sha256": "9454da1cecd64b926daa15c3c1f9546e2a1208a385a25eb8a455a328bfce429a", - "import_sha256": "c693b4a9882942f460fc6bc705c2b08f9b825109f2f9399993b50bb275a6a470" - }, - { - "source": "crates/resin-py/src/recipe.rs", - "path": "crates/orbitinfer-py/src/recipe.rs", - "source_sha256": "03860eb65b8e6f103dad4620411dedf5d9ee47fcb07ef49eb022e242e808ecf8", - "import_sha256": "5d377d19c7e6d6dccab514995f65d4e17d3614cbd30756f7159b15b6833db318" - }, - { - "source": "crates/resin-py/src/recipe_events.rs", - "path": "crates/orbitinfer-py/src/recipe_events.rs", - "source_sha256": "cf530b6c8a42fad25152faa23137475e3118d55d8165abbcc07050249c731cac", - "import_sha256": "e27cde0e5d1730bb3c8764a4b6ec7086349568a88ee0c415001fa60cad18cd08" - }, - { - "source": "crates/resin-py/src/recipe_media.rs", - "path": "crates/orbitinfer-py/src/recipe_media.rs", - "source_sha256": "3e4ef4d6e2e00389712f6191770a4ec980cb079b4373af32896008fd7ed870d5", - "import_sha256": "2c12a9314f030d5fa1ee717d3ac7c692f02505ccc6955252f1bfdcf7ca69e725" - }, - { - "source": "crates/resin-py/src/recipe_types.rs", - "path": "crates/orbitinfer-py/src/recipe_types.rs", - "source_sha256": "d0a1e83ef9c7b534d03a0f39ac1343e6b085e27148e804c4d2c2da58641d3dfb", - "import_sha256": "da2e0fe71ac2d5a7372d4f635211f93f21f944fcc2356873957e70ac547b5b33" - }, - { - "source": "crates/resin-py/src/runner.rs", - "path": "crates/orbitinfer-py/src/runner.rs", - "source_sha256": "f8f0737a0829d3b15425c81e2089547426bfc71e7dbcd78b2a139fbb5d8cb49b", - "import_sha256": "33c1ca8b956146fd95dde416eff2f9f91d5c4450f28419c10ca95ed032c34aa3" - }, - { - "source": "crates/resin-recipe/Cargo.toml", - "path": "crates/orbitinfer-recipe/Cargo.toml", - "source_sha256": "0b909513f81ba2164f6fc85ae7fa7eae292f6f686b42f9ed93ac8af1db589935", - "import_sha256": "97794541fa7ea5648fe299a56dbbda3eb931ae12b903da1791777c140f6a8d51" - }, - { - "source": "crates/resin-recipe/README.md", - "path": "crates/orbitinfer-recipe/README.md", - "source_sha256": "c08c74e2ffe34e4f6b221e925023e6d03a0491bd7dc8c4539d9c017119ac449f", - "import_sha256": "6e2e1de300064d6e928d00c03da7f95fd855e6cf88caa560ba73d7a95373db58" - }, - { - "source": "crates/resin-recipe/licenses/deepseek-MIT.txt", - "path": "crates/orbitinfer-recipe/licenses/deepseek-MIT.txt", - "source_sha256": "1c8f573e830ca9b3ebfeb7ace1823146e22b66f99ee223840e7637c9e745e1c7", - "import_sha256": "1c8f573e830ca9b3ebfeb7ace1823146e22b66f99ee223840e7637c9e745e1c7" - }, - { - "source": "crates/resin-recipe/licenses/glm-MIT.txt", - "path": "crates/orbitinfer-recipe/licenses/glm-MIT.txt", - "source_sha256": "30b85b6b9659f2e78aa259f8faf5d920a68dee7c9ced3fa6dba1f19f2bc4fca1", - "import_sha256": "30b85b6b9659f2e78aa259f8faf5d920a68dee7c9ced3fa6dba1f19f2bc4fca1" - }, - { - "source": "crates/resin-recipe/licenses/lfm2-OpenLicense.txt", - "path": "crates/orbitinfer-recipe/licenses/lfm2-OpenLicense.txt", - "source_sha256": "30adf9d6478191fb87f2424f63ba0728598335aaf99cd2848ef17e8e545fe94b", - "import_sha256": "30adf9d6478191fb87f2424f63ba0728598335aaf99cd2848ef17e8e545fe94b" - }, - { - "source": "crates/resin-recipe/src/deepseek/parser.rs", - "path": "crates/orbitinfer-recipe/src/deepseek/parser.rs", - "source_sha256": "22ee62bcca1a7c86467490bab0d9fec4e2aa099d69748d01818387fa8861e3df", - "import_sha256": "22ee62bcca1a7c86467490bab0d9fec4e2aa099d69748d01818387fa8861e3df" - }, - { - "source": "crates/resin-recipe/src/deepseek/tools.txt", - "path": "crates/orbitinfer-recipe/src/deepseek/tools.txt", - "source_sha256": "2ae2ff83edf990187ed419c4e2901bb9ea6ce2129eb503ac543b258e7de8231d", - "import_sha256": "2ae2ff83edf990187ed419c4e2901bb9ea6ce2129eb503ac543b258e7de8231d" - }, - { - "source": "crates/resin-recipe/src/deepseek.rs", - "path": "crates/orbitinfer-recipe/src/deepseek.rs", - "source_sha256": "e33a491ff6d198a1eb4391ebf78f8f7782b68ba691c11fd7d606c441f0b9d1fa", - "import_sha256": "e33a491ff6d198a1eb4391ebf78f8f7782b68ba691c11fd7d606c441f0b9d1fa" - }, - { - "source": "crates/resin-recipe/src/gemma/values.rs", - "path": "crates/orbitinfer-recipe/src/gemma/values.rs", - "source_sha256": "0400635ca4e65601b57be32fbd7b103b9e37e95d9a6a25dd9fc123e1c3cdff5d", - "import_sha256": "0400635ca4e65601b57be32fbd7b103b9e37e95d9a6a25dd9fc123e1c3cdff5d" - }, - { - "source": "crates/resin-recipe/src/gemma.rs", - "path": "crates/orbitinfer-recipe/src/gemma.rs", - "source_sha256": "097a6511fa6db84f39e5208428f3f19ed146fd0bbd67a82650093ae8b5ee2c15", - "import_sha256": "cdf242cff98ed184b3c9ee01c8069df1a858dce80fa1829c76926b13235c2125" - }, - { - "source": "crates/resin-recipe/src/glm/parser.rs", - "path": "crates/orbitinfer-recipe/src/glm/parser.rs", - "source_sha256": "f32f7044b1057a3e8196e1f55fc915b04b0444265c5030273443344fd088b1d8", - "import_sha256": "f32f7044b1057a3e8196e1f55fc915b04b0444265c5030273443344fd088b1d8" - }, - { - "source": "crates/resin-recipe/src/glm.rs", - "path": "crates/orbitinfer-recipe/src/glm.rs", - "source_sha256": "601c57f7b7028af36632b3be7e2363373c6d8f2a73c7dd764bdbc0efe504e79c", - "import_sha256": "601c57f7b7028af36632b3be7e2363373c6d8f2a73c7dd764bdbc0efe504e79c" - }, - { - "source": "crates/resin-recipe/src/gpt_oss/schema.rs", - "path": "crates/orbitinfer-recipe/src/gpt_oss/schema.rs", - "source_sha256": "5fc1bad4448d634149a523de63134c5b286c0f1134852e69c9f7918dd9e09d59", - "import_sha256": "5fc1bad4448d634149a523de63134c5b286c0f1134852e69c9f7918dd9e09d59" - }, - { - "source": "crates/resin-recipe/src/gpt_oss.rs", - "path": "crates/orbitinfer-recipe/src/gpt_oss.rs", - "source_sha256": "c2ab1afce69a9bbc7261dbdffeb6dbddd02542d701a9d0dd64f7f6e1ae9c96d7", - "import_sha256": "c2ab1afce69a9bbc7261dbdffeb6dbddd02542d701a9d0dd64f7f6e1ae9c96d7" - }, - { - "source": "crates/resin-recipe/src/inkling/render.rs", - "path": "crates/orbitinfer-recipe/src/inkling/render.rs", - "source_sha256": "f4b9c3f96049243954cd0bbcd7e2325f4b527466f33a0384f6a5a54c6f0b811f", - "import_sha256": "f4b9c3f96049243954cd0bbcd7e2325f4b527466f33a0384f6a5a54c6f0b811f" - }, - { - "source": "crates/resin-recipe/src/inkling.rs", - "path": "crates/orbitinfer-recipe/src/inkling.rs", - "source_sha256": "086f6ddcf8606dec4ffd4665ff1f8f3bcfd30dc9d91c254180452758da567ece", - "import_sha256": "086f6ddcf8606dec4ffd4665ff1f8f3bcfd30dc9d91c254180452758da567ece" - }, - { - "source": "crates/resin-recipe/src/json.rs", - "path": "crates/orbitinfer-recipe/src/json.rs", - "source_sha256": "02e42a877b630184a6b196dad5799111411e7c76d9660c05f36b03c3091ed588", - "import_sha256": "02e42a877b630184a6b196dad5799111411e7c76d9660c05f36b03c3091ed588" - }, - { - "source": "crates/resin-recipe/src/lfm2/literals.rs", - "path": "crates/orbitinfer-recipe/src/lfm2/literals.rs", - "source_sha256": "019eb94acac23d25ad5e4ffa8a115582d29d83adba11bfde48d9116f5d2f9205", - "import_sha256": "019eb94acac23d25ad5e4ffa8a115582d29d83adba11bfde48d9116f5d2f9205" - }, - { - "source": "crates/resin-recipe/src/lfm2.rs", - "path": "crates/orbitinfer-recipe/src/lfm2.rs", - "source_sha256": "23c61e20aea853f1e4b9a2148cdd80ca7831d22ce51464cc7d7f44f54f6f1b92", - "import_sha256": "23c61e20aea853f1e4b9a2148cdd80ca7831d22ce51464cc7d7f44f54f6f1b92" - }, - { - "source": "crates/resin-recipe/src/lib.rs", - "path": "crates/orbitinfer-recipe/src/lib.rs", - "source_sha256": "90afd2f180a9d35edaf16c375b17d5e41cb177b1c3dcf3b5cf411a68daf62dac", - "import_sha256": "90afd2f180a9d35edaf16c375b17d5e41cb177b1c3dcf3b5cf411a68daf62dac" - }, - { - "source": "crates/resin-recipe/src/media/deepseek.rs", - "path": "crates/orbitinfer-recipe/src/media/deepseek.rs", - "source_sha256": "8bb20e4e074cc56baeab9bd8ed3d418c33bd83573265f7c96990eddc8d205a27", - "import_sha256": "8bb20e4e074cc56baeab9bd8ed3d418c33bd83573265f7c96990eddc8d205a27" - }, - { - "source": "crates/resin-recipe/src/media/gemma.rs", - "path": "crates/orbitinfer-recipe/src/media/gemma.rs", - "source_sha256": "e0d91532528888d72ad2aca1e02d937e907b5568d71cd6fb16555a998761b88f", - "import_sha256": "e0d91532528888d72ad2aca1e02d937e907b5568d71cd6fb16555a998761b88f" - }, - { - "source": "crates/resin-recipe/src/media/glm.rs", - "path": "crates/orbitinfer-recipe/src/media/glm.rs", - "source_sha256": "3e4fba1a3849d6723f2bf9b09ce937b108ea84ee524fe95c358cfb4f5ca19eaf", - "import_sha256": "3e4fba1a3849d6723f2bf9b09ce937b108ea84ee524fe95c358cfb4f5ca19eaf" - }, - { - "source": "crates/resin-recipe/src/media/inkling.rs", - "path": "crates/orbitinfer-recipe/src/media/inkling.rs", - "source_sha256": "f01915b7de4c478142554d1eceda1323b2fcd81444331d90621a99ad81dae281", - "import_sha256": "f01915b7de4c478142554d1eceda1323b2fcd81444331d90621a99ad81dae281" - }, - { - "source": "crates/resin-recipe/src/media/mod.rs", - "path": "crates/orbitinfer-recipe/src/media/mod.rs", - "source_sha256": "563e720b210b624532677f073552a4161abef74ee52a30d47195aff7af58fa18", - "import_sha256": "563e720b210b624532677f073552a4161abef74ee52a30d47195aff7af58fa18" - }, - { - "source": "crates/resin-recipe/src/media/qwen.rs", - "path": "crates/orbitinfer-recipe/src/media/qwen.rs", - "source_sha256": "4aa5c0700aaadc9e5ecbb48e7bbd5d3affd64ce9a471cdb781d017abebe2cc2b", - "import_sha256": "4aa5c0700aaadc9e5ecbb48e7bbd5d3affd64ce9a471cdb781d017abebe2cc2b" - }, - { - "source": "crates/resin-recipe/src/parser.rs", - "path": "crates/orbitinfer-recipe/src/parser.rs", - "source_sha256": "43ec261b2920f34fc4c4c1ce68840ad7e0a8d5fc4f5702b6ca633361d51d39e1", - "import_sha256": "43ec261b2920f34fc4c4c1ce68840ad7e0a8d5fc4f5702b6ca633361d51d39e1" - }, - { - "source": "crates/resin-recipe/src/prompt.rs", - "path": "crates/orbitinfer-recipe/src/prompt.rs", - "source_sha256": "bcbd3336c1efb3039939a5de49b2f935f00e19644851c96db3c1e65002bb63f4", - "import_sha256": "bcbd3336c1efb3039939a5de49b2f935f00e19644851c96db3c1e65002bb63f4" - }, - { - "source": "crates/resin-recipe/src/provenance.rs", - "path": "crates/orbitinfer-recipe/src/provenance.rs", - "source_sha256": "1428873e71fd039fda442a084fe26f607bfaac540543e139a93a2b57b18d7534", - "import_sha256": "1428873e71fd039fda442a084fe26f607bfaac540543e139a93a2b57b18d7534" - }, - { - "source": "crates/resin-recipe/src/qwen/parser.rs", - "path": "crates/orbitinfer-recipe/src/qwen/parser.rs", - "source_sha256": "7d0919407f1e6f2ea1ae78e5433a725b85624b0271e28395ecfc716f3121f555", - "import_sha256": "7d0919407f1e6f2ea1ae78e5433a725b85624b0271e28395ecfc716f3121f555" - }, - { - "source": "crates/resin-recipe/src/qwen.rs", - "path": "crates/orbitinfer-recipe/src/qwen.rs", - "source_sha256": "a93e19b39267419329edf9aab38e97411a58aaf1857f0c348c7c71f497887ac2", - "import_sha256": "a93e19b39267419329edf9aab38e97411a58aaf1857f0c348c7c71f497887ac2" - }, - { - "source": "crates/resin-recipe/src/recipe.rs", - "path": "crates/orbitinfer-recipe/src/recipe.rs", - "source_sha256": "ad998ba2a584d49062469692e668cab9fcf40760d85be67d0ddc693a1a7427e1", - "import_sha256": "ad998ba2a584d49062469692e668cab9fcf40760d85be67d0ddc693a1a7427e1" - }, - { - "source": "crates/resin-recipe/src/render.rs", - "path": "crates/orbitinfer-recipe/src/render.rs", - "source_sha256": "cfae852536e483615416974ed01038dc39259680a662c1fdfb8c743263e1595f", - "import_sha256": "cfae852536e483615416974ed01038dc39259680a662c1fdfb8c743263e1595f" - }, - { - "source": "crates/resin-recipe/src/tokenizer/bytes.rs", - "path": "crates/orbitinfer-recipe/src/tokenizer/bytes.rs", - "source_sha256": "cafa0bdbb2132c7f352059cb7646ee78af21004beff4879ee301a9f2e07aebd4", - "import_sha256": "cafa0bdbb2132c7f352059cb7646ee78af21004beff4879ee301a9f2e07aebd4" - }, - { - "source": "crates/resin-recipe/src/tokenizer/trace.rs", - "path": "crates/orbitinfer-recipe/src/tokenizer/trace.rs", - "source_sha256": "8e2e62fceae526d7c0987bf38e89efde68dec4bf5c5af8ef7a3fc93682389bca", - "import_sha256": "8e2e62fceae526d7c0987bf38e89efde68dec4bf5c5af8ef7a3fc93682389bca" - }, - { - "source": "crates/resin-recipe/src/tokenizer.rs", - "path": "crates/orbitinfer-recipe/src/tokenizer.rs", - "source_sha256": "9f74aab1875fececebf2163aa9a2d41013753b8ed0316b87a84502b2d43c69fb", - "import_sha256": "9f74aab1875fececebf2163aa9a2d41013753b8ed0316b87a84502b2d43c69fb" - }, - { - "source": "crates/resin-recipe/tests/common/mod.rs", - "path": "crates/orbitinfer-recipe/tests/common/mod.rs", - "source_sha256": "ae99a11396e240f374e8fe3eedb7f239cc1a751264bb42fc8555aa4c33691bea", - "import_sha256": "812f20b52cb1804d4034858bdc9ff02319efd9554b3c59e9940d1b27da11e9fe" - }, - { - "source": "crates/resin-recipe/tests/deepseek.rs", - "path": "crates/orbitinfer-recipe/tests/deepseek.rs", - "source_sha256": "3aab0036ac0fd662e3bc01353df7381ab789ebe096d3e49edeee3c6f6efaa933", - "import_sha256": "4a6c4c0c69bb495b8e0df2be6e9e1ab824d228f8eda41aa02888e383cc5b5b58" - }, - { - "source": "crates/resin-recipe/tests/fixtures/deepseek_v41_prompts.jsonl", - "path": "crates/orbitinfer-recipe/tests/fixtures/deepseek_v41_prompts.jsonl", - "source_sha256": "7c42a694f4ceb0661bc2108516eb322aab213439037e1a7d1f22b9cb09900648", - "import_sha256": "7c42a694f4ceb0661bc2108516eb322aab213439037e1a7d1f22b9cb09900648" - }, - { - "source": "crates/resin-recipe/tests/fixtures/gemma4_prompts.jsonl", - "path": "crates/orbitinfer-recipe/tests/fixtures/gemma4_prompts.jsonl", - "source_sha256": "64bce4f1cda0721079cd8ccbca6ad29abffcbc4eb15529e522b157b21f16670c", - "import_sha256": "64bce4f1cda0721079cd8ccbca6ad29abffcbc4eb15529e522b157b21f16670c" - }, - { - "source": "crates/resin-recipe/tests/fixtures/glm_hf.json", - "path": "crates/orbitinfer-recipe/tests/fixtures/glm_hf.json", - "source_sha256": "8ced37e7b302ea33da0ef022ecf52dbf75f5aae5d11c70c90fbc8847b109770e", - "import_sha256": "8ced37e7b302ea33da0ef022ecf52dbf75f5aae5d11c70c90fbc8847b109770e" - }, - { - "source": "crates/resin-recipe/tests/fixtures/gpt_oss_harmony.json", - "path": "crates/orbitinfer-recipe/tests/fixtures/gpt_oss_harmony.json", - "source_sha256": "dd7a52e726f52adc48581becabf25ac689821dce2791649a78518d360f083eca", - "import_sha256": "dd7a52e726f52adc48581becabf25ac689821dce2791649a78518d360f083eca" - }, - { - "source": "crates/resin-recipe/tests/fixtures/lfm2_prompts.jsonl", - "path": "crates/orbitinfer-recipe/tests/fixtures/lfm2_prompts.jsonl", - "source_sha256": "a84d3201b775e79a8606def06fe8b7baf35199006a8785f510caac2c0cda897c", - "import_sha256": "a84d3201b775e79a8606def06fe8b7baf35199006a8785f510caac2c0cda897c" - }, - { - "source": "crates/resin-recipe/tests/fixtures/qwen_hf.json", - "path": "crates/orbitinfer-recipe/tests/fixtures/qwen_hf.json", - "source_sha256": "431f497f793fb411ff77b054aec543c42b469db705fc9e834d001591cc16c728", - "import_sha256": "431f497f793fb411ff77b054aec543c42b469db705fc9e834d001591cc16c728" - }, - { - "source": "crates/resin-recipe/tests/gemma.rs", - "path": "crates/orbitinfer-recipe/tests/gemma.rs", - "source_sha256": "b0fa7f3940a36a2eb8fe5d5aa5dd36df7eb5e057facd6d2cf5e2cb20cac6d7cc", - "import_sha256": "38c80f198ade64f2a665da4740a20d5ebbcf6b60a6616735410a9ace4928e778" - }, - { - "source": "crates/resin-recipe/tests/glm.rs", - "path": "crates/orbitinfer-recipe/tests/glm.rs", - "source_sha256": "cebdd7b3a57968cd166c1ef0033c836eeaff2f90f4ff180b8c778beb33bf8fe6", - "import_sha256": "09135ff3c908111061f4e1c82afc8c1d3e9fe54fad1574cb5f324b948595c324" - }, - { - "source": "crates/resin-recipe/tests/gpt_oss.rs", - "path": "crates/orbitinfer-recipe/tests/gpt_oss.rs", - "source_sha256": "5b1482df04c97ceb2ea8d3c27fcea6067b5634d2bc5f4ec70640842b04ffa609", - "import_sha256": "e6f4bb905eaad967c28bbea2aa6224727299b0e73f52220f8dbb95b677cd1b1a" - }, - { - "source": "crates/resin-recipe/tests/inkling.rs", - "path": "crates/orbitinfer-recipe/tests/inkling.rs", - "source_sha256": "c9dbae565873de53e26bac064ac966c92a9c8c381304e349e9ef49ce57e51792", - "import_sha256": "32a6c1163d267206fed08e8ffce378e74c58f2ca3fbef92636ee30b8e31b6f2b" - }, - { - "source": "crates/resin-recipe/tests/lfm2.rs", - "path": "crates/orbitinfer-recipe/tests/lfm2.rs", - "source_sha256": "be80cb9ac52ff46a4300f6f1a8ee9b81a883067f84042b4e00418e47092f56d0", - "import_sha256": "59fb75ce4f5af99f9845f5e6db514f33515c78b309e9ae87032e25feabb600e0" - }, - { - "source": "crates/resin-recipe/tests/provenance.rs", - "path": "crates/orbitinfer-recipe/tests/provenance.rs", - "source_sha256": "0a0fcdb8f8c118b5bab85b95109d9fc9ecd4d28b044ee0df1e991d51fe99b5e9", - "import_sha256": "6f530b60f70dfd5e1aebf37107c2829a808f0e4a5fdb6c69c26a4035daa571e2" - }, - { - "source": "crates/resin-recipe/tests/qwen.rs", - "path": "crates/orbitinfer-recipe/tests/qwen.rs", - "source_sha256": "b389487886257d162b0ef843354720ee600fafdb1cf03d33dbb8926044f46a99", - "import_sha256": "346b35016038ed7c98809dae3386e85e563fc6034d65cff0fd0f4bfec344b220" - }, - { - "source": "deploy/README.md", - "path": "deploy/README.md", - "source_sha256": "9713210c385e9b1d8ea3baa3b7c149c4c8bf1d730079d0dd6a710fceaefb3f64", - "import_sha256": "f6a4f02116d57682129e4ad58038d0fd48f52082548a3a2c105628fc2a6ab786" - }, - { - "source": "deploy/__init__.py", - "path": "deploy/__init__.py", - "source_sha256": "2d0248eb7cf6b6b1a9a57a928662da53f839bc475b2228bb5d3213479a90c653", - "import_sha256": "b75523787164121956d5af7f2032748e3e1c24b39049a681437c5cf1e7f5d6e6" - }, - { - "source": "deploy/checkpoint_cache.py", - "path": "deploy/checkpoint_cache.py", - "source_sha256": "64212420b1b21b28fe0d7e22277641ec4ff743801de2b98d1ea3ae6df4fba89a", - "import_sha256": "5bdb7b785ea97aa7453712da53beeea17d2595ca0f812411a09f2d488174abdd" - }, - { - "source": "deploy/jit_cache.py", - "path": "deploy/jit_cache.py", - "source_sha256": "003ccf6afc4fea75307368e94441e0ec79e4bae888196cf35ac42ab919fcb8b6", - "import_sha256": "1296d6ca0ecda39ba14682dc9cc13716e11be0e36ab328ac9e4d192c59fa7447" - }, - { - "source": "deploy/modal_bench.py", - "path": "deploy/modal_bench.py", - "source_sha256": "033e6f3d1ad3abe25c7b5322acfb91e39b005afbd04169c0f285eb9f3f5c325b", - "import_sha256": "3a2d25e013a17b376dfcc89d6a8c3b7d4b28159dce5c2d6b3ccb21e221b52130" - }, - { - "source": "deploy/modal_bench_scoring.py", - "path": "deploy/modal_bench_scoring.py", - "source_sha256": "30b1886454d3a6e021da365f74535a537bbbbc98f66a3c402e45dc6c1c61d3f8", - "import_sha256": "c640c44e8678e0aa0a7b2aca4e295c2a01f9002fab2b9953534e9094ea161261" - }, - { - "source": "deploy/modal_image.py", - "path": "deploy/modal_image.py", - "source_sha256": "deb30d06a92d8e8aac53b5c22c90b2706ce8353fc9472807e417c9ea004d80a7", - "import_sha256": "d8fdff613cfdb2922445c61a722dfc08be8eb0b5d7abe6e10cb79382abe8f2fc" - }, - { - "source": "deploy/modal_serve.py", - "path": "deploy/modal_serve.py", - "source_sha256": "1f4a84f81c221cfa4bbdd2d27936a7e0d8e31999a7cb350b3ef963baa5f95e56", - "import_sha256": "c379cd839c08d5fd7440e8059e6223587968010e6786f38af6ea63209ed1132c" - }, - { - "source": "deploy/models.py", - "path": "deploy/models.py", - "source_sha256": "6b68d1bb55ae70a5bb246852a500a8f5b01d68139c1cd32183c62830867827c0", - "import_sha256": "96a88a4ade71e7fda38b698ffcd10ca1335d10ac1aa1fa77c5256f59b91552ea" - }, - { - "source": "docs/architecture.md", - "path": "docs/architecture.md", - "source_sha256": "0cb5e04b9d3ed6d77a294d6444d163e68a6727213f263923c5e3e887f9fcf937", - "import_sha256": "48bfc4cebf0d11dd5e9781fcda588d93659ba313e41e26529622c7adbef01ae8" - }, - { - "source": "docs/benchmarks.md", - "path": "docs/benchmarks.md", - "source_sha256": "5835a268836467e966f0ac244437ae359d1d0b8c038abbd52f8b12304855f337", - "import_sha256": "4f91c0dd06bac988a55a2b928d6062879df0186ee430940e1780bbb7ad473f96" - }, - { - "source": "docs/cache-pool.md", - "path": "docs/cache-pool.md", - "source_sha256": "4968511e0902fefb2075d957b7b16265afc62c8150f98141c84721435b0b2bf8", - "import_sha256": "3151ef33a9467f79b3416c0a0ddddc7d23c0205a54a50a14de4503bc2e30318a" - }, - { - "source": "docs/cache-storage.md", - "path": "docs/cache-storage.md", - "source_sha256": "8d2ebefbf69390f1245b4b2d5205163f360d98679eba96824cddaad2885f89c0", - "import_sha256": "95a4fd365418ca10a1fb8924e565441000d5303190bdd94015bc0985b12e4369" - }, - { - "source": "docs/cuda-graphs.md", - "path": "docs/cuda-graphs.md", - "source_sha256": "b2b6de65c93d41d99cf8243414693c46e2fce7e9ed469a7da08e36d356855c7d", - "import_sha256": "9e76d00a3eb213084aa3964668ec162b8ac7533f31d59626c1a1579b298b9ad5" - }, - { - "source": "docs/distributed.md", - "path": "docs/distributed.md", - "source_sha256": "88fe468873a6504c0eeb0a56b4c5cbb4e321bd5164999af9d9a7b44f9d2d2ef1", - "import_sha256": "9bebcf15115395eefad733a5c67ccee5761c5b656ca493d222d213bbf5a29d36" - }, - { - "source": "docs/kernel-boundary.md", - "path": "docs/kernel-boundary.md", - "source_sha256": "687604823cae0541966a1eae984685fee7243b074db5e76964595e7561ed358d", - "import_sha256": "9fd34152e296ab02feaeb1afd7f8e2fa721befc447e967336c46c9bbec9a2712" - }, - { - "source": "docs/kernel-cache.md", - "path": "docs/kernel-cache.md", - "source_sha256": "09e87e04cf940795d9a26d199ebf53cc3846fd219090d99a4ae3792bff63d839", - "import_sha256": "594f9cbba4b2a1b02a297c4208c2b5b75e352160771bff4ac0d7b24e2137f03c" - }, - { - "source": "docs/logprobs.md", - "path": "docs/logprobs.md", - "source_sha256": "9adba53aeab8bf2eb7a7d4914d1676a05222a0e5d317f093e5ab43c5932d1d04", - "import_sha256": "783a8d98cc598c88fea8561ad705269fbf7045199682b249aa1c2fa339c379b5" - }, - { - "source": "docs/metrics.md", - "path": "docs/metrics.md", - "source_sha256": "3d7273b865952cb0da2cfe70cbf433e56ce254c2b2f0ff9420a63fbfa42e1487", - "import_sha256": "a5eff271df702aea89cd29394b214e398efd80022b6ba4ed253419959a02e671" - }, - { - "source": "docs/model-forward-pass.md", - "path": "docs/model-forward-pass.md", - "source_sha256": "5c05b10f1a316efd8be699f91c83b526506f0979839216bad1b2e3c50fb1116c", - "import_sha256": "b1297164e0e7280f18ad417d657a61435a81b2d6928b26e741e0c32a9865d04f" - }, - { - "source": "docs/numa.md", - "path": "docs/numa.md", - "source_sha256": "bd3f77cf80d42774cbf82e7fc4f344ac470739a06df886c228487c7ebccbe6d4", - "import_sha256": "1af986095cdd57503f183d67ae448cb8130e51c47272d1ccb234a645dffc4254" - }, - { - "source": "docs/scheduler.md", - "path": "docs/scheduler.md", - "source_sha256": "092b85fe718750935287fd3b22276d9715b8b34a9a4c628f61a85c3bc1802d22", - "import_sha256": "2f5d1a7596842c41a47ff9e4fdcf979ffcfae2975bf88938eac3c5f5e1f7b569" - }, - { - "source": "docs/scripts.md", - "path": "docs/scripts.md", - "source_sha256": "fcc30b70ed59b066e8f21b07c1fc54ab30924e3efdd69043bed0513b2ec84e2e", - "import_sha256": "5fe7543f08a674c6692f142235ffd5362118d98dedb2d1783c32e881c3d1dba3" - }, - { - "source": "models/cuda/ARCHITECTURE.md", - "path": "models/cuda/ARCHITECTURE.md", - "source_sha256": "6edde6dd9702b7ad72cc5e5763c8d51de37005d4d74c382aec5a1b9a89a2d401", - "import_sha256": "bccfe1cf143563c3ae454c00a8c0e67cbd87b8488a8f133a3e41306b2a9f9e28" - }, - { - "source": "models/cuda/pyproject.toml", - "path": "models/cuda/pyproject.toml", - "source_sha256": "e1fd8274913dca69bbe7ad8eab64b4469004dea07a8ff34953b49c713f034f26", - "import_sha256": "6b7625a551ee9347aab691c30d405d45f25c1b13d2c7df900aa6bca6ccc8cbf3" - }, - { - "source": "models/cuda/resin_models_cuda/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/__init__.py", - "source_sha256": "4fa1c36fb52141453902a1fe072afdf3f3ae0dce1d56464f07ec5a65cfdb3c98", - "import_sha256": "4fa1c36fb52141453902a1fe072afdf3f3ae0dce1d56464f07ec5a65cfdb3c98" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/LICENSE", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/LICENSE", - "source_sha256": "1c8f573e830ca9b3ebfeb7ace1823146e22b66f99ee223840e7637c9e745e1c7", - "import_sha256": "1c8f573e830ca9b3ebfeb7ace1823146e22b66f99ee223840e7637c9e745e1c7" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/README.md", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/README.md", - "source_sha256": "4381502c46c3f89c92dcecc05cc456a49333944808863fb6cb2aeed4c1be7ac0", - "import_sha256": "60035622a7aa0ddea54537d288b612e14a21a4c25faff2b24dfadc2ef7eb94c7" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/__init__.py", - "source_sha256": "33df8b487883194799b58227c143e7fb7302535fe9381d25942a2a727bf7bf23", - "import_sha256": "33df8b487883194799b58227c143e7fb7302535fe9381d25942a2a727bf7bf23" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/attention.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/attention.py", - "source_sha256": "d3078bef943a8599f76a2e81d1edaa241cd68759847e22e33ff8477c7c060ece", - "import_sha256": "ac23dddc28ac500f90c8bfd92d60dce458b6c8af70d9cd9efa83a60fbba05a2c" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/batch.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/batch.py", - "source_sha256": "7e2c41875dc591186eb58d41ca1bb073a8c7a5ff418e6d650822e0dec904f971", - "import_sha256": "e686b6fdf4230f541b709889b2baa9060cd7180395a9f7d241f28d3e6b4ba1b8" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/cache.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/cache.py", - "source_sha256": "7d86b8ee987077e783def144370c11b5543fe371132a1e178e567ba048357254", - "import_sha256": "5ec0a9e390a05eea4fd7690f2e41bfe88fc407507dc6c88b1ff82ecf41b6e188" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/checkpoint.py", - "source_sha256": "5b0880897866b79d4e73c232e574bc3a67dd8554ee3bad023c3aa6a6820949bf", - "import_sha256": "22e2486f00df70a0b33f868cd0635cfac3f070f184cb4b676f8997a065eec442" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/compressor.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/compressor.py", - "source_sha256": "e9c377e5dc820adedae42d057573799cf3a12052844430a04002e502bebe21b5", - "import_sha256": "1bd46d3f4afe6fdc96236e440d5b8e27abb94d194dc3142b601ef4d5dd758362" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/cuda.py", - "source_sha256": "75c1215ec7f6d2820fef175c1639d9dcb8dd0753cb7388d44f5015f3869bc72a", - "import_sha256": "0861c119ded08252c7e8b9abb8e25e344b19bd293e58894869cd5c4abde1aaf2" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/embedding.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/embedding.py", - "source_sha256": "fe7d76f103aa8f1fcc65a1a9b8ff1157ad80d2ef108c777a5f8dd913abaf803e", - "import_sha256": "850bee68362a8293cda1bdfa5ad0fbfe542ec847bdf1204fbd088874b9a91b53" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/embedding_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/embedding_cuda.py", - "source_sha256": "1b3941a62087845937df1ee12a13756abc784269ac8ad5a596f87bcbe4c077e3", - "import_sha256": "1b3941a62087845937df1ee12a13756abc784269ac8ad5a596f87bcbe4c077e3" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/engram.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/engram.py", - "source_sha256": "1c745ced5b97a0e9b5807abfdcf034632213298b961d99c563825139811dd900", - "import_sha256": "af5b770c519424559fd225eadb19611acd976630ca4c1a7e13342f1cc015e8ec" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/experts.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/experts.py", - "source_sha256": "0e6190c0f4710694fedcd6986f046319aa9037da0a0b1770a9a5a31b804c5f3b", - "import_sha256": "e8319e50216450588c3627dab71c5c70790af5d95c1049a53015336a455a85cc" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/gemm.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/gemm.py", - "source_sha256": "2b8b3231da24e64b147bbe578074bf3d5cf2e450d9a254c29aced3d593b6a239", - "import_sha256": "1075e437cbac52b9f422dc152e8fc00fc8cc3c7ed76c2c5546103291eeee0134" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/hc.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/hc.py", - "source_sha256": "329cadf6fa01b33c8512d8470f77675e6b2b491588ce000e4eb04a2736440618", - "import_sha256": "f3eeb19bbde7455d495a14dbb79f1bcc881c478fe198725c29478613d0c6e6dc" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/indexer.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/indexer.py", - "source_sha256": "49e3e72950554f984e116d8127a1b7d52605060e4ecdf69de623583a4f6cf2bc", - "import_sha256": "4f506232d9b91a5f95ba99b7440e3adc4bdfbfc5bf629248ce43aaf78e408674" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/kernels.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/kernels.py", - "source_sha256": "046b1c37528f76da65e34c7e977d1f0c307f210ad063e7588de68bfc39383e69", - "import_sha256": "046b1c37528f76da65e34c7e977d1f0c307f210ad063e7588de68bfc39383e69" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/linear.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/linear.py", - "source_sha256": "0261914ca9eb18cd6386e4d66729dc35b8559fd841008d175f51a61eac323919", - "import_sha256": "252a25ebd0db47e2105383eb4c72a83c23b1f76f1a2996d64ab11efc6bab0aa2" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/model.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/model.py", - "source_sha256": "0774be8c38c00d89b157dd1713c8eb4689648193ab878c5fecaa80759c63fde9", - "import_sha256": "1fca846bd02b655f249f10434239aa1ccdd4b2bada6949d8227cf4c0db9ac319" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/primitives.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/primitives.py", - "source_sha256": "5322e9817239ca832840e57ab0be19e1005782846c87bc2136434c03a92640bb", - "import_sha256": "a9fba2936fc0c252357f2e87f2730474c8fa9f13062409cd21a977bfca094d1a" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/tilelang_kernels.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/tilelang_kernels.py", - "source_sha256": "87bb958828b7ed587af66a0ed43720301c0f6e9654f600d2d7fe3676efa9f9d0", - "import_sha256": "87bb958828b7ed587af66a0ed43720301c0f6e9654f600d2d7fe3676efa9f9d0" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/LICENSE", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/LICENSE", - "source_sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", - "import_sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/__init__.py", - "source_sha256": "0e5de2d901ca3dbdb177e3f924282f9b00ca6f2c77ae1db3b20b4d5c2187f526", - "import_sha256": "0e5de2d901ca3dbdb177e3f924282f9b00ca6f2c77ae1db3b20b4d5c2187f526" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cache_common.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cache_common.py", - "source_sha256": "a7171a635409ac11dcab0a15c17bf79f533331c977a807f4293bfc023516e5f4", - "import_sha256": "a7171a635409ac11dcab0a15c17bf79f533331c977a807f4293bfc023516e5f4" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/candidate_blocks.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/candidate_blocks.py", - "source_sha256": "5d19700474d3f860f44b1fc56b7d2d829c1a73b1c0b50d5477560b26786f0a80", - "import_sha256": "0fa6b1f48725a6b30abfbe803a2f8e9825b36d852ac10be9c5864c52245cc529" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/common.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/common.py", - "source_sha256": "280c78ec4fd24945d455a1d45435242f36bebe0b3807b10995504f0ec20c4c6f", - "import_sha256": "280c78ec4fd24945d455a1d45435242f36bebe0b3807b10995504f0ec20c4c6f" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/__init__.py", - "source_sha256": "3d74000f200aa7d64b97a3a9850b3b4c24e6f755ad4578f2917ecab0fc306791", - "import_sha256": "3d74000f200aa7d64b97a3a9850b3b4c24e6f755ad4578f2917ecab0fc306791" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/batch_metadata.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/batch_metadata.py", - "source_sha256": "ac68569baaffdf4df32da115aa97c98bbe839e67e7f041f2af2efb1cb8827929", - "import_sha256": "ac68569baaffdf4df32da115aa97c98bbe839e67e7f041f2af2efb1cb8827929" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/candidate_blocks.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/candidate_blocks.py", - "source_sha256": "5b54809d80a1b312673c37b947c24f4d8b16fcaae36d4d1413bc58bde1b954b3", - "import_sha256": "5b54809d80a1b312673c37b947c24f4d8b16fcaae36d4d1413bc58bde1b954b3" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/compress.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/compress.py", - "source_sha256": "3bccf77bd0072eeecafe25b6d30fc1548ca0d9c28038995460753f1da4d41d91", - "import_sha256": "3bccf77bd0072eeecafe25b6d30fc1548ca0d9c28038995460753f1da4d41d91" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/flash_mla_utils.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/flash_mla_utils.py", - "source_sha256": "63ef4e584e2744189ff50717f624829f75d930415bda525978a4305585a98d56", - "import_sha256": "63ef4e584e2744189ff50717f624829f75d930415bda525978a4305585a98d56" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/fused_transform.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/fused_transform.py", - "source_sha256": "91346aae8aa976ac1b442b1742c08ade5224f04d90f6ffbc447287d56661684b", - "import_sha256": "ff6a00b8204ce8c424982802045f645becf0a927074a97fbe5e5d0fce329e005" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/get_kv_index.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/get_kv_index.py", - "source_sha256": "dd9338723082e1223d3f069513c7c184004c5162dad86b13cb7a5202f63502ec", - "import_sha256": "dd9338723082e1223d3f069513c7c184004c5162dad86b13cb7a5202f63502ec" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/cuda/quant.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/cuda/quant.py", - "source_sha256": "a36d3a00541a2b395d1d5fbd2b5eff509fbe3b418cef4a3b29a8384a0951887f", - "import_sha256": "a36d3a00541a2b395d1d5fbd2b5eff509fbe3b418cef4a3b29a8384a0951887f" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/upstream/fused_transform.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/upstream/fused_transform.py", - "source_sha256": "167f7fcbca66f90844cca8a6d28cc1afd6f31e26aa037eab1d69f5dbb849cc5b", - "import_sha256": "3bf703626e4821ba3e6d663d20f05e9680bc4678e29183bc3c534fe63d99c05f" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/vision.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/vision.py", - "source_sha256": "d37e4f8a426251d3d6007b3c7f395610e3d7144145bb6d21094cd7ba38d54d03", - "import_sha256": "d37e4f8a426251d3d6007b3c7f395610e3d7144145bb6d21094cd7ba38d54d03" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/weights.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/weights.py", - "source_sha256": "8e957148f2814fe965254e6e461448bf6b935740152c1a2e3046b919122d2515", - "import_sha256": "bf0689010e6847dbe0be95b9e1e277d90a670cde4a90d62acd6625dce1dab1e0" - }, - { - "source": "models/cuda/resin_models_cuda/deepseek_v41/yoco.py", - "path": "models/cuda/orbitinfer_models_cuda/deepseek_v41/yoco.py", - "source_sha256": "14ae76c0468d75842fcb34c31a774c4514756949f7b3d07541addf18ac5315ac", - "import_sha256": "50d2fabcbbc52bec198f0d905e9948375f9e7e12ecb6840f2547ea6f984fab14" - }, - { - "source": "models/cuda/resin_models_cuda/dense_parallel.py", - "path": "models/cuda/orbitinfer_models_cuda/dense_parallel.py", - "source_sha256": "a9f646badc6cb57141c88d3a00a2d85dd7b682668f3e730c7a10ba1374d4f9b9", - "import_sha256": "de2e1274bb613784dadb8bcc4344648bd3bf9888ad1e14df50652c8644e589e4" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/__init__.py", - "source_sha256": "f47304409c5cd8a4a8fadac3944f4fae9d0be3738785548ca721dacc068daa08", - "import_sha256": "f47304409c5cd8a4a8fadac3944f4fae9d0be3738785548ca721dacc068daa08" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/assistant_checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/assistant_checkpoint.py", - "source_sha256": "5e5f4d37dce224e10f7013ef7d0d5a3cf9c2f858a0c1c0d4271138cfc045cca2", - "import_sha256": "f5aed8f09d1093779b965ee4f77227da9271a80dc814771247cb7b852d0ae3b4" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/checkpoint.py", - "source_sha256": "5571e05d856c21475883bef1d470ed80d3b2ab08dceed841c8eeaa397f33109f", - "import_sha256": "51b21c528f6a05e77881a4b182807da534a404d55973a8b6398cc47890cc9d1d" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/cuda.py", - "source_sha256": "2db47d43ae6c0fb17beb49ce0ea333092e89f22f6dacc14302fb0c9853debb02", - "import_sha256": "10f4cfa55b354941533fcaf0b67e7f17c7d37b5061ed2da4629c7fa40ff09769" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_assistant.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_assistant.py", - "source_sha256": "3fdd3c3d5b1056a329069aafa883cf8454196bc1bbecb43ee17e4c85670dfd83", - "import_sha256": "c7ad1bd4917580eb065cdd2ed690a6eaa5f36065b0904c62cbdcafba904f9bdf" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_checkpoint.py", - "source_sha256": "7b7dff2ad9c7b87abf305af8a9c1efa219dd8ff7bfe4e4712d43bad3354e4c0a", - "import_sha256": "91e9ca0cc350fb16feee7643ecac1782533ea5a1b2072c3144877c5f64c7a790" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_cuda.py", - "source_sha256": "abf182ecad1dde64003a1b0f161afe553bbf8db5e2aaf2fc105960470e850fd5", - "import_sha256": "f58eabfe298e007348bf204c6cb3c4eaeb808e5787e29d8e32c868440f755677" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_kernels.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_kernels.py", - "source_sha256": "c32b0fd486956ca0012bac22c5141401cb573bed4f5b27cbdb6c997a564b434b", - "import_sha256": "c65319cf2631ed1f527e1a6d2af4cdbace30e6ccf390eafab3481731cd79639f" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_kernels_triton.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_kernels_triton.py", - "source_sha256": "786917a49eff847219a2abe5dd168f38e079f7759bde2197faaeb2078dde77db", - "import_sha256": "786917a49eff847219a2abe5dd168f38e079f7759bde2197faaeb2078dde77db" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/edge_model.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/edge_model.py", - "source_sha256": "94d83afa29546a260b4256a48893cc89a8957f49ec5f5fdd592bd647dfe19d38", - "import_sha256": "f438985da628af561addbe7d432fb2b0c5ba73ed62489b2efa83cd2ecc5020c9" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/model.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/model.py", - "source_sha256": "b5d3054d1b8d967840f4870ecd76917d3f37634ded11e68d54aa90f50b005417", - "import_sha256": "cdfbe08d8cd51e2d4a7c27053312282437a83849e5c3e8e0956f637d1e0014de" - }, - { - "source": "models/cuda/resin_models_cuda/gemma4/parallel.py", - "path": "models/cuda/orbitinfer_models_cuda/gemma4/parallel.py", - "source_sha256": "753a61d37eaa6a7f134161160aea8f54110c3a5938526fced217d86094522092", - "import_sha256": "7b2bd8df1a569503e6ee100ebf601e6a8f4222f16975ad19cc3f7c8a0bd76a52" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/LICENSE.fla", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/LICENSE.fla", - "source_sha256": "7bf7b5b53a1369cf5a561210285032d6cb310cad953c4421e349049fccee37a8", - "import_sha256": "7bf7b5b53a1369cf5a561210285032d6cb310cad953c4421e349049fccee37a8" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/__init__.py", - "source_sha256": "0631a62c8a4a67a96595e5ec37874824468d6378c914b95e27d419a2d47fd7d4", - "import_sha256": "0631a62c8a4a67a96595e5ec37874824468d6378c914b95e27d419a2d47fd7d4" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/attention.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/attention.py", - "source_sha256": "4bffbcc17a4b6ba760ce95f15e89c8b09b3882450e7259502c53dd36905a35f8", - "import_sha256": "a60b22ae7ff9f076cb58930d86c6822aa068ebc4b6f9e84e511276ff559707dc" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/cache.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/cache.py", - "source_sha256": "1f2be0322d5f4d320bc4124cdaa64d81b774ddf45fdc667d5ecef4f4602ea640", - "import_sha256": "9dc069c9c0b97ad5e2d98f254ac2f96a301f22f8681d515d7f42e53cec207396" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/cache_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/cache_cuda.py", - "source_sha256": "8a617f137bdfa14c1a08ec76aa44329821a3ffe38ad1b68ebddf6594ee0be175", - "import_sha256": "b3cec9d195b76aacb338ad623b389c52059b920971bae032fb812ae3584d7ec8" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/checkpoint.py", - "source_sha256": "b930100990d0ad0a45f2c499491d1e00e9aa11689040fd0ea66f3558f283f5c5", - "import_sha256": "abfcfdeee0c5097d276a6dd3c524efe4e59b83eef8a64ec38c47a7b31780cb13" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/cuda.py", - "source_sha256": "ba9436e2674807398e578d179bfe0503c7071fad6583e3e8f89bb1159e4bbdea", - "import_sha256": "09b54162fbd966e8373529edbf7f16eadbd17d6b337c2fd3f82333a989b07dd9" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/dsa.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/dsa.py", - "source_sha256": "5a66ac994f4cbab6c69cbdc19d8afc4eae22e51a784fe4138111d75cdb113366", - "import_sha256": "4d8da668d8d04e96e2f0337c212717adf0f73f693e83c596c9ed37c149463f40" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/dsa_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/dsa_cuda.py", - "source_sha256": "f17d835629e8a2a03975d3f24334d8ddff205c8f023836876822e86d1c4157d1", - "import_sha256": "f17d835629e8a2a03975d3f24334d8ddff205c8f023836876822e86d1c4157d1" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/feed_forward_fusion.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/feed_forward_fusion.py", - "source_sha256": "5c96b5803cfcc62443109819843eb4190c205cf51de0b68910632f3c6fd1bd5a", - "import_sha256": "5b96d02e5f102857e346353164c2aaf1f1d7bb3eec0137052f913719f28f6bd2" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/feed_forward_fusion_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/feed_forward_fusion_cuda.py", - "source_sha256": "328fb2669c60f9a19a9e35dedc367bb3c52e4c9f59b5b681a69d02ec0d059ee1", - "import_sha256": "328fb2669c60f9a19a9e35dedc367bb3c52e4c9f59b5b681a69d02ec0d059ee1" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/index_tail.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/index_tail.py", - "source_sha256": "3be2a06a319eed4b0e8d837b328557bef552808eff1e0f5ab4ed03031a3cc411", - "import_sha256": "c97dfdc9724780a726829c8d163165a3dc054f150016d5852a94b3971f6c2616" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/index_tail_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/index_tail_cuda.py", - "source_sha256": "65aa32e017c2944477a97cc040f082bf5a35442e773f4fee9b680c7a12fcca20", - "import_sha256": "911df82726a8837c8e4159c2b0532bbe04823924dfd4732705da9c27abc7d9c7" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/kda_state_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/kda_state_cuda.py", - "source_sha256": "98c759ddbb7c1385b4233744304c7d3353dfa4047ddf9cc29d855ce5b2346fc6", - "import_sha256": "f46788110b1279625eefe12c23b02d45df2aac436a0141d00627c8c04d27826f" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/kernels.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/kernels.py", - "source_sha256": "d44c28596f0e3550c2af6d97264a84a1b2348d70dbbe2c58066571a9996b1768", - "import_sha256": "90ed9e906e006eba37a5105938dad710d4088166491d7c41d30ac8f3999bf68a" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/kernels_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/kernels_cuda.py", - "source_sha256": "d454a2c6e99d456afcf3b6b13ec5356dfa69ee580d6a85e6278a7aae077055b5", - "import_sha256": "cbef7f51953aeebc0283ce5e4dc5df3364ea81f4113b61b8ccdc82892e150110" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/mhc_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/mhc_cuda.py", - "source_sha256": "a2c4578fd47cc1e5d120826993ce52852e87225329395106c04af62071278abe", - "import_sha256": "a2c4578fd47cc1e5d120826993ce52852e87225329395106c04af62071278abe" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/model.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/model.py", - "source_sha256": "503e2420c9ea08a09abad1c315a5db272ddb45e2f40bf922c442c0ff6efb2e6a", - "import_sha256": "9b8a02523b85c17d8505ee7f4f905bbfb41e700f434cca6fb055765474f11b22" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/normalization.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/normalization.py", - "source_sha256": "d4faf325ec798e8310907fe900061c0035c9cfb0334151d99670b35b2f37b066", - "import_sha256": "2f28dacbcd40ef65bb7397cd528d3177f9a4daf95059c024e5e42c12fdb38a11" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/normalization_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/normalization_cuda.py", - "source_sha256": "6cbf7e7b2fc0c1dc9fcb332d105dfbf14366b0aac49ee5f371bd7ef8020eeb9e", - "import_sha256": "6cbf7e7b2fc0c1dc9fcb332d105dfbf14366b0aac49ee5f371bd7ef8020eeb9e" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/projections.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/projections.py", - "source_sha256": "b5c90461cac0f50ae7926d8e403a48e69ba72b387c25086e3842e11641e0b12f", - "import_sha256": "83db2b43c7efe8e69967eab056b8758241c6b5b03686be71c327e51a10e37027" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/speculative.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/speculative.py", - "source_sha256": "3cb44a036948efc1ec22dc801e913d9429283d5cc1f9f2e1f436a3deceb525e5", - "import_sha256": "19b4332da5f9eb905b7d16bbc09dffc462ea6f13b3fb5c1a3277ae5be5a8fb87" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/speculative_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/speculative_cuda.py", - "source_sha256": "55bed2c2dce6838057034fcf3f3e38f6065ab314bd3adedffb787a064ee7c628", - "import_sha256": "55bed2c2dce6838057034fcf3f3e38f6065ab314bd3adedffb787a064ee7c628" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/staging_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/staging_cuda.py", - "source_sha256": "741c824ddaceefc98e5549b29eacf9a4a28cb48047d5b418bac9a0903fa672e0", - "import_sha256": "741c824ddaceefc98e5549b29eacf9a4a28cb48047d5b418bac9a0903fa672e0" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/state_cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/state_cuda.py", - "source_sha256": "1ed9c9b766e9f708bd68c8098fa34f1bfd859994afaca9000c36e71aef615bc2", - "import_sha256": "1ed9c9b766e9f708bd68c8098fa34f1bfd859994afaca9000c36e71aef615bc2" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/vision.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/vision.py", - "source_sha256": "b8fb4b8584e2457555e6fd9a9dc7e6a2572069a284684427f15c7979974458e0", - "import_sha256": "b8fb4b8584e2457555e6fd9a9dc7e6a2572069a284684427f15c7979974458e0" - }, - { - "source": "models/cuda/resin_models_cuda/glm5_3_flash/weights.py", - "path": "models/cuda/orbitinfer_models_cuda/glm5_3_flash/weights.py", - "source_sha256": "95227e7c01ce33bf2511f354c3d8821e6e595346d81b955ed4676b9b0fe3b5f5", - "import_sha256": "a2d6c285924172a08ff39d2e717ef6fd3aaee342e74abf62757addfb5e017457" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/__init__.py", - "source_sha256": "0814c270c97cb4475bf1ebcacb1b3132b0b8cc194dd79e6a0b48c2a836332045", - "import_sha256": "0814c270c97cb4475bf1ebcacb1b3132b0b8cc194dd79e6a0b48c2a836332045" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/checkpoint.py", - "source_sha256": "84c9132cc7c17728298a59f3802fb9a1c65cc76dbef572618b429cf039ae0ee4", - "import_sha256": "14054bcbe7860d2bbbb7e2bf3151e71d29896dae0785ac41a36617b9f60a5e4d" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/cuda.py", - "source_sha256": "c8be02092845acc442804dc03bb2ffbda57ff2996f455b38de98d83cdd698c2f", - "import_sha256": "68bfd04c11589beeb27329b6cf8569c938d187f2b6de0a0d4361f7bb2aa8a13a" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/cutile.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/cutile.py", - "source_sha256": "1bedef410d164f85f6c12ea5e07084892106cd234fed259f21f233468b189210", - "import_sha256": "1bedef410d164f85f6c12ea5e07084892106cd234fed259f21f233468b189210" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/flashinfer.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/flashinfer.py", - "source_sha256": "7cbdc04df2a106d8d0b67f2c56764067ef0d6d055c5f8eaeda2cd56843a9ba92", - "import_sha256": "7cbdc04df2a106d8d0b67f2c56764067ef0d6d055c5f8eaeda2cd56843a9ba92" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/model.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/model.py", - "source_sha256": "5458621352359d6dfc89877c581b22a5b607c80e9b7287ac50f3df92ea80e97a", - "import_sha256": "154b98eb6d0e352b17ca04854fe6af4565708776ceb322a3b1c1df3ac766dde0" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/mxfp4.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/mxfp4.py", - "source_sha256": "5331a2ccae2c3ff6842b4b5f323dc000833226ae3c0c26611524d14ff05196b3", - "import_sha256": "ffa22bd6534579b14b04a63f52ffbb13f1e4943483f7e625c01d2743aee645c7" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/paged_kv.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/paged_kv.py", - "source_sha256": "7f324a8f2d1eb2d5f3adc87788c2926bfd5a9d71db6050ff10a20e6c3d6e76bf", - "import_sha256": "3895f2b060de509d3982712886eca38570ab37750a9f072f06f5501fe3fe746a" - }, - { - "source": "models/cuda/resin_models_cuda/gpt_oss/weights.py", - "path": "models/cuda/orbitinfer_models_cuda/gpt_oss/weights.py", - "source_sha256": "9f3fb6811ce88cbc2e6082a9ddae1e24e6cb5c0b8ee927773768923caa7edf80", - "import_sha256": "7ec37552845dc412143a1f430550ae8c85157bddafb6007dc1a48801eca5abac" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/__init__.py", - "source_sha256": "df594b242f3214778f2fc711afb8eca3c36bee4207dfc1303b9b82e61337316b", - "import_sha256": "df594b242f3214778f2fc711afb8eca3c36bee4207dfc1303b9b82e61337316b" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/attention.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/attention.py", - "source_sha256": "9c44c4642d3969ebdbc457482c5b80f8d16a4dc8ebab833f6f120bce00350331", - "import_sha256": "404db831d02692fae4acff7e42ec72183bc5df2c08af15965fc1a33117304e0c" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/cache.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/cache.py", - "source_sha256": "df093181dd9410a88bfc09797ffa6cc31f8c6979f73b5ab14de81bb77f1c6b40", - "import_sha256": "4deb97b9ffaab9f8028c4300d72bb4ae4b47ad05014113d92d68af90bd8a616c" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/checkpoint.py", - "source_sha256": "f2bbe045c7b490cbdd001e6e8fc57591ef2ab0485772aeb8c721c7c49903e0c8", - "import_sha256": "bf76a35121ec58bcf6535072acd1deb15ff1c7572516da1c63ff5e7bf396ffb4" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/cuda.py", - "source_sha256": "6d5cc6bc316c60d7b798f2814c96e904fc7147bbe30ba6c378713a0fb429500d", - "import_sha256": "b544d9ffb94f211c136f3faab318eef6f9fbaa8c8f64a8dcb1bb08f759a09c73" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/cutile.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/cutile.py", - "source_sha256": "bf54a0a1f94b909f1cd92e3c1e81f1947d8246ca79b02bba772ec153832ddfa0", - "import_sha256": "bf54a0a1f94b909f1cd92e3c1e81f1947d8246ca79b02bba772ec153832ddfa0" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/dspark.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/dspark.py", - "source_sha256": "16d35e9b888e520551e2bcad24ddc0d264ecc5ad27e6f26057a240775a7096ed", - "import_sha256": "8f3254115bb07b6e7fea659d46b72f4af9825690c212bff13483955fcc2fbc3f" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/model.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/model.py", - "source_sha256": "c570a5bd536cb7261dd3f100cd508396a9d61a569c2cfb1c0418b48a2f42eea9", - "import_sha256": "90d5050285b3dc1895798dfd4461468d54621f3469eb2a51647bfcd8768df4f0" - }, - { - "source": "models/cuda/resin_models_cuda/inkling/weights.py", - "path": "models/cuda/orbitinfer_models_cuda/inkling/weights.py", - "source_sha256": "e0eeecbac96345e3ccc31645fdba58feb8dfd31a292f5889d07b32ce9ba8c24c", - "import_sha256": "208258397e82fa3a6f5fafa01947ab0cc0bd707fe6a12ac21f976b42e856cfa3" - }, - { - "source": "models/cuda/resin_models_cuda/mega_moe.py", - "path": "models/cuda/orbitinfer_models_cuda/mega_moe.py", - "source_sha256": "4e8ab52bd892d8ce8371e12fe4fd25326067e421b64fe3e9a9a87dca54cb0bb3", - "import_sha256": "5e7da60d2472a845e3f3082668858607e2b22dc963d4dfa8a878c7c207f02596" - }, - { - "source": "models/cuda/resin_models_cuda/nvfp4.py", - "path": "models/cuda/orbitinfer_models_cuda/nvfp4.py", - "source_sha256": "d02c2343a050333d503de37311f9deb3179e5cd05d37ce0c6b676aa2377aa402", - "import_sha256": "82b1411100862e4429fa0c80bf3b2745f80c83f703c3180b95dc9981fc277397" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/__init__.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/__init__.py", - "source_sha256": "b1a7cc243bb8fb6dd3a4bb5a00ec9ad108cb2482a9ad7d4dbc92c913887c40ab", - "import_sha256": "b1a7cc243bb8fb6dd3a4bb5a00ec9ad108cb2482a9ad7d4dbc92c913887c40ab" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/attention.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/attention.py", - "source_sha256": "63e36c0b565d7888e3e07b31042814b1d5b8b98b7547afec7dce92ee4fa0b43d", - "import_sha256": "a2eb13834996ca475eb81b36cb3ee8ce6b268b733036ceb0ce7e056b4556c0c2" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/attention_triton.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/attention_triton.py", - "source_sha256": "d273077795495d87bd5464a3d3948457aa88734cd53416f159ee19588c9a2bf7", - "import_sha256": "d273077795495d87bd5464a3d3948457aa88734cd53416f159ee19588c9a2bf7" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/cache.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/cache.py", - "source_sha256": "b397407edafa7f852367101b338f8e66a86c40aa9a197934452c4ffb6a0ecde9", - "import_sha256": "bb2647624c9d3bd43d1bdb93da485512f13225ed5f4095c9bb7e22ae5fb47be0" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/checkpoint.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/checkpoint.py", - "source_sha256": "bfc2ea01d7d56119f4baba8426fc129cdbb93d44786a5cabf4cc744676d7cf0f", - "import_sha256": "05a5c34d1f65860c56d2bbc4239f747a86dc9d0f555fd819df86c02bd50e849d" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/cuda.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/cuda.py", - "source_sha256": "b82ae8b34c72087d7620fabc7fe41d5ecb49f5f09105fb3d0ad6ffe860b8ad02", - "import_sha256": "2a17cd4a4ab28c2e5cf68d3cb5202d2c5d003e6dbd3eee79fc6de1f3144b9871" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/gdn_decode.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/gdn_decode.py", - "source_sha256": "5e371ccd41f189069403f6d55f9b003909a955020e539fc8c4de8abafd394ab0", - "import_sha256": "9aa5479917b74ebc9c5b5527126c434d2325eb78ebcc6d3267d22282afe070d4" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/model.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/model.py", - "source_sha256": "cfafa86b000b8361024e94b702df50bda4c33bafd5963512c1806b62a4185261", - "import_sha256": "c0a9a463f421c2345e7d50dcba35f1708559d12c72f795cd359e9c7f06ac8521" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/parallel.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/parallel.py", - "source_sha256": "523064b84b6872e4e04c30372bbf39d63bbd43202e7dd3c0450da55bdf29192a", - "import_sha256": "c153fa3b5ef9e636de01c3ad0046a0bb6dbed61c78ba355b21184f14bc7a3b2c" - }, - { - "source": "models/cuda/resin_models_cuda/qwen3_5/quantization.py", - "path": "models/cuda/orbitinfer_models_cuda/qwen3_5/quantization.py", - "source_sha256": "1b1f7fc8512c3d0bb2f3badf3485da39ffd710b31038ad2f2f42ace96ec39017", - "import_sha256": "1b1f7fc8512c3d0bb2f3badf3485da39ffd710b31038ad2f2f42ace96ec39017" - }, - { - "source": "models/cuda/tests/fixtures/README.md", - "path": "models/cuda/tests/fixtures/README.md", - "source_sha256": "6e6e7da0ce6ba39b0373b6042a4f10bb3c71c37fc619f0d583a9c4c055f2375e", - "import_sha256": "d8b29f28c5b862e7bde4af1da9b00badf9862ce2e063177f4ed2055c6b67e271" - }, - { - "source": "models/cuda/tests/fixtures/deepseek_v41/README.md", - "path": "models/cuda/tests/fixtures/deepseek_v41/README.md", - "source_sha256": "fe61ebe03168c299373a57dd2c60b0637b81224d64adca1a2878a42c5aff634a", - "import_sha256": "fe61ebe03168c299373a57dd2c60b0637b81224d64adca1a2878a42c5aff634a" - }, - { - "source": "models/cuda/tests/fixtures/deepseek_v41/config.json", - "path": "models/cuda/tests/fixtures/deepseek_v41/config.json", - "source_sha256": "8be45ce0476004a3f529fd896115a4a2e800a129ad2d3ec05b16050f52e21879", - "import_sha256": "8be45ce0476004a3f529fd896115a4a2e800a129ad2d3ec05b16050f52e21879" - }, - { - "source": "models/cuda/tests/fixtures/deepseek_v41/inference.json", - "path": "models/cuda/tests/fixtures/deepseek_v41/inference.json", - "source_sha256": "2e84f45cf1dac8c7fcbb200e96667d4b913275690668ed496f24c7747207a809", - "import_sha256": "2e84f45cf1dac8c7fcbb200e96667d4b913275690668ed496f24c7747207a809" - }, - { - "source": "models/cuda/tests/fixtures/gemma4_e2b_config.json", - "path": "models/cuda/tests/fixtures/gemma4_e2b_config.json", - "source_sha256": "1b28f3d2c3100f6c594754b81107428bd7b822a7f48272ca681dae9d2ec38330", - "import_sha256": "1b28f3d2c3100f6c594754b81107428bd7b822a7f48272ca681dae9d2ec38330" - }, - { - "source": "models/cuda/tests/fixtures/glm53_vision_hf.npz", - "path": "models/cuda/tests/fixtures/glm53_vision_hf.npz", - "source_sha256": "ed52aa57ad69d7934f8f06ce14933083e94170d17b157729fc307de4be472db4", - "import_sha256": "ed52aa57ad69d7934f8f06ce14933083e94170d17b157729fc307de4be472db4" - }, - { - "source": "models/cuda/tests/test_cuda_runner_graph_policy.py", - "path": "models/cuda/tests/test_cuda_runner_graph_policy.py", - "source_sha256": "d167a4ebdae1cb8bd7df36795f8f36cb1594a55a1dc5f02f565b64b618f62625", - "import_sha256": "a6cfe8bfeeface1e2e7237dd4a6b60fed4d2d5a0a078f2f15ab8c120d1151a8f" - }, - { - "source": "models/cuda/tests/test_deepseek_checkpoint_loading.py", - "path": "models/cuda/tests/test_deepseek_checkpoint_loading.py", - "source_sha256": "1a3c3c9879ad798f92815b06ce8ad29a7609025129f1aaef1b0d9d44ba134a78", - "import_sha256": "b79b6bb759b2966009832203554a0f0d2eacd02895f81105b72fdace49dc01b7" - }, - { - "source": "models/cuda/tests/test_deepseek_v41.py", - "path": "models/cuda/tests/test_deepseek_v41.py", - "source_sha256": "6f7a1107c5ca105f0e87c9f4c9774537a892403eca1f24b82e28988371f07949", - "import_sha256": "6f11ffe562f8e40f00b54192e84c640e11780663a008ab91b25f7efb9ecebb5d" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_attention.py", - "path": "models/cuda/tests/test_deepseek_v41_attention.py", - "source_sha256": "6431afff06aedfc2a8450de033f5307b850f9662627a6706bdc411f12988c170", - "import_sha256": "37fc557729c4f05fa03c48208393ea31c3a6cdd5c7aaa592053a3e2e6f0660b4" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_batch.py", - "path": "models/cuda/tests/test_deepseek_v41_batch.py", - "source_sha256": "f63928d1bd6cb74d3ec80b8167ed290a11bbca72a0f603a7fad38e48a81781b2", - "import_sha256": "bebc8f1f0788e6f0ad5277e70f615a6fea2c3a9a01b189c922b6b3a036e09a90" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_candidates.py", - "path": "models/cuda/tests/test_deepseek_v41_candidates.py", - "source_sha256": "a392728eecf53d361e1eee5057f54bc5fe7df0c05d2ebdb36a5d83ff3e7a6049", - "import_sha256": "b82efe738f954ca4df01b05ae6db5a175a77af7f8871216edcb01bc90affd157" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_compressor.py", - "path": "models/cuda/tests/test_deepseek_v41_compressor.py", - "source_sha256": "2f11ebfec429ecad0bcfd244938d5607976dc495ccd403e8e3769c6998847d4a", - "import_sha256": "8f46b5950b255b418ed4068815de348afe113705edbec4006604c30d9e51710d" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_contracts.py", - "path": "models/cuda/tests/test_deepseek_v41_contracts.py", - "source_sha256": "3c66fcc0eda462f3f3f69a748304de6e0835cfb1ba47eb60bcb67498f149c117", - "import_sha256": "345016dad621cdc017fab254f59b58789f366248a781bdf45ea3b08ce9629fe0" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_engram.py", - "path": "models/cuda/tests/test_deepseek_v41_engram.py", - "source_sha256": "d5863742521e1f5fa5292c7b1c997b767f1861b331500dc87ca23bbfe5fd70b1", - "import_sha256": "71ab5d33e2b4cc6d239c40c5e0c47c67972d69e78ea2f59f92686a9dcfbf844f" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_experts.py", - "path": "models/cuda/tests/test_deepseek_v41_experts.py", - "source_sha256": "9f3aa8dcdde3864eda3d23dd219fd73c05bce79736867d7c643f108b4ecc3e1d", - "import_sha256": "ee19e0288494cc08b3130e57c263aacfd3fcd3f25f111d2f34f39178fdfb1c4d" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_kernels_gpu.py", - "path": "models/cuda/tests/test_deepseek_v41_kernels_gpu.py", - "source_sha256": "b0b9638faafd4a150a2eee79b4959e0b89b59ebd1236c3558ff20e2b55b2633c", - "import_sha256": "e7c2c2bd728bebc27018a9c7eda62663e946f63a2c685b431ebbc26536e94352" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_media_transport.py", - "path": "models/cuda/tests/test_deepseek_v41_media_transport.py", - "source_sha256": "771e5f5596e320b09877984bc236c5e6811b5f8db0584c7d37a836577af9561b", - "import_sha256": "a22cfd609e3cf26a70f916a55b5221fc57105c38f8a83f5897f269de3426320d" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_model.py", - "path": "models/cuda/tests/test_deepseek_v41_model.py", - "source_sha256": "3b84d001d9c2939e4fad263c84dbb002da292cbc82fb626a0091cdf19aaf370b", - "import_sha256": "9e85aeab0ce3a36287985961c79b00a23c084bc89157b460e361c9965032936a" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_projections.py", - "path": "models/cuda/tests/test_deepseek_v41_projections.py", - "source_sha256": "456afbee37126cad5464ed6d910796f801be1a7639afc2c07a9a1bd273fac753", - "import_sha256": "167bad5e8867e3aa7604d626fc2a80a9d34f3b94c378da2e50acd99af410eeee" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_residuals.py", - "path": "models/cuda/tests/test_deepseek_v41_residuals.py", - "source_sha256": "62df2257599f8f0b307022fe3f45b5b91c25814b15dc054ec0e69ac476eff12c", - "import_sha256": "06e879532fb5da5884036e927a8c2c0670d1ff5bf779bfc86d374abc0e36166e" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_vision.py", - "path": "models/cuda/tests/test_deepseek_v41_vision.py", - "source_sha256": "c4ba9a5d62d29aa70149e51b7ca3a5a8b3c70571fbc10b9a2d9ac59c0e678f3a", - "import_sha256": "14fe49090102f722f083edaaa5ff83cdd6478ab8d160d0681352c067387d9f4b" - }, - { - "source": "models/cuda/tests/test_deepseek_v41_yoco.py", - "path": "models/cuda/tests/test_deepseek_v41_yoco.py", - "source_sha256": "2d9a95a152a1d8ad4b24f7c27404421b6b49d4b1cdb8d1704ff352664b9b1da6", - "import_sha256": "5f10a02e6d2d448175cb13a681335c6e7dcc8232fcc5bbbeb56cb82699f40fa9" - }, - { - "source": "models/cuda/tests/test_dense_parallel.py", - "path": "models/cuda/tests/test_dense_parallel.py", - "source_sha256": "3426774ef322a99f10a15bccfd20ed05a61de3a483cdb9f39cb21d6ca341d9fc", - "import_sha256": "b0d6f95026320b05482d55bfb148329866ffe931101ed5a692bb437ebeb996e5" - }, - { - "source": "models/cuda/tests/test_dense_parallel_gpu.py", - "path": "models/cuda/tests/test_dense_parallel_gpu.py", - "source_sha256": "cb1bf4b7a6834576d6060b477a022971c1199ffb797ce32280c7ffc7a025f1fb", - "import_sha256": "dca879997d4e4ef8fbd63eb0aa81669ad698c091fcc030b72ef4f9d3617c70dd" - }, - { - "source": "models/cuda/tests/test_gemma4_assistant.py", - "path": "models/cuda/tests/test_gemma4_assistant.py", - "source_sha256": "09298cab19499aff22a9764140a35ff7a0e7a4ca17334b591b8640bbfe5a3f53", - "import_sha256": "ade645fe4947014acaf03ba40ab52af589e39d2471bab872f586a7a13d8a77d5" - }, - { - "source": "models/cuda/tests/test_gemma4_checkpoint.py", - "path": "models/cuda/tests/test_gemma4_checkpoint.py", - "source_sha256": "07ca800bcb4bdd4f178bbbe768407e40fd743e7b7c33d77e228de5e46ed0bced", - "import_sha256": "3a3b4800462b245af6d56309f046d31e32babe0145dcf43abfca66e3a8d5c4f6" - }, - { - "source": "models/cuda/tests/test_gemma4_codec.py", - "path": "models/cuda/tests/test_gemma4_codec.py", - "source_sha256": "8d27c27e2eee47af87f7f13b5493d38dcefdd6d7e55379698ce4a9376761e709", - "import_sha256": "2944442a061e21c49c6abc83a968d973040564cc6aec24ba1c6334a2ed20f1e3" - }, - { - "source": "models/cuda/tests/test_gemma4_config.py", - "path": "models/cuda/tests/test_gemma4_config.py", - "source_sha256": "dcd5a8716bd67a311711798484de5f302985d9a5c60624b01107d7891ecf3b86", - "import_sha256": "f6ae2e6045dac6df80100c12d8839e05e053aaa8d2e523103468c8ea8add7c65" - }, - { - "source": "models/cuda/tests/test_gemma4_cuda.py", - "path": "models/cuda/tests/test_gemma4_cuda.py", - "source_sha256": "7931f6aa2b3f25190fecf764460befe81514c71e751cae43900046a7515ea45e", - "import_sha256": "9f6ce5e7967e2dfc936bf957be77a77c016e4c571defd38718c80158940a4c74" - }, - { - "source": "models/cuda/tests/test_gemma4_edge_kernels_cuda.py", - "path": "models/cuda/tests/test_gemma4_edge_kernels_cuda.py", - "source_sha256": "c4b092411a3af6e2f792515433a641ddb8aba8a0ea19d6c47a4960b68701997c", - "import_sha256": "470809e1014c67209c73d27024203519af9dbc4ce82212101b2382d7faea8538" - }, - { - "source": "models/cuda/tests/test_gemma4_edge_model.py", - "path": "models/cuda/tests/test_gemma4_edge_model.py", - "source_sha256": "a1fb0eca7eaca572851017532b8bf613e64c1551bc078a9680676e3c8fee6a05", - "import_sha256": "a65dcaf31293be07a8d7d46a7fd87c724f9496178fb0308d2d65728e162afbd7" - }, - { - "source": "models/cuda/tests/test_gemma4_edge_overlap.py", - "path": "models/cuda/tests/test_gemma4_edge_overlap.py", - "source_sha256": "5abb3b506c0123b8a469e99777b25e291518e8f01f7b369ae52d375dd8f60b06", - "import_sha256": "3d1e33a89a03e98e61a86af9cf9c0c3aaf4e42c6cb6cccfdc3b8300ae548af95" - }, - { - "source": "models/cuda/tests/test_gemma4_media.py", - "path": "models/cuda/tests/test_gemma4_media.py", - "source_sha256": "b827967c36ece42503926b484d5dc999992fb37a97aec97eab35fe5166288485", - "import_sha256": "4a3f5a6be37bc745347255f66e4c0345b32802ba2ce10ac91e6aba8f37eac6ce" - }, - { - "source": "models/cuda/tests/test_gemma4_model.py", - "path": "models/cuda/tests/test_gemma4_model.py", - "source_sha256": "e1df38e6a1c59e3b7f65ce9ab1139ebf69ca1b8c53955667dc7129355c228adc", - "import_sha256": "83f7e4efc3c97bf7f80796b8d698284341d210067aeb8668ed1ebc3c74c6dc51" - }, - { - "source": "models/cuda/tests/test_gemma4_nvfp4.py", - "path": "models/cuda/tests/test_gemma4_nvfp4.py", - "source_sha256": "6dfb754bb17d85410ae19cc2e1968122c337aa622d10456c0a77899f01fd450c", - "import_sha256": "1068eaa1a6f51d5db958b31679752aa28ba28e5280d21b4e1af370da0d15304b" - }, - { - "source": "models/cuda/tests/test_gemma4_parallel.py", - "path": "models/cuda/tests/test_gemma4_parallel.py", - "source_sha256": "aa61f29fdf531ebbde4da8535b117e55b2837575140d5a4e6f7f9bd5ad336d10", - "import_sha256": "d300bd0340b8df6e241c464cea5a331f830cb6743416aa7f49076e6a90273310" - }, - { - "source": "models/cuda/tests/test_gemma4_plan.py", - "path": "models/cuda/tests/test_gemma4_plan.py", - "source_sha256": "c240b99d358547966d6e2969e936bf83e7ecdc51f5a50c64ef3563b2d422dbb3", - "import_sha256": "8aaf195640547022add77740ad22dee50483372f136c0d347a6002c14e6c06f3" - }, - { - "source": "models/cuda/tests/test_glm53_attention.py", - "path": "models/cuda/tests/test_glm53_attention.py", - "source_sha256": "cfc5ab8dc9571ffe9bc497d6c36d1c9c1a638575e92d3464a78e87b86328c27a", - "import_sha256": "b3f8fcb56fbee37906bf5b57db4acb331190ea1ce239fd3536d71996bb3e57ce" - }, - { - "source": "models/cuda/tests/test_glm53_attention_compilation.py", - "path": "models/cuda/tests/test_glm53_attention_compilation.py", - "source_sha256": "877040b6e6c57231aabc736272d7b5000355232b241293c1b530cee61df52fdb", - "import_sha256": "6253d982e18fb7df1036a767239d9d385481c7780158bf03fd312b456562235d" - }, - { - "source": "models/cuda/tests/test_glm53_cache.py", - "path": "models/cuda/tests/test_glm53_cache.py", - "source_sha256": "e12c6c135d87e216782dbd84e3dbd5c10d827d8755ec2cf40b95a6de1b0f1dca", - "import_sha256": "5207ec250b688caf567a306f3bb109600a252fe991f9151a73b1bb2ddba15479" - }, - { - "source": "models/cuda/tests/test_glm53_cache_cuda.py", - "path": "models/cuda/tests/test_glm53_cache_cuda.py", - "source_sha256": "3d3f9ef339bf01298ca6550a437f0c0dd6733e69fe2409d513cf26289538f5f5", - "import_sha256": "b0d625bb257a44ef1914d6065f4bb4ee6f15ea0350c9b5bdd0e0ad890d574e3f" - }, - { - "source": "models/cuda/tests/test_glm53_dsa_fp8.py", - "path": "models/cuda/tests/test_glm53_dsa_fp8.py", - "source_sha256": "9009b7178ebfedee4c1e1d38409215624156c1adee2de05e03a6346d497bbd32", - "import_sha256": "4a310c92bf505a23f68df5e5924fa10d45ab1224379166d5d9ef23dd77f9545b" - }, - { - "source": "models/cuda/tests/test_glm53_expert_tp_gpu.py", - "path": "models/cuda/tests/test_glm53_expert_tp_gpu.py", - "source_sha256": "a422b55606d7b7a1fcf26742aa3ba367a2626ac9084ba4a99a68a89eaa30eb80", - "import_sha256": "1d1a51af733a70c9a7f5c6f0e8328f224dc7b8c3f037af6e2b242837e8d75aec" - }, - { - "source": "models/cuda/tests/test_glm53_feed_forward_fusion.py", - "path": "models/cuda/tests/test_glm53_feed_forward_fusion.py", - "source_sha256": "44c15c6e41a1ce7303de947c02f913bc68ef196dcdf00c84a2762b3b2ba8903c", - "import_sha256": "0caee208cfe1ee8ee4895a0a16c30719854202f3b479a3a05197f79dbe74ddcc" - }, - { - "source": "models/cuda/tests/test_glm53_index_tail.py", - "path": "models/cuda/tests/test_glm53_index_tail.py", - "source_sha256": "cf8e775a5fc4a78e392569db6e0682beca3618e7c912c6023ba872dd83b07ad2", - "import_sha256": "4d2767dfd8e8ca893ea37a3a16f75fb4826b564114544d92055fc08a926883b2" - }, - { - "source": "models/cuda/tests/test_glm53_kda_decode_workspace.py", - "path": "models/cuda/tests/test_glm53_kda_decode_workspace.py", - "source_sha256": "1717caed84ae5c26e947a9ec8e9f4ea8db415286ce66461dfd1d9d10a4c54f4c", - "import_sha256": "14ac082bd3c350b3e76e96eaf37b97fff2ce521b68ef9d7e13dab3d2231a8a14" - }, - { - "source": "models/cuda/tests/test_glm53_kda_state.py", - "path": "models/cuda/tests/test_glm53_kda_state.py", - "source_sha256": "3949846798be0f729944b40debdee754c8a6600279440ed0f6ffe7b85facac3d", - "import_sha256": "6b24778266939c712be1157f45bf40fa3784690d66f711263c930417d4bc8af9" - }, - { - "source": "models/cuda/tests/test_glm53_kda_workspace.py", - "path": "models/cuda/tests/test_glm53_kda_workspace.py", - "source_sha256": "15258fbe9a5abfd4772b041f7c82419b71fe906369fe63007b77a55aec61ffbf", - "import_sha256": "a5604af5df02261a3e61677c416c48c30340d9ef37258d8f5ce05e99d4324499" - }, - { - "source": "models/cuda/tests/test_glm53_model_gpu.py", - "path": "models/cuda/tests/test_glm53_model_gpu.py", - "source_sha256": "62453e1e96b6ab48cd511426926c0e8ebd3a61fa7f15f33ca9737fd8f47aa500", - "import_sha256": "e2f7183bc20f5b15f577d30e5d01a7f253ba3fbb88f5b3d9c717d16c13dcc2c2" - }, - { - "source": "models/cuda/tests/test_glm53_normalization.py", - "path": "models/cuda/tests/test_glm53_normalization.py", - "source_sha256": "37af06985f77dcba3f6e1c96c71798118a8d4eb2c303690755a890775fa47fed", - "import_sha256": "f2e329b8fb6b45f0262894055b831101c84c8218802d9684986e3a9081f93944" - }, - { - "source": "models/cuda/tests/test_glm53_projections.py", - "path": "models/cuda/tests/test_glm53_projections.py", - "source_sha256": "0cd8a3cf38ef678c738d76796c7118a7d662156b9b9b11093dc5e86fb71eea05", - "import_sha256": "d3b242ca98be4a091cfac7b2a0e2c0f0f3d1c379a42e796c7c8f57c650558a10" - }, - { - "source": "models/cuda/tests/test_glm53_runner_limits.py", - "path": "models/cuda/tests/test_glm53_runner_limits.py", - "source_sha256": "6e62ba1381e05532678b74347dce8c072140157691b089afd3fc35e42b694f10", - "import_sha256": "07549cb4440d812e4927af6d1518cf70c7b6ecc621bacdfc94838c1d848a3bca" - }, - { - "source": "models/cuda/tests/test_glm53_speculative.py", - "path": "models/cuda/tests/test_glm53_speculative.py", - "source_sha256": "bd451a863455c526389c73dddf7a7f858c58fb89c0bb9525c0ec122617a0c51a", - "import_sha256": "e5e7261de338132ed08c0f01e0dc019c65966930152793568a5f960eff85e46b" - }, - { - "source": "models/cuda/tests/test_glm53_speculative_cuda.py", - "path": "models/cuda/tests/test_glm53_speculative_cuda.py", - "source_sha256": "7b7f4eef17b6f889a051f93a2ed4eac72599a52b5585cbb05d8c64cefe3fd313", - "import_sha256": "911d5ab60e1fc7422a21aede07de1f538ab85f3f052619474536cef08dff19b9" - }, - { - "source": "models/cuda/tests/test_glm53_state_fusion.py", - "path": "models/cuda/tests/test_glm53_state_fusion.py", - "source_sha256": "2444ed0be70217086ce0deffaef68fbe43e3cc60c8a916c0410e8812c2dea6da", - "import_sha256": "282501afecda7f94f6cdd4aec8547261bf55a6d3f87fdc5515437d69ebf883d9" - }, - { - "source": "models/cuda/tests/test_glm53_vision.py", - "path": "models/cuda/tests/test_glm53_vision.py", - "source_sha256": "d4f81c60164e56701563014ae55c60e6a86c79146129de036de49847a95a0896", - "import_sha256": "83503e3e62a63e5c6ecd1215768e7f695b29e80048220b1bc0cef88605e2a91c" - }, - { - "source": "models/cuda/tests/test_gpt_oss.py", - "path": "models/cuda/tests/test_gpt_oss.py", - "source_sha256": "bfb4b4183a68dc1d6f391e2b890a1a654c23d098c7fd2472cb077ef5fb3cb0e7", - "import_sha256": "4e52b953d0a7eebc7f797f118f4523eddf988e4a1d5e61e0be235d4daf579940" - }, - { - "source": "models/cuda/tests/test_gpt_oss_cuda.py", - "path": "models/cuda/tests/test_gpt_oss_cuda.py", - "source_sha256": "789bc3b1a9c75efc55bd03a052da57ad1936b35b29201581fe63f5b8f2d30ca1", - "import_sha256": "04848195a357851af1d8fe233dd39489bea16e0b1d92cdacb21aba03c56d6f81" - }, - { - "source": "models/cuda/tests/test_gpt_oss_distributed_runner.py", - "path": "models/cuda/tests/test_gpt_oss_distributed_runner.py", - "source_sha256": "00f74a733d27854ca1dd1d095acd53c8566a44c6873b22137f4e4abc1397aeda", - "import_sha256": "8d068cc9ce45c9f61f936cb458b715532c10b4c7cf99522968cf22870365388b" - }, - { - "source": "models/cuda/tests/test_gpt_oss_expert_parallel.py", - "path": "models/cuda/tests/test_gpt_oss_expert_parallel.py", - "source_sha256": "00bb350ff1ca7417e3b3af291f4b2aff64313041370254e7ee79534a35fb695b", - "import_sha256": "414395d63e3db900e9301f4cd7910a8afa6f6b36ae342c64d415e48d4378fea4" - }, - { - "source": "models/cuda/tests/test_gpt_oss_flashinfer.py", - "path": "models/cuda/tests/test_gpt_oss_flashinfer.py", - "source_sha256": "d7079daf40f7da5f5b7d01c2f54f9e98f38d2c1c2db4c428ace284c702866f5e", - "import_sha256": "978109e316681013315249801fa2b95d1192655a610563b64639818677b0d351" - }, - { - "source": "models/cuda/tests/test_gpt_oss_flashinfer_gpu.py", - "path": "models/cuda/tests/test_gpt_oss_flashinfer_gpu.py", - "source_sha256": "88c9f5189720930c73cae49487d6a7d8b1439cb6d62bed92ea8471cc90469bb7", - "import_sha256": "7aa12455bc98d1d7810349ece1a341fb4d519b0f2c9f7d203eab8b88ca37b71d" - }, - { - "source": "models/cuda/tests/test_gpt_oss_packed.py", - "path": "models/cuda/tests/test_gpt_oss_packed.py", - "source_sha256": "20baeac1028e99d64b89f66cd192f0de5c4ee57ff7b7472cd29786860471d204", - "import_sha256": "d3750d77348a678463c75911f12acb7ff48d13da4bdbd4a1c8bf45079024eade" - }, - { - "source": "models/cuda/tests/test_gpt_oss_prefill_projections.py", - "path": "models/cuda/tests/test_gpt_oss_prefill_projections.py", - "source_sha256": "6d98488f07d35d680568102b9a2a64a354bae6a73ffe847524f608a5520ef978", - "import_sha256": "fa5ba27e3bd9fade1f446f02fec71549f2a583c9fc90b3eebdb4bdfa002186f0" - }, - { - "source": "models/cuda/tests/test_gpt_oss_routing_gpu.py", - "path": "models/cuda/tests/test_gpt_oss_routing_gpu.py", - "source_sha256": "675c81d5f3594ee638fe0b9476bf6c0c75991fa2d88b22a427f46f45ef62bbf8", - "import_sha256": "a49efd7db2c74a4e9c0ed9fb33a65a91ef860080ce451d26edf7669b4a19088d" - }, - { - "source": "models/cuda/tests/test_gpt_oss_weights.py", - "path": "models/cuda/tests/test_gpt_oss_weights.py", - "source_sha256": "3f44e1cf0542eec2538e9e1c1ed2fdf93aeb15db87fe4d33cdafc2ff02ecc421", - "import_sha256": "f99125dc1e80d271b4eb9945ce0b288a2f0739c3abc1d08e5d2e8227463db18a" - }, - { - "source": "models/cuda/tests/test_inkling_attention.py", - "path": "models/cuda/tests/test_inkling_attention.py", - "source_sha256": "1670ed119e50b1ac00458ab709f678be51d5e497f67875c5ad9674b2e2d33e5a", - "import_sha256": "3cd33ad48b9665e1d3b184c7d9ba41e3566f0a0796d6c5ed1d084e5f803a5dfb" - }, - { - "source": "models/cuda/tests/test_inkling_cache.py", - "path": "models/cuda/tests/test_inkling_cache.py", - "source_sha256": "680c666c81a82ec25eb116c05100fab7eddbc261579ef0d7ecfae01895720fa5", - "import_sha256": "66b872995ea19a2653ebd47a6b3a3fc3e125ece0f0eda63a573d8c339fefa70b" - }, - { - "source": "models/cuda/tests/test_inkling_codec.py", - "path": "models/cuda/tests/test_inkling_codec.py", - "source_sha256": "2979931085ccf41b40c888d555ffc7c22448c3ef9f10b48d1ffbfd87bc60dafd", - "import_sha256": "bb4decac8ccf434c5b4c0f600eac9e922c6d5b49e5129db9069df632d904340c" - }, - { - "source": "models/cuda/tests/test_inkling_config.py", - "path": "models/cuda/tests/test_inkling_config.py", - "source_sha256": "daef98c1b6668a7e6aa0c514b339d4824ea976e1b2ff0de75de91ec27402f93e", - "import_sha256": "30c6ed98a83e75c8b930cdc4506db1249a407e85fc3a7cbed812c7736a8dc64b" - }, - { - "source": "models/cuda/tests/test_inkling_media.py", - "path": "models/cuda/tests/test_inkling_media.py", - "source_sha256": "d523e4cedccf30734d7814b6c0247005ae12afa4663c8e05265e28bb1d64e1be", - "import_sha256": "0d2cf2170f47d49400565c895234d37eedefa0903311da4951ffa0a48ab27abe" - }, - { - "source": "models/cuda/tests/test_inkling_model.py", - "path": "models/cuda/tests/test_inkling_model.py", - "source_sha256": "34acc08b6e28203e17eb20fa0ae300e9a27f55cb4856c8cdf624842f21c21311", - "import_sha256": "6ef09c38c94ef41a66f89bcfdec96563a506cdcfaba30076736f4fb3bffa686b" - }, - { - "source": "models/cuda/tests/test_logprobs_cuda.py", - "path": "models/cuda/tests/test_logprobs_cuda.py", - "source_sha256": "b8a55af17f637144ee005fbfd14b836c9eb6347c2806801a21fe200e70d8f11f", - "import_sha256": "638e07a57132fd4fc2fd0392a6a6b30581844e23593ab40157654b35f360149c" - }, - { - "source": "models/cuda/tests/test_media_feature_reuse.py", - "path": "models/cuda/tests/test_media_feature_reuse.py", - "source_sha256": "9b0ce4f948184bb73a2a16585a8874ffc1dd9795740648eae11022b1ba8086b7", - "import_sha256": "ccc61f7d32a7aa9b9cb9421a3a480eed2f56a0dc7f317dab1d62f20850428250" - }, - { - "source": "models/cuda/tests/test_qwen3_5_cache.py", - "path": "models/cuda/tests/test_qwen3_5_cache.py", - "source_sha256": "fe758f39ed3d7dc4f235ba662a5c5200404dbed33aabe3e30ff12e9d7a72f417", - "import_sha256": "d9488f688856ff842149d2a9c5d6895be43fc5c7f3cacc78af4fe2059424a191" - }, - { - "source": "models/cuda/tests/test_qwen3_5_config.py", - "path": "models/cuda/tests/test_qwen3_5_config.py", - "source_sha256": "bd524cd081d7cab95ff544e56cd4e26a6d1dab97fcd1cd873fe72b08a5ba34f9", - "import_sha256": "e1da214a8c90e5a080acefabae6fd3b2467dbaf26f1a051ee9323638f3afd811" - }, - { - "source": "models/cuda/tests/test_qwen3_5_gdn_decode.py", - "path": "models/cuda/tests/test_qwen3_5_gdn_decode.py", - "source_sha256": "9c88ab2625e77d78ec32f10df033d7f9ec6a378762dce0352018da8d3b2fc704", - "import_sha256": "5b7a7d13ee19df0d8d2655d9785b3c034ce0af4f34fcdb35b3b8efec70916506" - }, - { - "source": "models/cuda/tests/test_qwen3_5_kernels_gpu.py", - "path": "models/cuda/tests/test_qwen3_5_kernels_gpu.py", - "source_sha256": "6c06a549e9731b168f3b84fa3113ace6172bc28083558cf2836d47e8a7d0246e", - "import_sha256": "3c65e1e686d4a15c0ce5fdc099e13e78d6b189516f1367db1940a078d6b1efcf" - }, - { - "source": "models/cuda/tests/test_qwen3_5_media.py", - "path": "models/cuda/tests/test_qwen3_5_media.py", - "source_sha256": "138a810946190e8114920ae00fe44ed9b587e79098917d220ee81ad6f3d5ee9a", - "import_sha256": "10f870c1dfa780f7abb63b8eb847e868f422d46831e8ad2a7b10fbd9ede275d4" - }, - { - "source": "models/cuda/tests/test_qwen3_5_model.py", - "path": "models/cuda/tests/test_qwen3_5_model.py", - "source_sha256": "aba107ea56787841187b56f49f29d1e9a86f76ff3f144dac73f30ea0b7f64ee6", - "import_sha256": "f06876b58b93d088e42cf109ba3f51f909de9c1bac57ab7fa48436eb569d627c" - }, - { - "source": "models/cuda/tests/test_qwen3_5_tensor_parallel.py", - "path": "models/cuda/tests/test_qwen3_5_tensor_parallel.py", - "source_sha256": "dc5f95e9343ff5b9ec4a466ad414d351b260311ab7abd74607c1186b64fbb494", - "import_sha256": "c1a9fda0a7d6f7750b694c910d5b676278ae7eafe98d46b73f298aaa7840dccd" - }, - { - "source": "models/cuda/tests/test_scoring_runtime_cuda.py", - "path": "models/cuda/tests/test_scoring_runtime_cuda.py", - "source_sha256": "4a7acafa5329fbe63cf8f35615d3a07bed88d163f2010862894553575957aba8", - "import_sha256": "8966403fae8f397380ca19a8b1559262581fae29a5e2610bcbe889392d4ab263" - }, - { - "source": "models/cuda/tests/test_verified_sampling_blocks.py", - "path": "models/cuda/tests/test_verified_sampling_blocks.py", - "source_sha256": "4201c6d9416245edea5529693df5dbfcf944b7056ce4bf3f35531ce98ffd708b", - "import_sha256": "cfff0975400c6adfdd676c24a5c20e6d6b4f15d931221c20046007dd32aefce1" - }, - { - "source": "models/cuda/tests/test_vision_attention_cuda.py", - "path": "models/cuda/tests/test_vision_attention_cuda.py", - "source_sha256": "2c9e0157d953e54478ec2ccf5c58137c14a02f4961b73b238b23449cd6f9b640", - "import_sha256": "205456ef30f53f0cc080be77b7ae3c682154a89ceb8190f155de62e5a6899091" - }, - { - "source": "models/mlx/.gitignore", - "path": "models/mlx/.gitignore", - "source_sha256": "c6b2660f1915b3c98f0874311ac5f23f0c7ebbab51cafd17fdbc83e708454474", - "import_sha256": "7a630dc5cb8d2e7e6a4416a9f0f9ad6e10c30bc8021c4c247924e18a42e98293" - }, - { - "source": "models/mlx/LICENSE", - "path": "models/mlx/LICENSE", - "source_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", - "import_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30" - }, - { - "source": "models/mlx/MANIFEST.in", - "path": "models/mlx/MANIFEST.in", - "source_sha256": "261c1efdc0fca88eefb31c72613fe6d59ee3a40ea7399be70a70c9d9576a0c0a", - "import_sha256": "261c1efdc0fca88eefb31c72613fe6d59ee3a40ea7399be70a70c9d9576a0c0a" - }, - { - "source": "models/mlx/pyproject.toml", - "path": "models/mlx/pyproject.toml", - "source_sha256": "cd908361104b5590ace82cd76827eab3f991b575e1aba985c91b975fb632a14d", - "import_sha256": "e26f51deb8de1ae902376beeaab667d99058719e2be6e295483d8868e1bb5970" - }, - { - "source": "models/mlx/resin_models_mlx/__init__.py", - "path": "models/mlx/orbitinfer_models_mlx/__init__.py", - "source_sha256": "1f1f94246cbd4380b13b3e038021ebe68e87d00492e09a4815a3aa39a0a7d6a0", - "import_sha256": "1f1f94246cbd4380b13b3e038021ebe68e87d00492e09a4815a3aa39a0a7d6a0" - }, - { - "source": "models/mlx/resin_models_mlx/_native/README.md", - "path": "models/mlx/orbitinfer_models_mlx/_native/README.md", - "source_sha256": "bb552dbe656a80e4bd61b25b3685f065cfc3a5b5fa026a3431b34866e4116797", - "import_sha256": "bb552dbe656a80e4bd61b25b3685f065cfc3a5b5fa026a3431b34866e4116797" - }, - { - "source": "models/mlx/resin_models_mlx/_native/__init__.py", - "path": "models/mlx/orbitinfer_models_mlx/_native/__init__.py", - "source_sha256": "f2fd509f8458f5a8289b537bd2390d15b7de8c02a91c8ea300f554c5478f6ae6", - "import_sha256": "fdd955c8d3c93ab16aa12350f39b2fefa174040e01bd5eb3ede0a3f650fc48a5" - }, - { - "source": "models/mlx/resin_models_mlx/_native/__init__.pyi", - "path": "models/mlx/orbitinfer_models_mlx/_native/__init__.pyi", - "source_sha256": "184b42ff055f1a405bdfa74828480402462d08b9a6d7640459285f7579aa5213", - "import_sha256": "ac05ab2c8e27da95d629124cf5b6c2f71d844508bae0d5e4213731748cf7bc5a" - }, - { - "source": "models/mlx/resin_models_mlx/_native/attention.cpp", - "path": "models/mlx/orbitinfer_models_mlx/_native/attention.cpp", - "source_sha256": "8564ef1e7579b6a25e556e87e9bf1c808934f2c565186da854d86b417fca8ada", - "import_sha256": "65d644e9aa7969f48ab1825bec6fedb1f93fe7b73e4ae395e5724bb4cf41385e" - }, - { - "source": "models/mlx/resin_models_mlx/_native/attention.metal", - "path": "models/mlx/orbitinfer_models_mlx/_native/attention.metal", - "source_sha256": "fdf01c7187e82225f255b6872e36af5c6dd854fb419e9c7c6d7460f38f8cb574", - "import_sha256": "7cd97ebfd181f0c7afbc4281d0cdaddaf6e75d18417fb889b246a08004591577" - }, - { - "source": "models/mlx/resin_models_mlx/_native/attention_kernel.h", - "path": "models/mlx/orbitinfer_models_mlx/_native/attention_kernel.h", - "source_sha256": "611c3272f4c0e18196485fb33be8459569774d5a5460e31c56d9eb362235f234", - "import_sha256": "a33e5644aa4606b1847b20fe4f0004717443561891f558adfc42432b00d99589" - }, - { - "source": "models/mlx/resin_models_mlx/_native/cache.cpp", - "path": "models/mlx/orbitinfer_models_mlx/_native/cache.cpp", - "source_sha256": "b1cceb2102d2fe0f9d199c505c498154b9a13a1107c98c9bcc6a23a491391df5", - "import_sha256": "021f13d5e7c72684ea1af780aa5380abf677ffd54d3debe961934502ba95c0dd" - }, - { - "source": "models/mlx/resin_models_mlx/_native/cache.metal", - "path": "models/mlx/orbitinfer_models_mlx/_native/cache.metal", - "source_sha256": "e35b0d3da24b11d6e60505ecd94dd5cb80d13f5626686046778f63757d3c0426", - "import_sha256": "2e748749bf7b6df14d7bcd7461057f5d17490fad7974075e7c7674434ed8788c" - }, - { - "source": "models/mlx/resin_models_mlx/_native/common.h", - "path": "models/mlx/orbitinfer_models_mlx/_native/common.h", - "source_sha256": "e4d1f186db1d1f277228f3a2ea40f16de12932802745bbe08f46adff124e2e68", - "import_sha256": "b07736ccce38d2ff1c6ddcd85290724a6c6bf0a6d9cd5d159d52ece86c39a2cf" - }, - { - "source": "models/mlx/resin_models_mlx/_native_cache.pyi", - "path": "models/mlx/orbitinfer_models_mlx/_native_cache.pyi", - "source_sha256": "0c8eb9486f6996b12229d5d9f75e6125a6f0a0330c04a959bf4f7a8333a309b6", - "import_sha256": "0c8eb9486f6996b12229d5d9f75e6125a6f0a0330c04a959bf4f7a8333a309b6" - }, - { - "source": "models/mlx/resin_models_mlx/cache_packing.py", - "path": "models/mlx/orbitinfer_models_mlx/cache_packing.py", - "source_sha256": "adb3eb2c76adb6196fcf98675b0ad23f5177e5c907d97b6a9072aaf83bb6c28b", - "import_sha256": "815b690696ba0523c0a924f9173597daf849e0b0bc4731d6e9f8a8c5c1a90443" - }, - { - "source": "models/mlx/resin_models_mlx/paged_attention.py", - "path": "models/mlx/orbitinfer_models_mlx/paged_attention.py", - "source_sha256": "c0a773be20e1b765a8be45bf2f8998cac395fde22e088720fed7eca1e5377961", - "import_sha256": "7b5c3c4c182853bcbf05daadfda13317655a554971a9eb72839a3c3d98ee30ca" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/LICENSE", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/LICENSE", - "source_sha256": "d31a1955307c2d52f55bc9d2d0569558d143d0571a2faf525a1b57b1458e3e59", - "import_sha256": "d31a1955307c2d52f55bc9d2d0569558d143d0571a2faf525a1b57b1458e3e59" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/README.md", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/README.md", - "source_sha256": "d2e50e162bd6dcb879f6e4e04fe77b93cf9224bf552c14da6c8294cc28cea34a", - "import_sha256": "c5b9b93155cafa41a9b6a06e3072ef0133d25a084354a8039a073f27100326a0" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/__init__.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/__init__.py", - "source_sha256": "6c47179f311c1f7300462aa5db514f4828685db8799ae21f53f3758dada85265", - "import_sha256": "c76fed50f7779034cd1821a71b4989a43f761c6b46f47c6fff14bf3f17ff69b4" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/cache.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/cache.py", - "source_sha256": "b2710464b931ba43d8d84850a59e928429e13de4ef09bfb545715aaad18592dc", - "import_sha256": "c0a4a2c3164797de013fc0cd877178bc81124a1351ce8ff015c4f338b1ee1af8" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/checkpoint.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/checkpoint.py", - "source_sha256": "47817baff1436bf778a25a63fbd14fc2e2c557850efcdb79e76deaf2e4558a63", - "import_sha256": "ee274b770ba59263b6a5069983fbadf026dac291e5182095533c78bf8c7073d2" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/fused_conv.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/fused_conv.py", - "source_sha256": "321b56374ac9a3af2dff904d765528062f6cb2d992ba25e4b96dd8c632eb5880", - "import_sha256": "feafec57f5215cb74d7bce95f317125d98ead4dd3d11c0ab9d4d13f7c61293bb" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/fused_norm.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/fused_norm.py", - "source_sha256": "822a85b76b605c7a30e866afcceae2d591074f5f31eee1c8bac41fd5f5e11567", - "import_sha256": "d6aee2da0b7277767f89fc492a3be91c1fd577815cdc48f3c622f190f57ede4c" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/gated_delta.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/gated_delta.py", - "source_sha256": "237fa395374d45f39785197e32099d899026adfded0cbbcb69b36d8b33d7c7b8", - "import_sha256": "237fa395374d45f39785197e32099d899026adfded0cbbcb69b36d8b33d7c7b8" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/media.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/media.py", - "source_sha256": "6d4f95d20eedc4f85984d1431b16ddf61af8d1e0c004f837a90fbae3d762e66d", - "import_sha256": "a0d2308a204f6a450b0ee7133cc451dc9dab5258a6b621eaa584558fa1193a90" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/model.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/model.py", - "source_sha256": "8235485170dac439817769c5f83d73e7c07b58a53d42151fe05fa6ec18ded1dd", - "import_sha256": "66de9d232aaac10921ac36302a42d6ab5b43fc1ea8cd398d60b396c1c8544d81" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/normalization.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/normalization.py", - "source_sha256": "6ebe0537e3504964a4e861e9e64fa6bd6ddac18926f9f6a79b72fc906c16e284", - "import_sha256": "6ebe0537e3504964a4e861e9e64fa6bd6ddac18926f9f6a79b72fc906c16e284" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/runner.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/runner.py", - "source_sha256": "96cdd62eb831640f10833c7b5644d68a6ed41fe8b69ba303112cca815774a321", - "import_sha256": "484cb99f852572e1362df912f5dcf0aa28ffbb7302b65138137cdebfb597f6fa" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/state.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/state.py", - "source_sha256": "c588fd95c3b4429d12569d70cf10647ffc017a750eec1b75fa913d43864b5586", - "import_sha256": "4cfc181c511375de6823bdfe3a6257417989c7c4f357bdb753897cb965036628" - }, - { - "source": "models/mlx/resin_models_mlx/qwen3_5/vision.py", - "path": "models/mlx/orbitinfer_models_mlx/qwen3_5/vision.py", - "source_sha256": "7f53a822120024348dc33fcc08dc3419ab6e603c2e3104a0af11b1e76e085586", - "import_sha256": "6413359eea2e4b44ffada482dceef92132b5c5cae4a1477d8bb32df48d8bd08f" - }, - { - "source": "models/mlx/resin_models_mlx/sampling.py", - "path": "models/mlx/orbitinfer_models_mlx/sampling.py", - "source_sha256": "8f9b81e925cfbad6f65694371b036511eb57ae64feb8b24b012357276c64ef83", - "import_sha256": "bd3026506c7938b13251cb9ae4b22448c54233f51762e4f4357c0005ad77d906" - }, - { - "source": "models/mlx/setup.py", - "path": "models/mlx/setup.py", - "source_sha256": "619d01cd9798364537da7d6c4a3528bd37498ceeb3030143de6560e97381fe52", - "import_sha256": "45ffa38a5dcf7ec4d2b4a1ba0cb5651d00936a54e87afdde779be8aa79a27b3f" - }, - { - "source": "models/mlx/tests/test_cache_packing.py", - "path": "models/mlx/tests/test_cache_packing.py", - "source_sha256": "ad2a1f1d436b3e517998e3d61406e0d4cfc40306eba7b2adde6da55d82ee83fb", - "import_sha256": "f1c764f754d3c3481aecd64751cb5bae5bf28e965d8ecd780ed6f805e87c31c1" - }, - { - "source": "models/mlx/tests/test_fused_conv.py", - "path": "models/mlx/tests/test_fused_conv.py", - "source_sha256": "e590689f0e4195440ced94916e3627c08d5f040ff5cc547d728dc24be2d9dde2", - "import_sha256": "de2aa90d64de1fb416158ac28d44975fdbc5a9dfe74415793c981b6c009c7919" - }, - { - "source": "models/mlx/tests/test_fused_norm.py", - "path": "models/mlx/tests/test_fused_norm.py", - "source_sha256": "ff2cb7f89f2e503ddb276d4ebc03ae08978df4f8e7852dfc0e781c706e5c3960", - "import_sha256": "6f017f9ecf19b9bb8d72deb878a549b671de925cc5ac0188c9b60c5043e6604f" - }, - { - "source": "models/mlx/tests/test_mlx_sampling.py", - "path": "models/mlx/tests/test_mlx_sampling.py", - "source_sha256": "c348d234c776ce5f4f586282de145b064262905a2a8e1bb3b616923d5eea9aa9", - "import_sha256": "29ceeb2c943d933f2243d8c4691afb9f26ef6256dca907760be99daad12d0002" - }, - { - "source": "models/mlx/tests/test_native_cache.py", - "path": "models/mlx/tests/test_native_cache.py", - "source_sha256": "6458cf13dafc8ac0083a29a37a1fb14c585ba3a9a7aeabe0018962b8239f6a0e", - "import_sha256": "c3b969a891cd48e93599a7661d9b165e70d9ec8fb173ff5195330d3e4bc6f082" - }, - { - "source": "models/mlx/tests/test_native_cache_ordering.py", - "path": "models/mlx/tests/test_native_cache_ordering.py", - "source_sha256": "82058362244566c2548297f08aa287ebc0b2a3342a374c8d4b0d219464fb1cfd", - "import_sha256": "57aec426e46db5154a5cc59e264047abaf50f120e345e2b48d6c691508a46332" - }, - { - "source": "models/mlx/tests/test_native_cache_rows.py", - "path": "models/mlx/tests/test_native_cache_rows.py", - "source_sha256": "6b0e33dc5ede8b1a8984e9a8bcb8afaf67fefcdb2c5c56f522e78c2242609f05", - "import_sha256": "c9bad9400d3cb5f2353e8e3fd32a77afa0aeed2f713b2235d708eb4bd7b2824c" - }, - { - "source": "models/mlx/tests/test_paged_attention.py", - "path": "models/mlx/tests/test_paged_attention.py", - "source_sha256": "3314c06aad935e0ef8313fb8cb7dd8d182b365ee8cda3f73d6bc38a35149ab3c", - "import_sha256": "29ea47fa33723c150eee298f8138aa7f724eb2cb57740ec4a37d6f59818e59fa" - }, - { - "source": "models/mlx/tests/test_paged_attention_layouts.py", - "path": "models/mlx/tests/test_paged_attention_layouts.py", - "source_sha256": "7811a1f6aac69c9d256db3df5e091e2d2fa87ac006437b379ec565f3c034d204", - "import_sha256": "1cd59bd297705043c28ceb617dd171200505563d7535c6d87c703f7f303f4548" - }, - { - "source": "models/mlx/tests/test_paged_attention_merge.py", - "path": "models/mlx/tests/test_paged_attention_merge.py", - "source_sha256": "82e91a91636a98be1830509c091b73b3a5e4af25d8980e5981b0cc4d1323219b", - "import_sha256": "72b4c3f47c677b6444418303f032197ab16be450c3a798d67599042c43ac1f8f" - }, - { - "source": "models/mlx/tests/test_qwen3_5_media.py", - "path": "models/mlx/tests/test_qwen3_5_media.py", - "source_sha256": "49bcbeefcbfae35f4c02f6e73ba54280776548125b435a886a23bb1618598489", - "import_sha256": "554f15855fa26548cb544c731c01868c41912c6b3d2ed54e643eb160815d553f" - }, - { - "source": "models/mlx/tests/test_qwen3_5_mlx.py", - "path": "models/mlx/tests/test_qwen3_5_mlx.py", - "source_sha256": "a970012b6584c89a1bd5d94624596d86d2636674fb1b5ce916f2fb7caa5801d9", - "import_sha256": "43073e23c80da2acc16c2782e47f4a291e51edd8789c192e335415465da96a28" - }, - { - "source": "models/mlx/tests/test_qwen3_5_vision.py", - "path": "models/mlx/tests/test_qwen3_5_vision.py", - "source_sha256": "e3b53dc160b81419afdc592b49bd32facbed0edc4afd2e696b017c8cbd405cea", - "import_sha256": "fe0e4d78fdd29d2f95cb1439dfcd7faaf7fb6a02bfb54dcbb67128d7ec3d477f" - }, - { - "source": "models/torch/pyproject.toml", - "path": "models/torch/pyproject.toml", - "source_sha256": "654706d8c00b231d704e2d67a11d9bf3a33977469804d6ed1448e1c04ce6e98b", - "import_sha256": "977930b9e51d89148a20342d371e91d102dedd5024e064e43f32c43ed2acbc82" - }, - { - "source": "models/torch/resin_models_torch/__init__.py", - "path": "models/torch/orbitinfer_models_torch/__init__.py", - "source_sha256": "494ba50d6c2d17dfbf7a8b0b39a53f9981626afe33ccc0a3cd131b439edf8618", - "import_sha256": "494ba50d6c2d17dfbf7a8b0b39a53f9981626afe33ccc0a3cd131b439edf8618" - }, - { - "source": "models/torch/resin_models_torch/lfm2/__init__.py", - "path": "models/torch/orbitinfer_models_torch/lfm2/__init__.py", - "source_sha256": "595b8005d46d2a23f288b17d3314212b484ca595855a612e97f78c2dd8963e72", - "import_sha256": "0570e3537f5b8407953cd4b5589061ad2aae2484593d87169244bf1cdcba218b" - }, - { - "source": "models/torch/resin_models_torch/lfm2/config.py", - "path": "models/torch/orbitinfer_models_torch/lfm2/config.py", - "source_sha256": "f429b9f98a84fe39c2acb2dfb4a917aa5f02feb4ae6623ee3538d4ba05e1b7ab", - "import_sha256": "f429b9f98a84fe39c2acb2dfb4a917aa5f02feb4ae6623ee3538d4ba05e1b7ab" - }, - { - "source": "models/torch/resin_models_torch/lfm2/model.py", - "path": "models/torch/orbitinfer_models_torch/lfm2/model.py", - "source_sha256": "2d0082238e523ba22fda6b2e8c10bdca4b920fdc7c54cc5d3f6fe37d26e45689", - "import_sha256": "1f6bc8f4277f25991b7da026d822dfcd2d560da612358ccec5d8de83fa8a4c92" - }, - { - "source": "models/torch/resin_models_torch/lfm2/reference.py", - "path": "models/torch/orbitinfer_models_torch/lfm2/reference.py", - "source_sha256": "7ad95d59571429534e6c8259b1469a43b8dd147aeffc74e2f089a2ef2f4a49a8", - "import_sha256": "4a3f0820f964ed29e6dc9ceefaff5a610b13df1a6f82ea70415aed93a7efed36" - }, - { - "source": "models/torch/tests/test_lfm2_torch.py", - "path": "models/torch/tests/test_lfm2_torch.py", - "source_sha256": "a20adeff8d4fe465ad5f1b02cf1c709f3b33868f9d6b30a729454ea594a058d5", - "import_sha256": "35f7033e3a7c079a579cea7a96de8845816dcc4850ed9f244ef613876c3e960a" - }, - { - "source": "pyproject.toml", - "path": "pyproject.toml", - "source_sha256": "f14674fd6a319a14e5a869c8c71a859603af28a72b84fcead4b32e97c5e33157", - "import_sha256": "de98c5bfc2b4d80805a97b07b907ab72f2c53b9388aaefcd30bacbfbdf50973d" - }, - { - "source": "resin/__init__.py", - "path": "orbitinfer/__init__.py", - "source_sha256": "a7b62b04fa3a0e91418e6467b8ba3ad1d18a7dc3b6d07b0f2c3c5ddb56c6013b", - "import_sha256": "cb91929df47c9c4195cdbf7140c37d63abf88d7206fc840c0998512a5268638a" - }, - { - "source": "resin/_cache_storage.py", - "path": "orbitinfer/_cache_storage.py", - "source_sha256": "3af0d1ec651c0885ce98e3e0703305f730417bae113916618edab17dd540ed9a", - "import_sha256": "0cdb8e28cae7d22450866b4e113bc03157d05cd61fe714c4676d2e3bf7f5d9cf" - }, - { - "source": "resin/_resin.pyi", - "path": "orbitinfer/_native.pyi", - "source_sha256": "fa0ee6f04eaa02feb3d60fb711c95e864a5a602127104963a33476e44f0e4ebb", - "import_sha256": "061af34f713c1ecc28a569c6fe9a2d8baf17ded811eee56d98988da59fc97edc" - }, - { - "source": "resin/backend/__init__.py", - "path": "orbitinfer/backend/__init__.py", - "source_sha256": "b9682913ddb034d537ba1de1b1c2d4b2b955b961a02ff9b710761d14367d1609", - "import_sha256": "8676ef8910a0f362678fb25a739c665197b6fd502c96fb5a768e6b45cea2610e" - }, - { - "source": "resin/backend/balanced_routing.py", - "path": "orbitinfer/backend/balanced_routing.py", - "source_sha256": "c860cc492a9a66cdbfda38f077881b55975582c7ef41f06adee78dc8e1b59be3", - "import_sha256": "c860cc492a9a66cdbfda38f077881b55975582c7ef41f06adee78dc8e1b59be3" - }, - { - "source": "resin/backend/cache_storage.py", - "path": "orbitinfer/backend/cache_storage.py", - "source_sha256": "7c015b1def47bb639b7dbac60067161aacf5f6c3c401f939765d97961f23ddcd", - "import_sha256": "7c015b1def47bb639b7dbac60067161aacf5f6c3c401f939765d97961f23ddcd" - }, - { - "source": "resin/backend/capacity.py", - "path": "orbitinfer/backend/capacity.py", - "source_sha256": "bb95d6daac84c5dada4f6aa73607004ed3aebea18c64a2facb63debed59cd81c", - "import_sha256": "bb95d6daac84c5dada4f6aa73607004ed3aebea18c64a2facb63debed59cd81c" - }, - { - "source": "resin/backend/collectives.py", - "path": "orbitinfer/backend/collectives.py", - "source_sha256": "3957a9d9290e1955a6a32432ebcc05799f8c77032bfa2a7d4545f73c43ae8865", - "import_sha256": "3957a9d9290e1955a6a32432ebcc05799f8c77032bfa2a7d4545f73c43ae8865" - }, - { - "source": "resin/backend/command_transport.py", - "path": "orbitinfer/backend/command_transport.py", - "source_sha256": "43284572dc6594d389789ebd20623d8265eec25879fa0b959fe7e7ed5591e3fb", - "import_sha256": "2ecb1990a4c2d98365c55c3e5219aeaea5110f6b4eaa61e537e2de50860ef2eb" - }, - { - "source": "resin/backend/control_plane.py", - "path": "orbitinfer/backend/control_plane.py", - "source_sha256": "42640d8bdb7f0dd62be8f35b2d74bceac7831e134f4995c9996741d2444bf33e", - "import_sha256": "921fedba0ca7bb8fa260d9fa258db850b3a137a00a83518733154d20dd69b3ea" - }, - { - "source": "resin/backend/cuda/__init__.py", - "path": "orbitinfer/backend/cuda/__init__.py", - "source_sha256": "8f1ff0beb34d07dffaa6bd418e575410cf0b27ddf293febd83a8528697d4564b", - "import_sha256": "8f1ff0beb34d07dffaa6bd418e575410cf0b27ddf293febd83a8528697d4564b" - }, - { - "source": "resin/backend/cuda/_sampling_cutedsl.py", - "path": "orbitinfer/backend/cuda/_sampling_cutedsl.py", - "source_sha256": "5427b0333474c6e4d906591d198904e0fdaaf747988263693d6f2e25ca697f4f", - "import_sha256": "4715397fe6a611f7a5be75d8e887bff738413de64679f3d3af6b0dd089666d0b" - }, - { - "source": "resin/backend/cuda/_scoring_cutedsl.py", - "path": "orbitinfer/backend/cuda/_scoring_cutedsl.py", - "source_sha256": "9e283a2d27e43799ca6c680ca04378403370a6e81c1dc3993b768695bc1b68dd", - "import_sha256": "9e283a2d27e43799ca6c680ca04378403370a6e81c1dc3993b768695bc1b68dd" - }, - { - "source": "resin/backend/cuda/completion.py", - "path": "orbitinfer/backend/cuda/completion.py", - "source_sha256": "11624844ad4e0f071cbcb5831e5433eb573254b32b0bc22bebd22ae184ed40f4", - "import_sha256": "81570c66a4e73b45a02f674e764f5532f150c003b15c9e9235ee05b6169a6181" - }, - { - "source": "resin/backend/cuda/context_parallel.py", - "path": "orbitinfer/backend/cuda/context_parallel.py", - "source_sha256": "34035ceca315205d8655f4d2f457e3bf0b7f5d8a68d5ba2d7f01b6df8058cfbb", - "import_sha256": "7d77a67d019f33d323df5db577eb7dad91fd3f6c80792d7a78f8d3e693d10704" - }, - { - "source": "resin/backend/cuda/expert_parallel.py", - "path": "orbitinfer/backend/cuda/expert_parallel.py", - "source_sha256": "079f3f5617de24e9c7595c0848a22c25bb5edf28542b52c2f8a23ce09c4092d4", - "import_sha256": "9bd7854c3a4886fc3ba6e026c4282718812ced7d6738640d09033a1600c828c8" - }, - { - "source": "resin/backend/cuda/expert_parallel_kernels.py", - "path": "orbitinfer/backend/cuda/expert_parallel_kernels.py", - "source_sha256": "a7c709a91d6298437dd49d8a8292465e75ee27230f98d799163ea5bdfd6127b0", - "import_sha256": "a7c709a91d6298437dd49d8a8292465e75ee27230f98d799163ea5bdfd6127b0" - }, - { - "source": "resin/backend/cuda/failures.py", - "path": "orbitinfer/backend/cuda/failures.py", - "source_sha256": "9b0319d2ec24428d47c3b7bcfe5f3248c2aa6197f574869eb1f55134fdd404c6", - "import_sha256": "05f44479698bc56a9acaea0cc85d06a56a0040b0f514dc2db2cb398a2e1d560f" - }, - { - "source": "resin/backend/cuda/filtered_sampling.py", - "path": "orbitinfer/backend/cuda/filtered_sampling.py", - "source_sha256": "6cca9268e0c1505cb5ab3e93dabc3bd755a442a376bea5df7b9d1c3eb7640d46", - "import_sha256": "2f1e60171073a184425127276c7b839a524d63f4eb4dbcdd48c116473c30dce2" - }, - { - "source": "resin/backend/cuda/graphs.py", - "path": "orbitinfer/backend/cuda/graphs.py", - "source_sha256": "70ec63152c16905668d702eb4734e2df82e350de3fb05269dfbbb7a5f97f04c3", - "import_sha256": "33d17ffd8806962c8a61e99c953a51841634d6e8053017d251a757fc86df1b10" - }, - { - "source": "resin/backend/cuda/logits_sampling.py", - "path": "orbitinfer/backend/cuda/logits_sampling.py", - "source_sha256": "d85d8d49c0fa1a3d2512727bfc0ee54f13eac080909eb1b56bdb3666f136fa3b", - "import_sha256": "d85d8d49c0fa1a3d2512727bfc0ee54f13eac080909eb1b56bdb3666f136fa3b" - }, - { - "source": "resin/backend/cuda/metrics.py", - "path": "orbitinfer/backend/cuda/metrics.py", - "source_sha256": "16977eef274deff70c5b825f704a3ee19670fda942fd4b702a35b014a7b4c3b7", - "import_sha256": "ee5c98956af0ef10030ec9b995eee07d44ed58c70e71c73f0b3238aedc89c077" - }, - { - "source": "resin/backend/cuda/parallel.py", - "path": "orbitinfer/backend/cuda/parallel.py", - "source_sha256": "1d2e055a7c36ff60d6b7d719ec55a09aae4cef8da3052a210ed238a948fc1000", - "import_sha256": "3a018548701afd62ff62cf7ed48ac1afd7784615bcf70b56c55e1cad628755df" - }, - { - "source": "resin/backend/cuda/runner.py", - "path": "orbitinfer/backend/cuda/runner.py", - "source_sha256": "73003f8aa03e600bd456ae32de525dd66a78b7f1a4b2ca49294d83dc8d72c80a", - "import_sha256": "5a5e00765bedcfe231446dd4aa7f7dcb8b67ce2ccf009f50176bef9bad2b129f" - }, - { - "source": "resin/backend/cuda/sampling.py", - "path": "orbitinfer/backend/cuda/sampling.py", - "source_sha256": "d3eafe98e9f132f09342d06c62eb8860a1aa47347bbc44e53975ded391c57db7", - "import_sha256": "6d309695f904b48540c73b9175e74bb07ac82bcd7c7c7a426513625b8f03acfa" - }, - { - "source": "resin/backend/cuda/sampling_blocks.py", - "path": "orbitinfer/backend/cuda/sampling_blocks.py", - "source_sha256": "67f1e6e8f4bc882c016cc6e4aef4f6ee00cd2c2f29100f606fc02bc6d3e20974", - "import_sha256": "2c9293e9ab7706146fad752e287dcd58510964761d65a248c6ffbdd24049ad82" - }, - { - "source": "resin/backend/cuda/sampling_native.py", - "path": "orbitinfer/backend/cuda/sampling_native.py", - "source_sha256": "28504c7e5a6974032b6e3a0e6a878f66458dfcf4106f6a432a972401b41aeaf5", - "import_sha256": "997aa512a401e73abe42b46dabbf12f07925fa35baf0de52fe6d6422802c30d8" - }, - { - "source": "resin/backend/cuda/sampling_topk.cu", - "path": "orbitinfer/backend/cuda/sampling_topk.cu", - "source_sha256": "16c00489ea3bfd434e6ed736a3f6833758a9b653c623ff01bc3b33f88f082490", - "import_sha256": "9a7138d07ed2feee437960549fee38b46a932aea46bf0ab0fba23673a86dc18f" - }, - { - "source": "resin/backend/cuda/scoring.py", - "path": "orbitinfer/backend/cuda/scoring.py", - "source_sha256": "fe768c917c1630c599087dc3ae3f484be5a141e204326db5a608761a80970a7a", - "import_sha256": "3ee108225c5863fbe2331cc0e693870d6e66b3405aae88aaec12a7f3f50829bd" - }, - { - "source": "resin/backend/cuda/shared_workspace.py", - "path": "orbitinfer/backend/cuda/shared_workspace.py", - "source_sha256": "1cd7743f6277d677526659cf549a5f881ad08fe959910d0c17e012c7201619c3", - "import_sha256": "2a213da78b12177ea12348ebd5a32b8e71f4ea12a3c82bca1a91c88ac7dc51c0" - }, - { - "source": "resin/backend/cuda/slab.py", - "path": "orbitinfer/backend/cuda/slab.py", - "source_sha256": "74887c58a6e8a8b125f20c132a9f309d13f11e63232d06ddaf73b7fc9575fe2d", - "import_sha256": "3dbc1eb0c6eea02afd0a8e9c566082bef0b65d95cb54e54f58fe9a4e0f79cb68" - }, - { - "source": "resin/backend/cuda/staging.py", - "path": "orbitinfer/backend/cuda/staging.py", - "source_sha256": "08bb6c5a4358abb755b1c6674a6a1aa14eecd5c3a81126a07d41a0c05aafb611", - "import_sha256": "08bb6c5a4358abb755b1c6674a6a1aa14eecd5c3a81126a07d41a0c05aafb611" - }, - { - "source": "resin/backend/cuda/state.py", - "path": "orbitinfer/backend/cuda/state.py", - "source_sha256": "6a37d34e774f0dbc9b696f9e3676bd755243e44107faaec10f465757a87b233e", - "import_sha256": "fe52837a2fad4ec4d3a7b20ad032c9c8fb74dd43ad6a0a5246dc7acdc88e4a44" - }, - { - "source": "resin/backend/cuda/state_triton.py", - "path": "orbitinfer/backend/cuda/state_triton.py", - "source_sha256": "ed2d5c9e72d0c75cc76d48c9d190e475075fb722838223de09a2948ed4407a30", - "import_sha256": "ed2d5c9e72d0c75cc76d48c9d190e475075fb722838223de09a2948ed4407a30" - }, - { - "source": "resin/backend/cuda/tensor_graphs.py", - "path": "orbitinfer/backend/cuda/tensor_graphs.py", - "source_sha256": "48ef6150dd9d5df4fc21462eab0f4f44baa679c048c1f02f80fbd98b902be8d7", - "import_sha256": "991cbdab560b79a403d6d25ddd298e38f4988995ede7d9db3ff82940a3ce057d" - }, - { - "source": "resin/backend/cuda/vision_cache.py", - "path": "orbitinfer/backend/cuda/vision_cache.py", - "source_sha256": "f0fd361b5f581e8d51d3ccc4241390d0f2de90ec1d9f64e1c9b49c96ae59c4fc", - "import_sha256": "f0fd361b5f581e8d51d3ccc4241390d0f2de90ec1d9f64e1c9b49c96ae59c4fc" - }, - { - "source": "resin/backend/distributed.py", - "path": "orbitinfer/backend/distributed.py", - "source_sha256": "bd996b7536405ad6eb4ea24baab17633f7f24593fa91e148734d6f1d8f2d8af3", - "import_sha256": "48ef580916172bb9ddc92cfbfd43e8bd8a3067849081aa2461bff57164787286" - }, - { - "source": "resin/backend/expert_parallel.py", - "path": "orbitinfer/backend/expert_parallel.py", - "source_sha256": "e68d62c31d2a5fc317f0b1512260d8a4c6fd67b759f0c4a41eeba18bf416640b", - "import_sha256": "be7d1472aced89cd10b738c6fc6837684614f5e2ea7109629cd0ce2936ddc268" - }, - { - "source": "resin/backend/layers.py", - "path": "orbitinfer/backend/layers.py", - "source_sha256": "07ef8ff3813ea11eaf73497de92dd17d2272819a805297767166081b98db8837", - "import_sha256": "07ef8ff3813ea11eaf73497de92dd17d2272819a805297767166081b98db8837" - }, - { - "source": "resin/backend/loading.py", - "path": "orbitinfer/backend/loading.py", - "source_sha256": "fd746ccc211bb52708f44b7a071fdea5131f472e791781f50a414d3d36eaa2a9", - "import_sha256": "8ef5e1cb6131a5068e179e668d1e305fde36a39d202c83155545fe48ccd0d8c2" - }, - { - "source": "resin/backend/metrics.py", - "path": "orbitinfer/backend/metrics.py", - "source_sha256": "4b0216880c1a0ce8d5b3ab1e8c0588579574f77d323cd2a5f09e077e35e787d1", - "import_sha256": "20fbb19bfd51076d5a76fb50a27d8d600fbc673acea8c4cf9be43fc01d2c725a" - }, - { - "source": "resin/backend/plan_transport.py", - "path": "orbitinfer/backend/plan_transport.py", - "source_sha256": "52c631c4b5430b2385bafd2bd775fe54bd4cd39eab15831e909dd72cf5be5dbf", - "import_sha256": "8033a6e91def0ea1ea65a760a9ac134fa83a96ec5374e289589f408de42057b8" - }, - { - "source": "resin/backend/plans.py", - "path": "orbitinfer/backend/plans.py", - "source_sha256": "2da82634b769cf99c9206030fb6d68f5f5713c7d2a28978a4bc3bd03c86d865b", - "import_sha256": "2ca7dfe1e802f98f7a03d957485329ef8ddf798727e0a818f6e1fce995ed229b" - }, - { - "source": "resin/backend/resources.py", - "path": "orbitinfer/backend/resources.py", - "source_sha256": "31dde77fa84b6a7b7926d13763af6f8fd3519b5c82d41f30fe759c5df20d013f", - "import_sha256": "31dde77fa84b6a7b7926d13763af6f8fd3519b5c82d41f30fe759c5df20d013f" - }, - { - "source": "resin/backend/safetensors.py", - "path": "orbitinfer/backend/safetensors.py", - "source_sha256": "440626cc5a108dbf19fad372a0603f3624e7741b3eb6b38b31c63dd6c5511645", - "import_sha256": "440626cc5a108dbf19fad372a0603f3624e7741b3eb6b38b31c63dd6c5511645" - }, - { - "source": "resin/backend/sampling.py", - "path": "orbitinfer/backend/sampling.py", - "source_sha256": "ae0ebcd24fbb9794237fd7a9b154d4333c5aff91904319aca0ff25266ca46dfc", - "import_sha256": "e5c410a6dfebb97e9b0fc5d8fe448c1580e7862cfb603fa7bcab19fed24a91f9" - }, - { - "source": "resin/backend/speculative.py", - "path": "orbitinfer/backend/speculative.py", - "source_sha256": "322320849bdab41fac9165e90b4ac2b91fdc3af1c1468d1e5d331711ba23d466", - "import_sha256": "322320849bdab41fac9165e90b4ac2b91fdc3af1c1468d1e5d331711ba23d466" - }, - { - "source": "resin/backend/topology.py", - "path": "orbitinfer/backend/topology.py", - "source_sha256": "8d0f65909c3b09f63b182f8ab2d4f774700012b8c46b4d9bf3c70db2bd62d7c4", - "import_sha256": "6a4eb639969f814d9576ef9a4e4e6fdefd190b97e0e5dcb2c025f249211a5266" - }, - { - "source": "resin/backend/torch.py", - "path": "orbitinfer/backend/torch.py", - "source_sha256": "fc10521b892dc1ec82e3241f635cfe9fbb747bb32b62b9b8bb5b5bd73aad5b0d", - "import_sha256": "fc10521b892dc1ec82e3241f635cfe9fbb747bb32b62b9b8bb5b5bd73aad5b0d" - }, - { - "source": "resin/bench.py", - "path": "orbitinfer/bench.py", - "source_sha256": "c3be17b5d508d2772b0883f14ae6d1276188df79a732363ab569fce91ce7c3e9", - "import_sha256": "f33509b957ffea9cc166dd352baf941db16b540d03a8a6425c6c6043c82333a7" - }, - { - "source": "resin/bench_prompts.py", - "path": "orbitinfer/bench_prompts.py", - "source_sha256": "0daeea6c0b4535291b7825f680c06d7e8020f530df7f77c1e69a61673cfebcdd", - "import_sha256": "4917e50637a5b77591802e0bba8bbb76aa0914ba06a5477501d28d4a598e9dd0" - }, - { - "source": "resin/cache_storage.py", - "path": "orbitinfer/cache_storage.py", - "source_sha256": "bbdde100e79f17484321420b69546fea779ec5f989b09ecbb9147441cffb3062", - "import_sha256": "bbdde100e79f17484321420b69546fea779ec5f989b09ecbb9147441cffb3062" - }, - { - "source": "resin/kernel_cache.py", - "path": "orbitinfer/kernel_cache.py", - "source_sha256": "38ff1aa2eb078e4c08f75ea42f8527f569bb656fffaaac12c92b1c3605dfd8f0", - "import_sha256": "ce7f0f3adb2c2b70d826fe93562e6409247d8b702d28bb53369be77834e854da" - }, - { - "source": "resin/models/__init__.py", - "path": "orbitinfer/models/__init__.py", - "source_sha256": "6c22ae97b0ab451897288f248f5295fee5a5b64fb449dcf24a2016d93cf9fa49", - "import_sha256": "6c22ae97b0ab451897288f248f5295fee5a5b64fb449dcf24a2016d93cf9fa49" - }, - { - "source": "resin/models/deepseek_v41/__init__.py", - "path": "orbitinfer/models/deepseek_v41/__init__.py", - "source_sha256": "d3d77ca73ebab0ee914271bfaf0409b289ec97d7a67871948995eb4f0ce7e332", - "import_sha256": "8d702deca66d96634e27d55b7cb4e44da2e009d2f784eb297d24904ccaf728b3" - }, - { - "source": "resin/models/deepseek_v41/config.py", - "path": "orbitinfer/models/deepseek_v41/config.py", - "source_sha256": "94ff1d8d1a5eeb49ca8be44bd25625ed0920590712f9fbb9b6d49c15b4d2c595", - "import_sha256": "94ff1d8d1a5eeb49ca8be44bd25625ed0920590712f9fbb9b6d49c15b4d2c595" - }, - { - "source": "resin/models/gemma4/__init__.py", - "path": "orbitinfer/models/gemma4/__init__.py", - "source_sha256": "341a3ee8e58c4187c58777644f80daf890c55b37db4d500d6a33a7f7a1b8d847", - "import_sha256": "341a3ee8e58c4187c58777644f80daf890c55b37db4d500d6a33a7f7a1b8d847" - }, - { - "source": "resin/models/gemma4/assistant.py", - "path": "orbitinfer/models/gemma4/assistant.py", - "source_sha256": "acb211ef1f7abc69e5fc5b736c6661b3afa30375831055905c0c3ebf19b8a346", - "import_sha256": "00a5342665d526b464da657541ad3129ebe657206b650805b363a2f3bd0aaab3" - }, - { - "source": "resin/models/gemma4/config.py", - "path": "orbitinfer/models/gemma4/config.py", - "source_sha256": "c7b15af1013e356346d13b916d52d2230e9deedff10cef40a53504ab2e073bc5", - "import_sha256": "c7b15af1013e356346d13b916d52d2230e9deedff10cef40a53504ab2e073bc5" - }, - { - "source": "resin/models/gemma4/edge_config.py", - "path": "orbitinfer/models/gemma4/edge_config.py", - "source_sha256": "d32be51946c151d4443aa7a0a5abdfc97ced8d27578d854e05c8257f8ef41de7", - "import_sha256": "efdfaa7e843071bdb9d3759da1f64128534353ab47803740103f7ac6d8809b81" - }, - { - "source": "resin/models/gemma4/model.py", - "path": "orbitinfer/models/gemma4/model.py", - "source_sha256": "921eaf1191441b3e3820afbe1326918477040777351722a1a5f228465d132d33", - "import_sha256": "3b9a5808e36caf594ba12f5482197db86b0f19827d260957c2bbe0f1311b6797" - }, - { - "source": "resin/models/gemma4/plan.py", - "path": "orbitinfer/models/gemma4/plan.py", - "source_sha256": "701f5049d5378f896476bd8b4cd126e5bee57034e51b3950f41be8ee7a51af48", - "import_sha256": "38f25ddd4a896c3bc310632186bdbebbdf211fbfc78deb0ae6c97b57fd5d75ca" - }, - { - "source": "resin/models/glm5_3_flash/README.md", - "path": "orbitinfer/models/glm5_3_flash/README.md", - "source_sha256": "c3dc5631c22443f8fda6775e039c71919ffc18e3e12bae4752f3848ba0ded180", - "import_sha256": "796e8bb6239dab288a5f878a7454b9b4132b833f9fa6f4b6ad0c15c2225f2621" - }, - { - "source": "resin/models/glm5_3_flash/__init__.py", - "path": "orbitinfer/models/glm5_3_flash/__init__.py", - "source_sha256": "2622606b3532d52f059394de46d3e83edaf4a2978be2e8229006c812e213b77a", - "import_sha256": "2622606b3532d52f059394de46d3e83edaf4a2978be2e8229006c812e213b77a" - }, - { - "source": "resin/models/glm5_3_flash/config.py", - "path": "orbitinfer/models/glm5_3_flash/config.py", - "source_sha256": "f635d934153616ba8f4eec353c0e7e1e6ae065ba836d9c9d504a29b51055fbab", - "import_sha256": "f635d934153616ba8f4eec353c0e7e1e6ae065ba836d9c9d504a29b51055fbab" - }, - { - "source": "resin/models/gpt_oss/__init__.py", - "path": "orbitinfer/models/gpt_oss/__init__.py", - "source_sha256": "0e29860b8120147a669825c53cf79e0f97ba89e63388549fa0f8b61c214cffa6", - "import_sha256": "0e29860b8120147a669825c53cf79e0f97ba89e63388549fa0f8b61c214cffa6" - }, - { - "source": "resin/models/gpt_oss/config.py", - "path": "orbitinfer/models/gpt_oss/config.py", - "source_sha256": "3b4b67b596340e84caf3ddc77b0aa342a0957a033ef87a44bd95cc54574bfac1", - "import_sha256": "3b4b67b596340e84caf3ddc77b0aa342a0957a033ef87a44bd95cc54574bfac1" - }, - { - "source": "resin/models/gpt_oss/plan.py", - "path": "orbitinfer/models/gpt_oss/plan.py", - "source_sha256": "45a8094e485d6c3ae0e7106c6fa50fa37826ac7aabc29228550635e09bba44e9", - "import_sha256": "36652f7438b91f0f425fc0bd2e7ddbc39430bc9444dd9fd0cc22fa63e27fa74f" - }, - { - "source": "resin/models/inkling/__init__.py", - "path": "orbitinfer/models/inkling/__init__.py", - "source_sha256": "794ef1d200aed5bf56d658542f5a997b9c962779de9607b5ba050325d6b2dd40", - "import_sha256": "794ef1d200aed5bf56d658542f5a997b9c962779de9607b5ba050325d6b2dd40" - }, - { - "source": "resin/models/inkling/config.py", - "path": "orbitinfer/models/inkling/config.py", - "source_sha256": "1807f25fa739efe026df932b3e660cc8be7211651dd0b2599feb3c4c68eacaef", - "import_sha256": "1807f25fa739efe026df932b3e660cc8be7211651dd0b2599feb3c4c68eacaef" - }, - { - "source": "resin/models/qwen3_5/__init__.py", - "path": "orbitinfer/models/qwen3_5/__init__.py", - "source_sha256": "073fe2371b3a678253afdfd1f51c9275c7a676a8eac638fff2639b99f2e5a44f", - "import_sha256": "073fe2371b3a678253afdfd1f51c9275c7a676a8eac638fff2639b99f2e5a44f" - }, - { - "source": "resin/models/qwen3_5/config.py", - "path": "orbitinfer/models/qwen3_5/config.py", - "source_sha256": "b5cf65fdc5bc4b0167601f22344df92cacdd685899aaf1724b67589adad8531b", - "import_sha256": "b5cf65fdc5bc4b0167601f22344df92cacdd685899aaf1724b67589adad8531b" - }, - { - "source": "resin/numa.py", - "path": "orbitinfer/numa.py", - "source_sha256": "acb04b8e8e012eb264fbac34825bf079f338d34e04c2a1c3dafc49c86c25fbb6", - "import_sha256": "db9dc1146c14289e0729b6bf0dd930140d93d2e552e4d17e5ef55d387506e7d8" - }, - { - "source": "resin/process.py", - "path": "orbitinfer/process.py", - "source_sha256": "45adb5a4570e0d16773d63a7fb383e654a5fc6eafc33e96e022766cd3148e354", - "import_sha256": "1e9bc4c658f291df0e9783ddafbbf1710a97cf7aa3f537d8fa8f5ed4599101cc" - }, - { - "source": "resin/protocols.py", - "path": "orbitinfer/protocols.py", - "source_sha256": "37fb234c2fbff5c74a496e1f8ca6c3d353350a680bb26776e666371edfd6d739", - "import_sha256": "5cccda44fd6cf6e3c229e8b364ebd2134c32e19933f9d12946b768a6c2bc4561" - }, - { - "source": "resin/py.typed", - "path": "orbitinfer/py.typed", - "source_sha256": "b67803b13847a368f4ae62dce6c2e0bc3ab85f5605287d374cb67a25be829a37", - "import_sha256": "a86df4ef76ba577278f345487b13d36b8fe14b9adb87a450b29ef20f73a244bc" - }, - { - "source": "resin/recipe.py", - "path": "orbitinfer/recipe.py", - "source_sha256": "ed165d972b9b0eb9cf0699bb96a5ff8047b04c570dbd1aa2d57a733fb87bf7b9", - "import_sha256": "ff14b9ea53b0b2f1fb15a1ae945b678ecac62d9bf86ebfd3111da0d2694fdcfe" - }, - { - "source": "resin/server/__init__.py", - "path": "orbitinfer/server/__init__.py", - "source_sha256": "6038c7a651d511e29bd7e835fe1e96847dce5d8ee9d6badfe12f98d004457606", - "import_sha256": "0b4d50177747b5baf57cd6aaf66629f2d2fbc204a67f0b02e400bb2f80436ae1" - }, - { - "source": "resin/server/distributed.py", - "path": "orbitinfer/server/distributed.py", - "source_sha256": "8408b7270d89b2fb4e65f1808a011b9b2655756a357cdbac3cccde2568af49fd", - "import_sha256": "e1e13632222291bd725c5e5fd8a04e1265bd5c328ab05adc9c3dd14c2adbbc82" - }, - { - "source": "resin/server/transport.py", - "path": "orbitinfer/server/transport.py", - "source_sha256": "9b9afb5f3a9b2715727e75a402bea6ee947043c11ef0f3be198a5c28e8b469dd", - "import_sha256": "d38e96cc9ec76359d5a3b8f2cb02f6c5269a1cd8a1aaa91ad572902ccaa9056a" - }, - { - "source": "scripts/bench.py", - "path": "scripts/bench.py", - "source_sha256": "d1d4ead4ae3ab19149a96223244b0544b7e644e2e663f11bfb372431fe8f9cf1", - "import_sha256": "38fef695cdd0ddc82494f5b1538fe78d7a4911ff16b36c5e04b02dbcff7a345a" - }, - { - "source": "scripts/bench_scoring.py", - "path": "scripts/bench_scoring.py", - "source_sha256": "e9e516d1bb5e7960ae9cf2883aef0f58f282b280d4296319d4820d913e6ffb46", - "import_sha256": "95cc4e615ba430826c7b73799d07a89e1fd93bd00bc27619c209163d10c95f62" - }, - { - "source": "scripts/bench_startup.py", - "path": "scripts/bench_startup.py", - "source_sha256": "ed835a99d13e3300d6d95d21bf0b8cf02c9030125a4ed26f1495ad94d79ad5f5", - "import_sha256": "10ef5d013d9b3ddf3fc96488a2a3a4bcf7a153022340ce8ad092633fc78f3d7c" - }, - { - "source": "scripts/check_server.py", - "path": "scripts/check_server.py", - "source_sha256": "232b3978d13097ab7531c636ff3d1d20bcc36cff5d5873c8d70b7171657d36b4", - "import_sha256": "c9baee1344e9c7b3b56d6e8f7f5ee39143b9002b64524a781e7976cebf98b680" - }, - { - "source": "scripts/cloc.py", - "path": "scripts/cloc.py", - "source_sha256": "18119bcb6b2e5ca919e8cdae606440618cc7d5dce5964d4cf0ef67dffd61a3a4", - "import_sha256": "c7c5a236fb0093990ebd08d648db7588ba314a6961e7074ea0395789bf81fcf6" - }, - { - "source": "scripts/lfm2_chat.py", - "path": "scripts/lfm2_chat.py", - "source_sha256": "824b97439e3d7e9be3e4838283283a6e84b4b7d597a46ee3ec466b06dc713831", - "import_sha256": "155fc64fc8a4c404a70b4571d68956eb3de7305e5604a917398531791ee81a93" - }, - { - "source": "scripts/lfm2_serve.py", - "path": "scripts/lfm2_serve.py", - "source_sha256": "53f812489cf95da1598d4f27692e0f566335e155f305fdb303f5b4e9871fd2a0", - "import_sha256": "20ea43b47701be8c7e252f547944efa289ff586354a2c93e44fe0403ab5c930c" - }, - { - "source": "scripts/merge_bench_traces.py", - "path": "scripts/merge_bench_traces.py", - "source_sha256": "9a4de886b3bb95a4edab23f5edab908d5b4511e56c389db6352b8cd8099d5a9e", - "import_sha256": "c43e5d9a1b6f5560eb0a687da900463a617feed433ed3f0ef3dfb202fb8f1926" - }, - { - "source": "scripts/mlx_serve.py", - "path": "scripts/mlx_serve.py", - "source_sha256": "1837d2a9bb4a9f6dc88bf076ffc3ed5e2c57429a659bb623d6dbd4d3eee5af17", - "import_sha256": "49aaff646fbdc3562c33d74e639929beaa01eda35f55ebbe7217d1c73e7ddb99" - }, - { - "source": "scripts/normalize_stubs.py", - "path": "scripts/normalize_stubs.py", - "source_sha256": "3c035257a7f5df40e0613c2d2a1de13234756a2c27977b377067db112ce73734", - "import_sha256": "7dedab593aabf698d7608eebed1d33e1fca8a14edd90f749d8447959e21518bd" - }, - { - "source": "scripts/plot_bench.py", - "path": "scripts/plot_bench.py", - "source_sha256": "6d5f5d764bb2f29c9df08be13c41234b5c4eb3d4fb4f361851b999672edd3a30", - "import_sha256": "bce52dc8855bfb5f3cb4c775a19e8bb7d1f0b66d6a8de520e5582b8d9797ee49" - }, - { - "source": "scripts/rust_build_and_types.py", - "path": "scripts/rust_build_and_types.py", - "source_sha256": "e67ad68ef6bc4753d996f90ce655db746116a6a3049dadad4865a4622a005326", - "import_sha256": "fb3dfba0942b934fe735b4f53ab2e7f1188ebb1ed78021a900637ad0170e0df5" - }, - { - "source": "scripts/stress_http.py", - "path": "scripts/stress_http.py", - "source_sha256": "c10a6404b040d4109cd472976074b3bcaadf87f2bcf21ae16b7f2d7615b19243", - "import_sha256": "d6d52842b6e3727aba9ae8c71ff53a99f7e6d7404431607eeb368fa98a31696d" - }, - { - "source": "scripts/sync_type_libraries.py", - "path": "scripts/sync_type_libraries.py", - "source_sha256": "811e44e4b8e8b6418567eece27d7ddd9672888fc8d2b20a9428227e03afc4d71", - "import_sha256": "811e44e4b8e8b6418567eece27d7ddd9672888fc8d2b20a9428227e03afc4d71" - }, - { - "source": "skills/add-resin-model/SKILL.md", - "path": "skills/add-orbitinfer-model/SKILL.md", - "source_sha256": "e58cf8be4e47a411b621cd22b3f4484323f9f392b169a085a8fbba8d751af565", - "import_sha256": "4006fa395a04a95da1f7d508e50b1ca07de5ef3e8173502dbb67461c3c1be618" - }, - { - "source": "skills/add-resin-model/agents/openai.yaml", - "path": "skills/add-orbitinfer-model/agents/openai.yaml", - "source_sha256": "b1e00952689b8409c3c15c42c2dbb6b689521fcae1acda18cfc5673b4212632a", - "import_sha256": "7e1787850c11788cc5c9896b2f2721fc12794e38398c6ef8b2e6402cb2bc156a" - }, - { - "source": "skills/optimize-resin-performance/SKILL.md", - "path": "skills/optimize-orbitinfer-performance/SKILL.md", - "source_sha256": "57f7c7787e57ad2941b2f372d579761654358ada0a638cbde1fe46393c45dedf", - "import_sha256": "1a8813aea7c9f3467135186e458817cf6f1868cfe0ffa865847c7fe2a177400c" - }, - { - "source": "skills/optimize-resin-performance/agents/openai.yaml", - "path": "skills/optimize-orbitinfer-performance/agents/openai.yaml", - "source_sha256": "92b069be6bf1f054d3ec63981149c42c4d6617289604aaef29ad8018afa81009", - "import_sha256": "d7a82f201a701c291e706b5568f093c020b42040a0078cdde91cb367e2796c71" - }, - { - "source": "skills/optimize-resin-performance/references/lessons.md", - "path": "skills/optimize-orbitinfer-performance/references/lessons.md", - "source_sha256": "9fc76b73080f170152b511995848edacc3ba4e31e90cbebd90254d51ded032dd", - "import_sha256": "588090cdcdfb6313b6a7676e4c9c39210c3a46a0dc61416dbd31e75dc0bf72b1" - }, - { - "source": "tests/fixtures/deepseek_v41_images.json", - "path": "tests/fixtures/deepseek_v41_images.json", - "source_sha256": "d92708c1e481ddaa0b63375ac221c0889ecfe24af8a52eb0cdc511732a219f38", - "import_sha256": "d92708c1e481ddaa0b63375ac221c0889ecfe24af8a52eb0cdc511732a219f38" - }, - { - "source": "tests/http_packets.py", - "path": "tests/http_packets.py", - "source_sha256": "1f40c45f5e9587edc27f7160facc05966370830706f2c94ce1d2dfa5ddfd8b81", - "import_sha256": "c6f7969561054bd55229e27b91fd72a3da2693a294115c20400b7dcd1ab6ff13" - }, - { - "source": "tests/output_helpers.py", - "path": "tests/output_helpers.py", - "source_sha256": "ea6c5223aa6a842b42c035b3d4e3f3cad6c11ff4c880b99571cf1df4e24111cf", - "import_sha256": "69dfc0faad4ba3f1cd0505113d1460f80a2049ef71ab9a1ad3b7f168632604d6" - }, - { - "source": "tests/recipe_fuzz/README.md", - "path": "tests/recipe_fuzz/README.md", - "source_sha256": "a5f9a8ea6ec8b2d7ff82d16bf49c0cb9c25f3035a582162add0872e958b64a9b", - "import_sha256": "4af92654ac65bbcc271c805e07e66f851ad7e916d1d6a42e23b217ff3c5112a1" - }, - { - "source": "tests/recipe_fuzz/__init__.py", - "path": "tests/recipe_fuzz/__init__.py", - "source_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855", - "import_sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855" - }, - { - "source": "tests/recipe_fuzz/deepseek.py", - "path": "tests/recipe_fuzz/deepseek.py", - "source_sha256": "0580e69dfe02e0a8304d5a2fda173d16c69fde0b36930c2f634d54f6a1519030", - "import_sha256": "000c530104976c2cb7f5fcd68090c68b36f52e0b0d5c6209775f46a313791c25" - }, - { - "source": "tests/recipe_fuzz/gemma.py", - "path": "tests/recipe_fuzz/gemma.py", - "source_sha256": "bbddb59191e7e94e61713ecd1a30c7767e4504d697318604807e9c1f229ff97b", - "import_sha256": "ba1aea20e423da333d7ff8dee91d1079d79c9d6459132ad9a83e41638974ad9c" - }, - { - "source": "tests/recipe_fuzz/gpt_oss.py", - "path": "tests/recipe_fuzz/gpt_oss.py", - "source_sha256": "7d6a9eb32b9155dccbac6d6d29dc8b830f10eb09b499c9677b21cfe8c3e57813", - "import_sha256": "155121aec492ca55ddb8094dd537529ebdb13989e493a089917929aedbd74fda" - }, - { - "source": "tests/recipe_fuzz/harness.py", - "path": "tests/recipe_fuzz/harness.py", - "source_sha256": "aa0f666e9d4fd5407c88b4cab00c4ee7e17469af5b5c513cebcb22213f8386b5", - "import_sha256": "2a7ec0ddfbeeadb0bf9169aedb8124710390c879610050939a36ade3cbedf244" - }, - { - "source": "tests/recipe_fuzz/inputs.py", - "path": "tests/recipe_fuzz/inputs.py", - "source_sha256": "a971952592b6fb6d87bccdf4d3721832a6117d84f27d0c1d1f3381416a5175ff", - "import_sha256": "00873d559dfd9adf0027979d46755fe8462fa6db444385791bf12d91071a460d" - }, - { - "source": "tests/recipe_fuzz/lfm2.py", - "path": "tests/recipe_fuzz/lfm2.py", - "source_sha256": "dcdc239c846bbab670baeab7f3b4f801d88bbd13d8b10b9d42fef8fa030ac52a", - "import_sha256": "d1bd3ee2e71a1fa663ca1826c5411f36e3dd36111b9fa9c67734452c3fdb36de" - }, - { - "source": "tests/recipe_fuzz/qwen.py", - "path": "tests/recipe_fuzz/qwen.py", - "source_sha256": "8bfe9f3abaae360f63a63eb908ec3c05fdf511e2d0bf8539f51f0784a556da8f", - "import_sha256": "e0d1b569f389414c26cc66bf67f0f3fb170e1f8c4ab0d3ccd9e969d49a9d781e" - }, - { - "source": "tests/recipe_fuzz/references/README.md", - "path": "tests/recipe_fuzz/references/README.md", - "source_sha256": "0956c87e5bdeed54468286bb501dc81efb37ccd11dc1eeb4a082dea37204853b", - "import_sha256": "4b81b5c2eab5b8449beaac852b8f8d9b59b0c444dc27a2487009cd98dd2c64bf" - }, - { - "source": "tests/recipe_fuzz/references/gemma4-LICENSE.txt", - "path": "tests/recipe_fuzz/references/gemma4-LICENSE.txt", - "source_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", - "import_sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30" - }, - { - "source": "tests/recipe_fuzz/references/lfm2-LICENSE.txt", - "path": "tests/recipe_fuzz/references/lfm2-LICENSE.txt", - "source_sha256": "30adf9d6478191fb87f2424f63ba0728598335aaf99cd2848ef17e8e545fe94b", - "import_sha256": "30adf9d6478191fb87f2424f63ba0728598335aaf99cd2848ef17e8e545fe94b" - }, - { - "source": "tests/recipe_fuzz/references/lfm2.jinja", - "path": "tests/recipe_fuzz/references/lfm2.jinja", - "source_sha256": "c5b67247c9f736f6c40eaa91ae1ef27dc67f5b041eedaadb97a1bea5030fb0ff", - "import_sha256": "c5b67247c9f736f6c40eaa91ae1ef27dc67f5b041eedaadb97a1bea5030fb0ff" - }, - { - "source": "tests/recipe_fuzz/templates/README.md", - "path": "tests/recipe_fuzz/templates/README.md", - "source_sha256": "1ceede467d3e36164dc5f1e937208e2825bb4b7c77952841cc6c9fbefcf022a9", - "import_sha256": "6786e139d4d7f5ac9a387e2fda67a59ef975119bbbb45850aaa65186ec5a90ee" - }, - { - "source": "tests/recipe_fuzz/templates/gemma4.jinja", - "path": "tests/recipe_fuzz/templates/gemma4.jinja", - "source_sha256": "94899c0f917d93f6fe81c95744d1e8ddab2d21d39228d2e4aec1fb2a25bff413", - "import_sha256": "94899c0f917d93f6fe81c95744d1e8ddab2d21d39228d2e4aec1fb2a25bff413" - }, - { - "source": "tests/recipe_fuzz/templates/glm53_images_hf.npz", - "path": "tests/recipe_fuzz/templates/glm53_images_hf.npz", - "source_sha256": "fd5bee144cc80832637c4a36d1628200358c7dcdc07525936a56d6110389c10a", - "import_sha256": "fd5bee144cc80832637c4a36d1628200358c7dcdc07525936a56d6110389c10a" - }, - { - "source": "tests/recipe_fuzz/templates/glm5_3_flash.jinja", - "path": "tests/recipe_fuzz/templates/glm5_3_flash.jinja", - "source_sha256": "34d5ee66b12fa6446cdae131c352b8f68cd85369e0e6fda115583805fada3891", - "import_sha256": "34d5ee66b12fa6446cdae131c352b8f68cd85369e0e6fda115583805fada3891" - }, - { - "source": "tests/recipe_fuzz/templates/inkling.jinja", - "path": "tests/recipe_fuzz/templates/inkling.jinja", - "source_sha256": "0aa1aa0c729d90176dcaa00c440c8faffca2957ffb2cc4b79456ee6d02bcf43b", - "import_sha256": "0aa1aa0c729d90176dcaa00c440c8faffca2957ffb2cc4b79456ee6d02bcf43b" - }, - { - "source": "tests/recipe_fuzz/templates/ornith.jinja", - "path": "tests/recipe_fuzz/templates/ornith.jinja", - "source_sha256": "182e77dd83bd8e9ca818b240b82e28f243762cd5dda32e6eef327df7b1cd107e", - "import_sha256": "182e77dd83bd8e9ca818b240b82e28f243762cd5dda32e6eef327df7b1cd107e" - }, - { - "source": "tests/recipe_fuzz/templates/qwen3_5.jinja", - "path": "tests/recipe_fuzz/templates/qwen3_5.jinja", - "source_sha256": "a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715", - "import_sha256": "a4aee8afcf2e0711942cf848899be66016f8d14a889ff9ede07bca099c28f715" - }, - { - "source": "tests/recipe_fuzz/templates/qwen3_8.jinja", - "path": "tests/recipe_fuzz/templates/qwen3_8.jinja", - "source_sha256": "c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041", - "import_sha256": "c3cf9e34abf4f9e36c2d72165aa9c132d3e2a725b6c2586aaa3a8af9d7a81041" - }, - { - "source": "tests/recipe_fuzz/test_deepseek_images.py", - "path": "tests/recipe_fuzz/test_deepseek_images.py", - "source_sha256": "5a58cd56a709de1d3fd413af14b65e95de98e3a85fb9d2e259bbdcb3f15827ad", - "import_sha256": "6112241ea62c46bbabf4d7d1cc7d31cc35c8b4c7ea51db285a1ff2f8b0d499d8" - }, - { - "source": "tests/recipe_fuzz/test_differential.py", - "path": "tests/recipe_fuzz/test_differential.py", - "source_sha256": "59232147f198475fba306a719cbee398d7eed2641d4431e65e003a1a071cd383", - "import_sha256": "a1c496efa20b763bef17006339864c3ae827118e294104eb05f0eafc0ff350f7" - }, - { - "source": "tests/test_backend_capacity.py", - "path": "tests/test_backend_capacity.py", - "source_sha256": "06e6cac66acaed4d15cf9edf09b8c8bf17e4226caf39a1b2d0f369970ef7bf55", - "import_sha256": "7232b5f272629f6e8ba9ae2707dfece46f2f4ad9d7081e1e75930ed76ffd9ce1" - }, - { - "source": "tests/test_backend_loading.py", - "path": "tests/test_backend_loading.py", - "source_sha256": "1c9c79537e7b877015fa98fa820ee2a7944c41547bea9782c786ac1d895793e8", - "import_sha256": "133bb072fefec52a7ea98906567083aa0fbea4a9a8b19f647e092a0c53a72219" - }, - { - "source": "tests/test_backend_plans.py", - "path": "tests/test_backend_plans.py", - "source_sha256": "631d32fc78698af1869a706e023b96c638f08aef52ce9a64dd1d8f4d048fa1f8", - "import_sha256": "f93440ce89e66f9835bf50595727cbbc61f240240954d8acbdd07c46804072a2" - }, - { - "source": "tests/test_bench.py", - "path": "tests/test_bench.py", - "source_sha256": "1fcbeef95a3f8dc3e35a19abdc2230eb31bc18bfaa12a9edc979d1dfc5828ea5", - "import_sha256": "74b492c49ccab1981ac61378a7bcc3705f17cd1d448d1b69b05f3698267e5816" - }, - { - "source": "tests/test_bench_cli.py", - "path": "tests/test_bench_cli.py", - "source_sha256": "68cec265928a5b22b0230a792f71a279b7269776976802b6a03971f1b2c26037", - "import_sha256": "68cec265928a5b22b0230a792f71a279b7269776976802b6a03971f1b2c26037" - }, - { - "source": "tests/test_bench_plots.py", - "path": "tests/test_bench_plots.py", - "source_sha256": "fcb6053ce92a97f0a5e526917c61ee193ec8d4b3a932cb59424a2d504242c874", - "import_sha256": "fcb6053ce92a97f0a5e526917c61ee193ec8d4b3a932cb59424a2d504242c874" - }, - { - "source": "tests/test_bench_preflight.py", - "path": "tests/test_bench_preflight.py", - "source_sha256": "ab1db1b784399bf9dc003dd48def7911a77d36465c29d0952dbfa848ea827399", - "import_sha256": "b3a2e82fbdfbefeef6e865ec3769551ea1507431e73b58e38f9467d3b9437fbe" - }, - { - "source": "tests/test_bench_trace.py", - "path": "tests/test_bench_trace.py", - "source_sha256": "16607063e3641b98eb53ca93c368085616e6b464aa17f8af26ae8c97e4d6a768", - "import_sha256": "5b74b2f2e247ab8958089f1922aa447753e59bb58277055cfdaf83e527f22589" - }, - { - "source": "tests/test_bench_trace_headroom.py", - "path": "tests/test_bench_trace_headroom.py", - "source_sha256": "31b2e271f29fbe3bfd6be3d539dbd888ecd4436e9c293c984f0ac2769a81f040", - "import_sha256": "c352d205080f64764966913c14a755e54a8d48810c0b2ee080c31b999be851dc" - }, - { - "source": "tests/test_cache_salt.py", - "path": "tests/test_cache_salt.py", - "source_sha256": "057068f3ef0e8cdfac279922385e0f9a2cbf69e03b687094f5651d31e5770ead", - "import_sha256": "ba25873ed6394590204e775d760de8a2c7eeb37b4beb4f15f6188e8abfe0a4fa" - }, - { - "source": "tests/test_cache_storage.py", - "path": "tests/test_cache_storage.py", - "source_sha256": "0c06d51b83841efd37745332567ef1e3810ba513d917ebfd82586809743faae0", - "import_sha256": "18264e089039cad6bcc654edc461c7342be7748ad69999626dd8c6dfcd68bfdd" - }, - { - "source": "tests/test_check_server.py", - "path": "tests/test_check_server.py", - "source_sha256": "6d1f2a6bae2a956c1aecaf622b7a9b952721e63c165dd7165d072526711418fc", - "import_sha256": "6261172b14f2d992bbceaec792e1e1d472d3ddcb783bbbdc47060a8a3dea0178" - }, - { - "source": "tests/test_checkpoint_cache.py", - "path": "tests/test_checkpoint_cache.py", - "source_sha256": "3c55250ec5a0daecde78cd9d5498d48f9e10b9174d749221383a3f064892d9ae", - "import_sha256": "3c55250ec5a0daecde78cd9d5498d48f9e10b9174d749221383a3f064892d9ae" - }, - { - "source": "tests/test_composable_expert_parallel.py", - "path": "tests/test_composable_expert_parallel.py", - "source_sha256": "305d033fa7e80fe31c5908312a1c6ef2327ffb71e58c904d3258d0a2935e7a7b", - "import_sha256": "f430fe7c322aee57b9b816c6cc33c4aca0df875f62df3621b9b290fe26f1099c" - }, - { - "source": "tests/test_context_parallel.py", - "path": "tests/test_context_parallel.py", - "source_sha256": "8d7f7731e93a2592dce694a0dee7053e8739a1353c7e069eb310b38a4f2a76b4", - "import_sha256": "f3619e3660f63d1eaea96075bc6b4aa40c564d945a7948a4d0674c89f01e064c" - }, - { - "source": "tests/test_context_parallel_gpu.py", - "path": "tests/test_context_parallel_gpu.py", - "source_sha256": "14185de4c12731f33e4e947d988e6a3f159a1edaefefc5b965465fa8bf17f339", - "import_sha256": "8de5bfaa0aa42277d18414efea94116d65d83579d4ce26ffa4c813e8c6b68133" - }, - { - "source": "tests/test_control_plane.py", - "path": "tests/test_control_plane.py", - "source_sha256": "103432e9052dfe6b8ceef753ceb0356a9fdaab5e99a766f679f7b478c7d94319", - "import_sha256": "c9fc6ef7390b06223b48ddc5e393a8a79a59d863252dd2687b92c65c644108d1" - }, - { - "source": "tests/test_cuda_completion.py", - "path": "tests/test_cuda_completion.py", - "source_sha256": "fd09b9db282b5c99c9fe3e37cae8791fb84d93d74e70e19d138365ff4ad4cf46", - "import_sha256": "61174131559252aac97987601ace6375381846fe46d2598ed7d1f63835ab9e55" - }, - { - "source": "tests/test_cuda_failures.py", - "path": "tests/test_cuda_failures.py", - "source_sha256": "a9fa300d7a4e9d00208064fbe6d16c13439c73cec86e1f487d4d85c4b386ab42", - "import_sha256": "f3e6030cb30f1aa35e3ae86297580980756dc84a5f9b52146164a6649701f74c" - }, - { - "source": "tests/test_cuda_graph_pool_gpu.py", - "path": "tests/test_cuda_graph_pool_gpu.py", - "source_sha256": "badab350f73cb52f93b1af90d00853a970cb7a69d4ef7ef210083f5af270e43c", - "import_sha256": "bd3fc75befe96c05ce5f0d1767855dcd9dcc92b31965dd840c3fc192b8d6ca2f" - }, - { - "source": "tests/test_cuda_graphs.py", - "path": "tests/test_cuda_graphs.py", - "source_sha256": "e3d056fd519e35888047a2553c4c6251d336253d942d3aedceb22ee71e9e59a4", - "import_sha256": "8258f853040f9af995a3402dae2750869d5005a3e00a642da875e6cc5262dd2a" - }, - { - "source": "tests/test_cuda_metrics.py", - "path": "tests/test_cuda_metrics.py", - "source_sha256": "0682a614385c1b5e2ee2efc315eebc1bbf9d3955c563465046da09862d9fff5e", - "import_sha256": "8296c26ac70411c20da936326852dfe08e62fc8df5a1a7db8f1f8e27cbbefebe" - }, - { - "source": "tests/test_cuda_plan_validation.py", - "path": "tests/test_cuda_plan_validation.py", - "source_sha256": "94619e085e482a57629c6b24ae0c475840d694f37dfae20374a8d9507e621742", - "import_sha256": "abaac77e4eb78a00bbe97e963b64d448561e03131ced6701168299331ce45c0b" - }, - { - "source": "tests/test_cuda_prefill_graphs.py", - "path": "tests/test_cuda_prefill_graphs.py", - "source_sha256": "a2af7d4a6b01d64f7f0beab7ae2ef2b56450449c8277066d6a2d5b07b7aca0c5", - "import_sha256": "e4ac590d27782e6e66961a504c0f8a17a1296fdc50fdbb7ab426b2d59c806166" - }, - { - "source": "tests/test_cuda_slab.py", - "path": "tests/test_cuda_slab.py", - "source_sha256": "136a6edef6907bd0eeba41bc5c92b47a612584278d3f0c114f89d4bf31c74f8d", - "import_sha256": "760eb68a59a687907d7cb4018d0886286c121c8147045d427d91471c194da87e" - }, - { - "source": "tests/test_cuda_staging.py", - "path": "tests/test_cuda_staging.py", - "source_sha256": "57dd61dc4276fc6d0759db6c15fba0792624c93625b1a67e3355585e211aed08", - "import_sha256": "69e7e448c18b62a2d751ea8c030da190a6ed22d456b2dcd983ea772d9974fb49" - }, - { - "source": "tests/test_cuda_tensor_graphs.py", - "path": "tests/test_cuda_tensor_graphs.py", - "source_sha256": "9d87b70ecda484089e7c95e9e311db8148fec302adec4513e936db7ef1c04cae", - "import_sha256": "918e752c029e5ce6f0d7a7a8464d78b784a74b3d981e8f0caf6482945638014f" - }, - { - "source": "tests/test_deepseek_attention_graphs.py", - "path": "tests/test_deepseek_attention_graphs.py", - "source_sha256": "13365d72e5a1c5fb2389f0304f763028a5a51745191f62fd7d389ba2340d63fe", - "import_sha256": "53581550ee3f48de88b552205654fa162158f415812f79725ec5b242c0c938fe" - }, - { - "source": "tests/test_deepseek_auxiliary_graphs.py", - "path": "tests/test_deepseek_auxiliary_graphs.py", - "source_sha256": "474b42a59192d374b1d10d0643619a1b990989263719cee66b98c804e9959650", - "import_sha256": "3b55bce525ad524c9fd655812b21a9a85b5ed29100bc73c1ad602976127cbbc5" - }, - { - "source": "tests/test_deepseek_engram_graphs.py", - "path": "tests/test_deepseek_engram_graphs.py", - "source_sha256": "d23cd9a766cbe91ab29fef3b742c0b5618d1ebb5bf1bbbaec4517b6f7f1c9e9f", - "import_sha256": "f6f18d94f538f498292bf35f63e05e3dcf3a1c9257fd255a57f1b2f7a60c5d83" - }, - { - "source": "tests/test_deepseek_hc_warmup.py", - "path": "tests/test_deepseek_hc_warmup.py", - "source_sha256": "9c662a616e8806342ff180665d80861bf37d667269688997a71579a6935290ad", - "import_sha256": "499390bd49ea9a422de7ad44578591d17964c5cf3ce05b76ce80aadd47ae216b" - }, - { - "source": "tests/test_deepseek_v41_codec.py", - "path": "tests/test_deepseek_v41_codec.py", - "source_sha256": "ff8bb60f0604d2f3403951b1a18a50c891da6368b460a69d5abec6beb20764ce", - "import_sha256": "224e0f1181d386f745946e95d39b9108beceb40b9608c71e5270c121ad2f2ecb" - }, - { - "source": "tests/test_deepseek_v41_media.py", - "path": "tests/test_deepseek_v41_media.py", - "source_sha256": "80d7cbed95c856f03f864b525190b80b7c2a987b86edf0601927d6957dcb74ee", - "import_sha256": "97121a2a70c4c8114e40a60c66ffffb35c25bd985c577de89d2305594e017b0b" - }, - { - "source": "tests/test_deploy_models.py", - "path": "tests/test_deploy_models.py", - "source_sha256": "bff6be1fdc07d6d252a39306654d868bb44cf81f87622ee84f8889e7fc0ba84e", - "import_sha256": "db572a47b08a0a0eba05f799e765e7da2a08a06880b5ee23bf891c98127a323b" - }, - { - "source": "tests/test_distributed.py", - "path": "tests/test_distributed.py", - "source_sha256": "4c55c26562830b17f1584d26a01e2a6d244f9855dcd305c766693a14d7295b77", - "import_sha256": "dbc7cddfe6ee8e53e90f73eaf459e6fc8264251548afa7c1176fa9388435e958" - }, - { - "source": "tests/test_distributed_cuda_graphs.py", - "path": "tests/test_distributed_cuda_graphs.py", - "source_sha256": "8a3199aac446e17afd7a588e0043a0f87e4861e28d0c0ef86af1a9bcba6d5e08", - "import_sha256": "5c3ed13bd2e48dbe3d41c6ee19debfee191b1e45361f0c7b292593320104f657" - }, - { - "source": "tests/test_distributed_http.py", - "path": "tests/test_distributed_http.py", - "source_sha256": "79a664f727e6885ea9c4dfb76e4144f4fee410012c3de2e8f51a6440ec0a8ecb", - "import_sha256": "37725527c67f39d8de7e1032ba5e483bb4adfe6bac70df710ea5944cce3cb6c0" - }, - { - "source": "tests/test_distributed_metrics.py", - "path": "tests/test_distributed_metrics.py", - "source_sha256": "d5754ff908d4b54311ef2c2164312ab317ad24aa500b2508797c82c956fb2919", - "import_sha256": "320cd411d5818a98fc473b0a35c5f2dc8aed05be717deb990c4e98f42f41adab" - }, - { - "source": "tests/test_distributed_overlap.py", - "path": "tests/test_distributed_overlap.py", - "source_sha256": "0b8101ef5f46851f6175c245dcd3e0aa0552c4ee3df4a200b401c4d8b1fab420", - "import_sha256": "c34f7a0b6ec0de3560cc5eda22931745067eab832ff2fda0a08356f8e4e62021" - }, - { - "source": "tests/test_distributed_transport.py", - "path": "tests/test_distributed_transport.py", - "source_sha256": "f251eb8f52d900dffb29af3f29b7fdc404e1800de46b55afc1a051bda68f5526", - "import_sha256": "98075b05d883410086beb3bbd4c6eb30b471ffdde359ffd772e3ec18c6c16996" - }, - { - "source": "tests/test_engine.py", - "path": "tests/test_engine.py", - "source_sha256": "e6fcb959dc297cd7625358b822254ac6c6f25d5a10ed0ec958fafd52a8377fad", - "import_sha256": "025c6a6a7cb4b9713f12838645323bc1b753ec88c94867b1732170093549be3b" - }, - { - "source": "tests/test_engine_execution.py", - "path": "tests/test_engine_execution.py", - "source_sha256": "a0f9f05b2ee3d7f1fa4bd129f4989ec41430404871afbfbcc0a42801d030c08f", - "import_sha256": "73f2c00863cf209e345b80cc8610106d95e5e3ea87fb98b38df533a04980179d" - }, - { - "source": "tests/test_engine_overlap.py", - "path": "tests/test_engine_overlap.py", - "source_sha256": "52b4dc953b3c00e02af908d2d193a04f2e2fa3d62d9a2aa262aef45d53fc72fc", - "import_sha256": "a7afb9db4d1b1239df5662b4c47934c826ee8401a311d29dc24597f336350c10" - }, - { - "source": "tests/test_expert_parallel.py", - "path": "tests/test_expert_parallel.py", - "source_sha256": "00e3bb0cc05bc1884820989c92101c988dd361bca4de6010c8fc9f9d35e4254c", - "import_sha256": "15e394ccc0bd54a6dce73f954cee9934e48b64db8de1f6bc9a67eae3340e0d7e" - }, - { - "source": "tests/test_expert_parallel_gpu.py", - "path": "tests/test_expert_parallel_gpu.py", - "source_sha256": "a9ffd8d0289da6301a7fee411efb0ef37f5b46efb9775052b73b038d23cf5fb8", - "import_sha256": "d3c7ce70b3a28abe786c756a3f3d4a973a75c4e1b48d07f299d90b8b48fbc4b8" - }, - { - "source": "tests/test_expert_parallel_preparation_gpu.py", - "path": "tests/test_expert_parallel_preparation_gpu.py", - "source_sha256": "2cafa545c43869ea758e37699d6af87b1241913008d728ebf88723d2021730b0", - "import_sha256": "986ac0536b4b539bfbc05b91bdee69144b77462a7156b3f63c16211e3ba97dfa" - }, - { - "source": "tests/test_flat_cache.py", - "path": "tests/test_flat_cache.py", - "source_sha256": "eb2f0a5fc397e532c4891bbe95cb1b5d4d32731fa5fd141197a42ac4ae73d8d9", - "import_sha256": "553dda72882fd0c0fd2ba2de1e34ebf58cc16b891a9c9d4db2a36b47e829e0e7" - }, - { - "source": "tests/test_glm53_expert_sharding.py", - "path": "tests/test_glm53_expert_sharding.py", - "source_sha256": "387aa1578cb0f1feb29b59db6ef6cacf342099f633e0fa4ddab558dc64164710", - "import_sha256": "5ada7c6761ae19fbe51bc6d0c6933ecf74d064c8581cc9658ef0dc842d521475" - }, - { - "source": "tests/test_glm53_model.py", - "path": "tests/test_glm53_model.py", - "source_sha256": "24493bb5020039a9cdabe4ea8c9b35052bc4d4fd9ddb05ea2d212d823c64da51", - "import_sha256": "2e853d7cb4be6480c30dca39d4f76411e70d1a826ebc80311bbda76429fb71ef" - }, - { - "source": "tests/test_glm_media.py", - "path": "tests/test_glm_media.py", - "source_sha256": "2e4a427b9392d70527b427df6425b944306be6a02fb1947eff143710deaad582", - "import_sha256": "e4a619242b3d21bbff227ecc8859d19deb06934afbe54192665d6b162acdcda4" - }, - { - "source": "tests/test_glm_recipe.py", - "path": "tests/test_glm_recipe.py", - "source_sha256": "f8d583fbc51491b42abb1e1f1d17c6307d1b6203ac965a2022008b4b76ca15db", - "import_sha256": "03fd86be288e415afd9aa80d153353765bfd38bae4ebcffd5e78c38b96e09e52" - }, - { - "source": "tests/test_http_lifecycle.py", - "path": "tests/test_http_lifecycle.py", - "source_sha256": "82329976db5b34838103a10583610fb6ee31411a1cccf6603b3ecad5b95a63d3", - "import_sha256": "0b56ec13f1b28dbadd21a8a5567433dbaa556a44ed71bb7c4b39aef0ad663fe9" - }, - { - "source": "tests/test_http_logprobs.py", - "path": "tests/test_http_logprobs.py", - "source_sha256": "eb65b54a2969ef278e82c0f12e8729b61a07b08f2d385a5b4bc4a05201455018", - "import_sha256": "f233c6b6295ae36c25968a8a756b4e062ef8325280ad9637b0881c646c6e35ae" - }, - { - "source": "tests/test_http_metrics.py", - "path": "tests/test_http_metrics.py", - "source_sha256": "32a8a612c7e9e99eb0c85149db8c49a9bad4bdda219c3e73f85591c089907080", - "import_sha256": "e1d2462e22e9faab49f0314b955fefe03a057cf66aa8b7a4e634ca0c1b9776fa" - }, - { - "source": "tests/test_http_server.py", - "path": "tests/test_http_server.py", - "source_sha256": "53ef625acf906f7b158a0728fa01468d35c964aeef691dd04cc5d0dc0eab93ec", - "import_sha256": "c952f90e0bbf11de106be4650fc71e9cb213ecfd84c13cf962933b7dedbee883" - }, - { - "source": "tests/test_http_session.py", - "path": "tests/test_http_session.py", - "source_sha256": "33ceb97be5dcde2f51eed5832674eaae106074cf2fe66de3c114ba5b1bec3061", - "import_sha256": "62c1cb5a705db927ffbed26f5c4e876c5f95bdfc9fa8be6ef7ae59854e5ad813" - }, - { - "source": "tests/test_http_stops.py", - "path": "tests/test_http_stops.py", - "source_sha256": "7665a5880637acec7fdc432e9d760e39d1ee86294d5db20cf9da0b2370dcf8f8", - "import_sha256": "a28d146bf7a14a5a9bdde75bf29b66783548c2e8d2323bd50e35c010a51b28d4" - }, - { - "source": "tests/test_http_stream_interval.py", - "path": "tests/test_http_stream_interval.py", - "source_sha256": "eeba09a47d5888b08803e2ce2ed2bfdd90dac6a94323233d88e15bd4eefacb88", - "import_sha256": "c45f834f101c7e23da9b126b466a25152d9367ec3222d22464316166491c4053" - }, - { - "source": "tests/test_http_stress.py", - "path": "tests/test_http_stress.py", - "source_sha256": "320702a9036e3218c511fd10497d6de001095cf7b1e97cfb1d95260e71eae751", - "import_sha256": "216625ecf8db94906d1fecce277795a57cc8e3165b7bbbcc9da3606e4ddd5e40" - }, - { - "source": "tests/test_http_token_counts.py", - "path": "tests/test_http_token_counts.py", - "source_sha256": "67530ca18ae58a697713e98b421f7f99664a213e01a64e5414b5862a20ddcfad", - "import_sha256": "1726646e032e997c8b89fe2b64eddf499011239106b9de8b86143d481a68dfcd" - }, - { - "source": "tests/test_http_token_limits.py", - "path": "tests/test_http_token_limits.py", - "source_sha256": "12bf0ed4b887f97b82afab0bcfd250c28fc0b4914a2a9e6df2668629ffa5c6eb", - "import_sha256": "c53ae0576e70a2f026c3d9675a345b32e28eb08d45c91bddc1f50e4857650418" - }, - { - "source": "tests/test_http_transport.py", - "path": "tests/test_http_transport.py", - "source_sha256": "1d2199ea39d0bf7748fab1876e0e8c8b92bbe115b01a6cbbbd674c97101f2790", - "import_sha256": "06772654e1130e1d93e9c476f646c9d78fdcddcb6fb82b04abd637606f5e67c6" - }, - { - "source": "tests/test_jit_cache.py", - "path": "tests/test_jit_cache.py", - "source_sha256": "bb68314a41546bf64e9eea79be5a98cb0ba2876badcc95ad835ab0ac6aece0d2", - "import_sha256": "bb68314a41546bf64e9eea79be5a98cb0ba2876badcc95ad835ab0ac6aece0d2" - }, - { - "source": "tests/test_kernel_cache.py", - "path": "tests/test_kernel_cache.py", - "source_sha256": "1762f129ab1914215897a6e860463805b8ba6d3594b90a43df6bb2e1a4f8a18f", - "import_sha256": "7a8763e9d70a4c40db51266147f1bd1b2828a6608d207e2f334574788b0338a4" - }, - { - "source": "tests/test_lazy_imports.py", - "path": "tests/test_lazy_imports.py", - "source_sha256": "aeb469d8ca0cf29d15e6e9bd0c93e21f091ea9286bdea34ed97e85b9b92ea8fa", - "import_sha256": "14d2125fc8c12eef638b5d12b11469fe0aa21232391b2550e36c0b407352c94a" - }, - { - "source": "tests/test_logprob_limits.py", - "path": "tests/test_logprob_limits.py", - "source_sha256": "51e24f41c784c933827ad72597a65b47142d275ff5e10ec3fb2cba49f92da775", - "import_sha256": "60eced9e8908e33172ce5cb0826254e24c0b432cb3a244114cb79ea2c19d88c2" - }, - { - "source": "tests/test_logprobs.py", - "path": "tests/test_logprobs.py", - "source_sha256": "9655b03c580d328756b5056c883e0fa97517b82884e3bbc811b9d117a2290bb9", - "import_sha256": "807766f72bfdeccd9b0105fcd1117a930ebbb41c5241a9c94027164f431ed9c4" - }, - { - "source": "tests/test_mega_moe.py", - "path": "tests/test_mega_moe.py", - "source_sha256": "24527dc528b120319a68d907f7ac733705cf3f8c1a68c5a2c5abf526312db1bf", - "import_sha256": "13af6ed156f2ad15f117f07a05e16baef8c9497e1b9ba0f5b7acac73e815e5dd" - }, - { - "source": "tests/test_merge_bench_traces.py", - "path": "tests/test_merge_bench_traces.py", - "source_sha256": "7ec62c2a7a63ff7743257116f4ab87cb949f8661e43a33f6902fe3d864ba6be1", - "import_sha256": "7f0141f86c73ebaf6c0c7e60be880badc8f732a598b1c71d517ed2a79451ff22" - }, - { - "source": "tests/test_metrics.py", - "path": "tests/test_metrics.py", - "source_sha256": "d1a2877a11a7e88a8a157388e423628cd07848598f101038a800e30b3010db3f", - "import_sha256": "fc559cfa4b44383642c263c8622186eff43da5a51ff097ff0ac5ddec615dc968" - }, - { - "source": "tests/test_modal_serve.py", - "path": "tests/test_modal_serve.py", - "source_sha256": "b6ec0ab5285667893ed54439a22604ca1a285c5906cb4cadef629a5832db7b75", - "import_sha256": "52cf32a6cf088644d671b10cd7ec1cae698788d09e1231c7cc326ffe8d796aba" - }, - { - "source": "tests/test_native_recipe_types.py", - "path": "tests/test_native_recipe_types.py", - "source_sha256": "5de50feffeb8033fa19cb1791fee2faabb3f553b9b9479465ce03e73e623f98a", - "import_sha256": "0d42ca31daaef00a47fe4299c72ceeb2aa73ed3b8a3275802d5db25f8a2326fa" - }, - { - "source": "tests/test_numa.py", - "path": "tests/test_numa.py", - "source_sha256": "a615a1b3ddbd008d895574571802bacbe0d29e59775a88ab860c141345198b46", - "import_sha256": "351bca727b2d6e01a9d339182d48a0e3d87377066de3d92838a30e00c6db03bc" - }, - { - "source": "tests/test_parallelism.py", - "path": "tests/test_parallelism.py", - "source_sha256": "867797211c473c5cccd333f79c7d15120d7e1a27cf4196b994452f8c0510f7f0", - "import_sha256": "366a8e5282a12571106acb6b5eea4905e90c6a628b3689cbc3d2375d8158864e" - }, - { - "source": "tests/test_plan_metadata.py", - "path": "tests/test_plan_metadata.py", - "source_sha256": "74314fa35387f6f631e69cb6917d6723af39dc4ce88b83caac5e18ea49b74027", - "import_sha256": "956b074b297ffc506b6a8941c858bf66bea6764cafea18ffee6ce1728be6c75a" - }, - { - "source": "tests/test_plan_transport.py", - "path": "tests/test_plan_transport.py", - "source_sha256": "5428732be321237f0579f444479567ba1be6ca7f514225197b8530181c92f815", - "import_sha256": "66e723b48fdbb9bee1b3fcf24ec43e7c1a93563a9685cc5c7d6e2ba741fe9023" - }, - { - "source": "tests/test_process.py", - "path": "tests/test_process.py", - "source_sha256": "9e03cccec2df8515f005d6f83992043ae09892ef60f1525b10dff14006a31854", - "import_sha256": "d7c33819e5b7c17e463bf756cbb8ffd4ac9720e376cfd10dcbe7028aa0391aae" - }, - { - "source": "tests/test_prompt_logprobs.py", - "path": "tests/test_prompt_logprobs.py", - "source_sha256": "5299128eac2dbbfa46c76b633bd4b601d9722de0e03200cf6d81592b0d4d2f85", - "import_sha256": "ac990408898081a3edf3707ef19471168423f3c423ae5d54e460f83c92f891ee" - }, - { - "source": "tests/test_qwen3_5_moe_config.py", - "path": "tests/test_qwen3_5_moe_config.py", - "source_sha256": "3d734a1e9e2d5b8e5ed73aeaa5e9f400b18d03aaaf661538cfd553ec32ad6635", - "import_sha256": "5ce502e62850153fdb24e639296b487e7201262cb3662897add63ddfd9b62c21" - }, - { - "source": "tests/test_recipe.py", - "path": "tests/test_recipe.py", - "source_sha256": "489e32ddf73c67eb2173d6fc5b5d7dde48a27c6dca02cf9e35c2dd5847660d17", - "import_sha256": "c66bdad1807a65095212dc666e2a7dea06b7998423f7f7ae952e983c0700957b" - }, - { - "source": "tests/test_recipe_http.py", - "path": "tests/test_recipe_http.py", - "source_sha256": "87b95f35bc78e50390c41e67af8c0432de2dbd8563dc92f9f6368887bc1b5934", - "import_sha256": "d0f6e6cb186fc54702cb79c3635dddaa1d65f8e5a4bf79357fdc2f5a7098bad3" - }, - { - "source": "tests/test_recipe_media.py", - "path": "tests/test_recipe_media.py", - "source_sha256": "ca258bfa5f62f2d5b24065d4887e3a2631f5d417b929343a46682ebc859505f0", - "import_sha256": "8db73932f8100665f315d4c3f7b97db0e2579ca82f8ef7ff29983f8da02b54a5" - }, - { - "source": "tests/test_recipe_media_cache.py", - "path": "tests/test_recipe_media_cache.py", - "source_sha256": "6ea1d87652b00d3214700eca92b0ce70c1802174c11cb2f952a43505cdae0d70", - "import_sha256": "2e92f0ca4dbf07bf6e71de6bca9adbdd9514772a8f87adb5ddfe720beecfa46d" - }, - { - "source": "tests/test_recipe_media_preparation.py", - "path": "tests/test_recipe_media_preparation.py", - "source_sha256": "dd98974d575f4b48cc0bc7e6216c5569f62266fd31e72b41da1b4bac60d3c027", - "import_sha256": "b5dca8a5ee6fe3afdbf3dd076b3ccaa8081dc9ade5127e95b2cd08d0a2968300" - }, - { - "source": "tests/test_runner_metrics.py", - "path": "tests/test_runner_metrics.py", - "source_sha256": "0b5ab3a37b7f7e1f25f3f52a41847d1c05f9674f723263cd2853594a5a80abe2", - "import_sha256": "a500228f3c78b8d1bb42034517e5facd6f5e7a0441b84d61a01f3e9ca3371832" - }, - { - "source": "tests/test_sampling.py", - "path": "tests/test_sampling.py", - "source_sha256": "709f1dcdfe69f90b9303c359945e50e3f40d75ae0480b21fc150e9c4299da0b3", - "import_sha256": "851830862890910f83e6602777a41fce6cb56ed8ac1b6ccf81f00ba66bbebb69" - }, - { - "source": "tests/test_sampling_blocks_cuda.py", - "path": "tests/test_sampling_blocks_cuda.py", - "source_sha256": "dfa6e004e8d43d5dda8c3e360b01578507a92b5be98868a083134e42ccbf53ef", - "import_sha256": "0f05fee8fb091718f1f929d22ba537a4827f4ac3855ddb129f46a1f9b97fe8b3" - }, - { - "source": "tests/test_sampling_completion_cuda.py", - "path": "tests/test_sampling_completion_cuda.py", - "source_sha256": "65515e4d3745c8d81b86fcc47336e504ce4574f5514d315837699cefbadb2008", - "import_sha256": "f1d28b4e94d89c32794c604ac17abaf7ff3c6fe8f959f7775dc286652f0f85dc" - }, - { - "source": "tests/test_sampling_cuda.py", - "path": "tests/test_sampling_cuda.py", - "source_sha256": "915f2eab677af54b0ca4fc9525216c08953f315076141bfb149778934a602140", - "import_sha256": "2acf1f9c30184f447986124a5b588b73d5cdaad17bc9987f73de8dbb2fd7ace1" - }, - { - "source": "tests/test_sampling_filters_cuda.py", - "path": "tests/test_sampling_filters_cuda.py", - "source_sha256": "3809c8b0fbba5c7c754d0530c79d42093649a616664e94a56338a134b7102a20", - "import_sha256": "f346ad33e98deb2f5dd591827967cda2f1aced441a4285eb8da8f3416d95ea78" - }, - { - "source": "tests/test_sampling_fused_cuda.py", - "path": "tests/test_sampling_fused_cuda.py", - "source_sha256": "61f9914b88dee8398387cbcd092f9fdb525e22c07f6e85597b926a981596fc19", - "import_sha256": "d8d98ef470902a7445491e63b6c04167c0007a1c153e103188235f8f920c765f" - }, - { - "source": "tests/test_sampling_guard_cuda.py", - "path": "tests/test_sampling_guard_cuda.py", - "source_sha256": "241c3500be9ff31b8d06fa247243dce6bc71ff73a26352d2f3b74ca5d0186caf", - "import_sha256": "85bcc68976cae15aef497e049df39f73214037c21be0471f54637b1a3b2a22c2" - }, - { - "source": "tests/test_scheduler.py", - "path": "tests/test_scheduler.py", - "source_sha256": "48e4b121944e3e9568d8ab8e09d02b385e652e7f05d15b34ffff50a54175b65c", - "import_sha256": "1881c109f9f9eb124328dbecc446b0431e6c51f879333de02a48f2ef93a777e9" - }, - { - "source": "tests/test_server_media.py", - "path": "tests/test_server_media.py", - "source_sha256": "9d4830988712713679d465dc7d876e36a98fee41b324eeba8cc8b27ffe0e4d8b", - "import_sha256": "fb243364f7af6b3809ef288c0cc142c6df52e42e8fdf4daaf586f580da829062" - }, - { - "source": "tests/test_sharded_embedding.py", - "path": "tests/test_sharded_embedding.py", - "source_sha256": "c285fd91fdfea58f4ec152402ac3b8deca3f9d0eea5c599d2689d5db4a332b4b", - "import_sha256": "2d1903d3e2d4df5acb351e8919044b5eab33483f5cfe43915e03be0a399eb687" - }, - { - "source": "tests/test_sparse_expert_transport.py", - "path": "tests/test_sparse_expert_transport.py", - "source_sha256": "8168ea0ad4d657f15fa70a5dc4f8928849c594f0b9b93ae947a4e12fc341edad", - "import_sha256": "c387dd777c3da9a61d19cd8fd1e049397cf14fe5ca95aded05439c3b28f72145" - }, - { - "source": "tests/test_topology_groups.py", - "path": "tests/test_topology_groups.py", - "source_sha256": "e3ef4b2e7b776700a27891238734508d77fa32c3808663388316abf1ab6439ea", - "import_sha256": "55679f97fc811adb4838300b43b4a7ebd3e95086f461038ee311b4b36563b31f" - }, - { - "source": "tests/test_type_libraries.py", - "path": "tests/test_type_libraries.py", - "source_sha256": "9d98a6ea58599c6ceccc8ce196aff21ac97fc31e3cf11c9edbf362717e805706", - "import_sha256": "9d98a6ea58599c6ceccc8ce196aff21ac97fc31e3cf11c9edbf362717e805706" - }, - { - "source": "tests/test_vision_cache.py", - "path": "tests/test_vision_cache.py", - "source_sha256": "6091a66c208ff82f5a3a5e04797420e559df53c6257414cb67eb869127188b3b", - "import_sha256": "071722870daff845638a714b6a28afd87b2eb363439a1efa14f3eb0612a7b9a7" - }, - { - "source": "uv.lock", - "path": "uv.lock", - "source_sha256": "abcb67c16b81d7569fa678ce0cf73b59ef41a17f675daaad3cdce3f07c23f29b", - "import_sha256": "dc19dc608337f68357af9ab87e790a8070f9388afe383f83268b2ff920af2a20" - } - ], - "excluded_source_paths": [] -} diff --git a/third_party/upstream/source-audit.md b/third_party/upstream/source-audit.md deleted file mode 100644 index 2e53aa7..0000000 --- a/third_party/upstream/source-audit.md +++ /dev/null @@ -1,255 +0,0 @@ -# OrbitInfer:源码审计、架构与实施计划 - -核查日期:2026-11-14。2026-11-15 已按此架构导入首批源码,详见[移植记录](source-import.md)。GPU 验证和 OrbitKV 接入尚未完成。 - -**从 Resin 的 Rust 调度与资源核心、现有 DeepSeek CUDA runner 起步,以 DeepInfer 为模型执行参考和对照基线。** OrbitInfer 补充 NVIDIA 产品配置、DeepSeek 验收和 OrbitKV 适配。第一版保留 PyTorch/CUDA 算子。 - -首个硬件目标已确定为 **SM100 / Blackwell**。首个模型固定为源码实际支持的 `deepseek-ai/DeepSeek-V4.1-Flash-1017`,权重 revision `1bae0e1088993b9b82de08f59de5be5a34e42170`。先做文本生成;多模态、其他 DeepSeek checkpoint、Qwen、Kimi、GLM 分别验收,不能凭品牌名宣称兼容。 - -## 1. 核查范围与版本 - -| 项目 | 本次固定 revision | 主要用途 | -| --- | --- | --- | -| [DeepInfer][d-repo] | `a679bb86ce36e883a98a97733492b686e5b9a08e` | 官方模型执行、CUDA 算子、并行与数值基线 | -| [Resin][r-repo] | `a1989594d81921be3bffe43804eb0550b216ca64` | Rust 请求生命周期、调度、页面租约和 CUDA runner | -| [OrbitKV][k-arch] | `01d9e36c6eed2910368e86cff3cd0f6b3df75ed1` | 外部缓存身份、DRAM/SSD/peer 层、查询与恢复协议 | -| [OrbiTE][t-copy] | `fbf21da8bfd7f2aec1614bc71126f09297f0cffa` | 注册内存、字节传输、CUDA 事件和物理退休 | - -这是源码审计,不是性能实测。DeepInfer 从本机只读 checkout 核查;其余按上表 GitHub revision 读取。本机 OrbitKV/OrbiTE 副本落后于远端,本设计使用远端版本,未改动两个项目。 - -## 2. 源码结论 - -### DeepInfer 提供什么 - -公开版包含 C++ 调度器、本地 GPU KV 管理、Python 执行层,以及 FlashMLA、DeepGEMM、DeepEP、DeepSelect、TileKernels/TileLang 算子。NVIDIA 路径要求 SM100、CUDA 13;上游还有 Ascend 路径,OrbitInfer 不构建它。[D1][d-repo] - -一次执行的关键链路是:`LLMEngine.step` → C++ `scheduler.schedule` → 状态复制/清零 → `worker.execute_model_async_step` → 当前 forward 与前一步采样结果回收。`DeferredRelease` 和 `pre_sampler_sync` 控制序列、状态池及复制事件的释放;`zos.hpp` 准备重叠执行元数据。Rust 并不会自动比这些 C++ 调度代码快。[D2][d-engine] [D3][d-worker] [D4][d-scheduler] - -公开发布不等于完整线上系统:线上 serving wrapper 和完整外部缓存/P-D 集成没有全部开放。`init_pd_split`、`post_schedule`、`do_save_kvcache` 等保留扩展点。它已有本地 `KVCacheManager` 和 `GPUCacheInterface`,不能把这些说成缺失;缺的是完整的外部缓存产品链路。[D2][d-engine] [D5][d-cache] [D6][d-qa] - -### Resin 已经省掉哪些重写 - -`resin-core::EngineCore

` 不依赖 Python/Torch。它已经拥有准入、优先级、prefill 分块、batch 规划、统一逻辑页面分配、prefix 复用、reader/writer lease、结果提交及错误隔离。Python runner 拥有权重、物理 GPU arena、CUDA graph、算子和 collective。双方通过不可变执行计划交接。[R1][r-arch] [R2][r-engine] - -原生核心已有 `PrefixStorage` / `PrefixStorageOperation`:lookup 和 transfer 分开,作业数量有上限,未完成的 load/store 保留页面。`Some(0)` 对 transfer 表示已经结束的失败,`None` 表示仍未取得完成证明。这个接口适合接外部缓存,但不能把 OrbiTE 的 `Unquiesced` 转成 `Some(0)`。[R3][r-cache] - -Resin 也已有 `DeepSeekV41CudaRunner`,无需从空白模型实现开始。其 prepared forward 参考 DeepInfer `774ce244b87cc0eb8d4b94ab9f5de2ab22049f01`;本次检查到该 revision 至上表 DeepInfer revision 的 engine/worker/model_executor 路径没有新增变更。仍须固定两边依赖与权重后运行数值对照。[R4][r-model-readme] - -### 不能直接视作等价的地方 - -| 源码事实 | OrbitInfer 决策 | -| --- | --- | -| Resin 默认 `swa_replay_tokens=140`,遵循 DeepInfer 的近似 SWA 重建;完整 producer KV 不代表 consumer 的全历史计算完全相同 | 首个正确性基线用 `swa_replay_mode="off"`。`mask`/`compact` 以后单独测质量,不能只做两个近似实现互比 | -| Resin DSpark 使用固定草稿长度;加载 confidence head 不等于实现 DeepInfer 的自适应长度策略 | 首先关闭 speculative decode;随后验证固定长度,最后才评估是否移植自适应策略 | -| 默认 expert backend 是 FlashInfer;DeepGEMM 路径要求 `attn_dp=ep=world_size`、CP1、expert TP1;单 KV-head MLA 的 `attn_tp=1` | 启动时校验具体 mesh。禁止静默切 backend、并行形状或精度 | -| DeepGEMM persistent grid barrier 要求整个 GPU;runner 会等待 Engram 的 NCCL fetch 后才启动该计算 | 不承诺低 CTA 拷贝核可无条件并发。背景 Restore 先用 DMA,干扰测试后再启用 SM kernel | -| Resin 当前没有 pipeline parallel;Engram 的 host-RDMA fetch 未实现 | 首版限定一个能容纳模型的设备域,不做跨节点 EP、CPU expert/Engram offload | -| `EngineCore::close(&mut runner)` 要求调用方提供 runner;core Drop 不负责 device runtime。Python 层保存失败 completion,但对象销毁后的物理持有仍须验证 | 把 failed-event + 最后一个 Python 引用释放列为接入硬门槛。不能仅因存在 `poisoned`/`quarantined` 标志就认为 DMA owner 永久安全 | - -以上来自模型说明、执行源码与失败处理:[R4][r-model-readme] [R5][r-model-cuda] [R6][r-py] [R7][r-failure]。最后一项是待验证的生命周期边界,并非本次已复现的上游漏洞。实现前须补 CPU 假 completion 和 GPU 故障控制;若确认缺陷,再形成最小修复。 - -## 3. 三个项目的职责 - -OrbitInfer 是推理引擎;OrbitKV 是外部 KV 缓存;OrbiTE 是传输引擎。三个项目都属于 AI infrastructure,但各自拥有不同资源。 - -```mermaid -flowchart TB - A[请求、tokenizer、流式输出] --> B[OrbitInfer:复用 Resin EngineCore] - B --> C["DeepSeek CUDA runner
权重、HBM arena、算子、graphs"] - C --> D[NVIDIA GPU / NCCL / expert collectives] - B --> E["薄 Rust OrbitKV adapter
语义身份、页面租约、完成状态映射"] - E --> F[已有 OrbitKV native client] - F --> G["每主机一个 OrbitKV Manager
缓存策略、DRAM / SSD / peer 状态"] - F --> H[OrbiTE:进程内 Restore 拷贝与事件] - G --> I[OrbiTE:file / peer / GPU 字节传输] - H --> C -``` - -| 资源或决策 | 唯一负责者 | 其他项目只做什么 | -| --- | --- | --- | -| 请求次序、prefill/decode、batch、采样进度 | OrbitInfer 的现有 Resin core/runner 分工 | OrbitKV 不调度 token,OrbiTE 不决定生成策略 | -| 逻辑 HBM 页面、COW、request/page lease | Resin core | adapter 持有租约,不另建一个 page allocator | -| 权重、物理 HBM arena、graph workspace、collective | NVIDIA runner | 原生传输注册必须保留实际 tensor/allocation owner | -| 缓存语义 key、可恢复范围、层级策略、DRAM/SSD 副本 | OrbitKV | OrbitInfer 声明布局与可恢复边界,不复制 Manager | -| 缓存 membership、目录及源节点授权 | OrbitKV/部署者 | OrbiTE 接收 endpoint 和权威 epoch;不做服务发现 | -| GPU↔DRAM、GPU↔SSD、peer bytes 和物理退休 | OrbiTE | adapter 不写 memcpy/kernel/socket/file fallback | -| 模型的 all-reduce、expert dispatch/combine | runner 及 NCCL/既有 expert 库 | OrbiTE 不充当 collective 库 | - -OrbitKV 当前使用每主机独立 Manager、UDS bootstrap 和 iceoryx2 控制协议。raw DRAM Restore 可在推理进程内通过 native client + OrbiTE 执行;Publish、SSD/codec Restore 保持 Manager owner。这里直接接已有路径。[K1][k-arch] [K2][k-restore] - -OrbitInfer 是新消费者,不要求 OrbitKV 放弃 vLLM/SGLang,也不改变 OrbiTE 的 NIXL 上游路线。首版不做 P/D;以后若做,应沿既有 NIXL 设计及其退休条件推进。 - -## 4. 直接复用、适配、移植与排除 - -| 部分 | 做法 | 原因/边界 | -| --- | --- | --- | -| `resin-core` engine/scheduler/planning/resources | 固定 revision 复用 | 已有事务式提交、页面租约与回滚;不重写同一套调度 | -| `resin-py`、CUDA completion、graph 工具 | 复用并做窄补丁 | 保留 runner 绑定和异步 completion;补生命周期证明与 native storage 注入 | -| `DeepSeekV41CudaRunner`、cache/batch/model 文件 | 作为首个模型包适配 | 默认配置收窄到文本、确定 mesh、关闭近似及投机优化 | -| FlashMLA/DeepGEMM/DeepEP/DeepSelect 等 | 按 runner 实际依赖选择并固定 | 使用成熟 CUDA/C++/TileLang 实现;不为语言比例翻译内核 | -| DeepInfer `scheduler.hpp`、`zos.hpp`、`llm_engine.py` | 行为参考与测试对照 | 不把其调度/worker 与 Resin core 并排运行 | -| DeepInfer 尚未覆盖的模型优化 | 后续逐项移植 | 只有数值或 profile 证据证明需要时才做 | -| OrbitKV native client 与 OrbiTE | 调用已有 owner | 新写的只有模型布局/身份/租约适配,不复制传输和缓存层 | -| Resin `resin-http`、`resin-recipe` | serving 阶段复用 | 先离线;不再造 HTTP、聊天模板、流式解析框架 | -| Ascend/CANN、MLX/Metal、ROCm、CPU 模型执行 | 排除构建与依赖 | 仅 NVIDIA 推理;CPU 用于调度、工具和测试 | -| 新动态 plugin ABI、通用 tensor runtime、分布式 coordinator | 不做 | 第一个 runner 静态注册;真正需要第二个实现时再扩大接口 | - -**Rust 的范围:**请求/资源生命周期、调度、缓存 adapter、协议及 serving 控制。**暂留 Python 的范围:**checkpoint 装载、PyTorch tensor 与既有 runner/kernel glue。GPU 数学仍由 CUDA 库执行。独立模型包已经提供扩展点,不需在第一阶段再定义 ABI。 - -这不是零改动接入:Resin 原生 `set_prefix_storage` 可接 Rust provider,但 Python 构造器目前接的是 Python storage 对象。需要在同一个 `resin-py` 绑定里加入窄的 native provider 注入点,并交接 arena owner;不创建第二个 EngineCore。先做该接口的最小验证,再决定补丁是否长期维护。[R2][r-engine] [R6][r-py] - -初期保留 Resin 已有的 core、Python binding、HTTP、recipe 四个 Rust crate;后两者是绑定的真实依赖。源码已迁入 OrbitInfer 命名空间,保留固定 revision、原始文件哈希和许可证;不增加 scheduler、allocator、transfer、registry 等空 crate。OrbitInfer 固定 Rust 1.96,不继承 OrbiTE 的 1.85 承诺。首个对照环境用 Linux、Python 3.12 和共同验证的 CUDA 13 工具链,逐项锁定 Torch、CUDA bindings 和内核依赖。 - -Resin 和 DeepInfer 顶层采用 Apache-2.0。复用时保留文件来源、版权及 NOTICE,并逐项记录内核、checkpoint 与 tokenizer 的独立许可证;此设计不代表权重再分发许可已审计。 - -## 5. 缓存接入:先正确持有,再异步重叠 - -### 不把前缀 hash 当作完整身份 - -Resin Python storage adapter 把 token/namespace 编成 byte keys;OrbitKV 已有 `StateKey`、模型/计算/布局身份和恢复契约。应使用原生结构化 PrefixStorage 边界,交给 OrbitKV 的身份实现生成 key,不能反向解析 Resin 私有 hash 或另造一套 hash 协议。[R8][r-storage-adapter] [K3][k-identity] - -namespace 至少绑定 checkpoint 内容、tokenizer/模板、runner 实现与版本、精度/量化、缓存布局、并行分片及影响状态的配置。同 token 前缀不保证 bytes 可互换。变更权重、布局或计算策略后切换 namespace。 - -DeepSeek 的 cache manifest 必须覆盖 SWA window、压缩 KV 和索引的共享 source、FP32 compressor state、token history 及启用 DSpark 后的额外状态。Resin 的 `CacheGeometry` 将它们组织为四类 group,其中压缩 KV 与索引共享布局;不能当成普通逐层 dense K/V 两个张量。[R9][r-model-cache] - -OrbitKV 的 Rust 契约能描述 full/window/checkpoint 恢复;当前 serving 入口仍拒绝 recurrent/SSM,不应把类型系统的表达能力当作已发布模型支持。DeepSeek 的这些 group 必须通过实际布局注册、restore 和冷计算对照;不支持的 state 表示应启动失败,不能补零后当作命中。[K3][k-identity] [K4][k-hybrid] - -### 最小生命周期 - -1. runner 一次性分配并绑定 arena;adapter 注册实际 backing,持有 Python/CUDA owner。 -2. core 为请求取得逻辑页面租约;lookup 未命中时按普通 prefill。lookup 本身不授予写入任意页面的能力。 -3. 命中后先固定目的页面和源 lease,再验证完整布局;Restore 交给 OrbitKV native client。未取得证明前不把这些页面重新分配。 -4. 首次实现等最终 drain,再把成功通知给 PrefixStorage。以后才加入每层 readiness:把事件依赖插入模型 stream,不把“事件已发布”当作“复制完成”。 -5. forward completion 验证后提交请求进度。Publish 保留 producer readiness 和页面读租约,直到实际传输退休。 -6. 取消或关闭先停新工作;等待当前 runner、Restore/Publish 和 native registration 的释放证明。失败对象留在同一 owner 的有界隔离状态,容量仍计费。 - -| 情况 | 对外结果与复用规则 | 必须有的负向控制 | -| --- | --- | --- | -| 明确未提交的验证失败 | 返回错误,可回滚租约;无设备副作用 | 提交计数为零,guard bytes 不变 | -| CUDA/Restore 已接受、尚未结束 | pending;页面、tensor、源 lease 保留 | cancel/timeout/early drop 后地址不能被重新分配 | -| `wait_restore_enqueued` 已返回 | 仅证明 readiness 事件发布 | 事件未完成时不能释放 backing 或源 grant | -| 失败且确认所有访问已结束 | 返回失败,可按合法边界重新计算 | 原目的页释放前已收到最终 receipt | -| `Unquiesced` 或 unknown effect | 不返回“已退休失败”;停止该 owner 的新工作并保留预算 | 查询/销毁失败后无重试 destroy、无重放不确定 WRITE | -| Manager/peer 死亡 | 按本地复制、远端授权分别求证;死亡/超时不自动等于 drain | 活跃 DMA 的页不回池,新 epoch 拒绝旧描述符 | -| 模型 step 已改写 persistent state 后失败 | 隔离该执行状态;禁止原地再跑同一步 | sampling 失败后不重复推进 compressor/window | - -OrbitKV 的现有文档明确区分 Ready、Drained、Reaped;局部 readiness 不能替代源端回收确认。[K1][k-arch] [K2][k-restore] - -还有一个接入门槛:OrbitKV 当前 Publish 使用 engine block ID,不能把 context/source epoch 当作每页重新分配的 generation。必须由 Resin 的真实 page lease 把页面固定至退休;若协议需要 generation,它只能由真实分配 owner 签发。adapter 自增计数不能制造回收证明。[K3][k-identity] - -取消和最后引用释放的安全逻辑归现有 Engine/runner/transfer owner;不加一个后台“回收协调器”。失败保留的数量和字节有硬上限,达到上限停止准入,并记录诊断。安全关闭不能依赖 Python GC 的执行时机。 - -## 6. DeepSeek 优先,如何扩展模型 - -首版只注册 `deepseek_v41`。模型适配器声明 checkpoint/config 校验、cache group/retention、允许的 mesh、精度、算子要求、graph 能力及采样/投机能力。公共调度只看这些事实,不写模型名分支;模型特有的 RoPE、MoE、压缩和 checkpoint 恢复在 runner 内。 - -| 扩展目标 | 已发现的可用参考 | 加入条件 | -| --- | --- | --- | -| DeepSeek 首个 checkpoint | Resin DeepSeek runner + DeepInfer | 完整冷执行、恢复一致性、生命周期与 SM100 安装测试 | -| 其他 DeepSeek checkpoint | 单独审计配置/权重/算子差异 | 不因同一品牌自动复用缓存 namespace | -| Qwen | Resin 已有 CUDA `qwen3_5` 包;并非只有 MLX 实现 | 先选精确 checkpoint;混合 attention/state 需 OrbitKV 恢复控制。MLX 不进入依赖 | -| GLM | Resin 已有 CUDA `glm5_3_flash` 包,声明 SM100/103 | 明确量化、路由、layout 与实际 GPU;不能推广到全部 GLM | -| Kimi | 本次固定树未发现同名专用 runner | 先读所选 checkpoint 的官方实现,再评估 MLA/MoE 复用;不预设只改配置即可 | - -用户目标是 DeepSeek 先完成,再扩展 Qwen、Kimi、GLM。后三者独立排期,不把它们塞进首个版本。现有 runner 是降低工作量的候选,不是 OrbitInfer 已支持的证据。[R10][r-qwen] [R11][r-glm] - -## 7. 分阶段实施与退出条件 - -阶段按验收推进,不给缺少硬件条件的日历承诺。每个阶段拆成可独立审查的 PR;一个 owner 一次变更。 - -| 阶段 | 交付与代码范围 | 完成条件 | -| --- | --- | --- | -| P0:复用可行性 | 固定源码/依赖/许可证;最小构建;Resin storage 注入与 arena-owner 持有原型;记录唯一调度/分配 owner | CPU conformance 通过;poison、drop、部分提交后所有实际 backing 可追踪。若现有接口不能满足,先给窄补丁,不启动完整引擎重写 | -| P1:DeepSeek 离线正确性 | 文本 runner、启动配置校验、权重装载、普通 prefill/decode;`overlap=false`、draft=0、SWA replay off、外部缓存关闭 | 在足够 HBM 的 SM100 设备域跑完整 checkpoint;对照 DeepInfer logits/生成/长上下文及 batch invariance;明确数值容差与未实现配置 | -| P2:现有执行优化 | 逐项打开 CUDA graph、overlap、已有 prefix reuse;沿用既有 worker/domain | eager/graph、同步/重叠结果对照;取消、边界跨页、终止后 successor 仍在飞的资源控制通过;没有第二个调度器 | -| P3:OrbitKV 接入 | 窄 Rust PrefixStorage adapter、模型 identity/layout、page lease 与 readiness;先本机 DRAM,再 SSD/peer | 完整恢复等价于冷计算;命中/部分命中/不兼容/源重启/早退均通过。DRAM、SSD、跨主机分别安装验收,不合并成一个“缓存已支持”结论 | -| P4:最小服务 | 复用 Resin HTTP/recipe;有界队列、流式输出、超时、关闭、必要指标 | 两进程与真实 HTTP 控制;断流取消不释放仍在使用的页;TTFT/ITL、队列等待、HBM/外部缓存占用可解释 | -| P5:测量驱动优化与模型扩展 | 固定长度 DSpark,再评估自适应草稿;近似 SWA 独立质量开关;按顺序审计后续模型 | 真实接受 token 的端到端收益和质量报告;每个新模型完成自己的数值、状态恢复与生命周期验收 | - -P1 不要求“一张卡跑完整模型”。先测权重、Engram、activation、graph、KV 和通信 workspace 的实际峰值,确定设备数与 mesh。小形状 CPU/GPU 测试不能代替完整 checkpoint;源码里示例的多卡数也不是最小部署承诺。 - -首版只保留一个已验证的 expert backend 配置。默认 FlashInfer 可先做 runner 基线;若所选 SM100 profile 验证使用 DeepGEMM,则固定其受支持的 mesh。两者不要同时作为首版默认去优化。 - -## 8. 测试映射与新控制 - -下表是源码核查得到的验收清单。已移植测试的 CPU 结果见[移植记录](source-import.md);GPU、跨项目和完整模型验收仍待执行。 - -| 上游测试 | OrbitInfer 要证明什么 | 环境 | -| --- | --- | --- | -| Resin `rejected_execution_restores_requests_and_releases_batch_access` | 未接受失败才可重排原 cohort;不影响其他请求 | CPU | -| Resin `failed_release_keeps_request_ownership_quarantined` | release 失败保留实际 request/page owner | CPU + GPU 安装 | -| Resin `frontend_event_failure_quarantines_before_native_commit` | Python 结果构造失败不造成半提交 | CPU | -| Resin `test_cancel_drains_accepted_work_and_preserves_other_request_events` | cancel 只取消目标,不丢其他结果或提前释放 | CPU + CUDA | -| Resin `test_stop_result_precedes_successor_completion_but_release_waits` | 逻辑结束与仍在飞 successor 的退休分离 | CPU + CUDA | -| Resin `test_async_failure_quarantines_the_original_writer_lease` | 错误仍持有原写租约,非一个无 backing 的标记 | CPU + CUDA | -| Resin `test_failed_read_fence_preserves_unretired_buffers` | 外部 Restore fence 失败不丢目的内存 | CPU + OrbitKV/OrbiTE | -| Resin `test_queued_reads_retire_before_eviction_and_close`、`test_storage_inflight_limit` | 驱逐/关闭等待真实退休;存储并发有界 | CPU | -| Resin `test_engine_quarantines_state_advanced_before_sampling_failure` | 已改变模型状态的 step 失败不重新执行 | CPU + CUDA | -| Resin `test_ratio2_rejected_tail_needs_no_relocation_or_accepted_state_mutation` | speculative reject 不污染先前接受的压缩状态 | 模型算子 | -| Resin `test_compact_metadata_preserves_page_ownership_masks_and_placeholder_history` | 将来打开近似 replay 时保留页面、mask 和完整 producer 语义 | CPU + 独立质量门槛 | -| DeepInfer `test_round_trip_consistency`、`test_swa_only_state_round_trip` | window/compressor 等状态保存恢复完整,不只检查 KV 主体 | GPU | -| DeepInfer `test_batch_invariant`、`test_kl`、`test_offline_inference` | batch 形状不改变合法结果;完整模型数值与生成对照 | SM100 安装 | -| DeepInfer `test_interleaved_exit_no_leak`、`test_forbid_across_threads` | CUDA 同步限制在并发调用后正确恢复 | CPU;独立 GPU 检查 | - -来源:[R12][r-core-tests] [R13][r-overlap-tests] [R14][r-storage-tests] [R15][r-failure-tests] [R16][r-contract-tests] [R17][r-yoco-tests] [D7][d-state-tests] [D8][d-tests]。 - -还要新写四组跨项目负向控制: - -- 同 token、不同 checkpoint/layout/量化/并行分片:必须 miss 或拒绝,不能读取错误 cache。 -- Restore pending 时 cancel、GC、注销、Manager 重启、page allocator 压力:旧页及其 generation 不复用,guard bytes 不变。 -- event query/synchronize/destroy 出错后丢掉最后一个 Python owner:实际 arena、completion、registration 和预算仍在;明确关闭返回未退休。 -- 部分层已获得 readiness、后续层传输失败:禁止报告整段成功或重放不确定写入;已启动 forward 的影响进入该请求的错误隔离。 - -CPU CI 运行 Rust 生命周期、fake runner、计划/布局、绑定和协议控制;SM100 安装门槛运行实际内核、完整模型、guard bytes、fault 和数值检查。时间敏感控制采用事件/屏障构造交错,再重复执行。每个故障路径验证 owner 数量和预算,不只检查错误码。 - -## 9. 性能比较与边界 - -比较对象是固定版本 DeepInfer,以及同版本未修改的 Resin runner。固定模型/权重、精度、mesh、内核版本、GPU/driver、prompt/output 分布、batch、warmup、cache 状态和近似/投机开关。先关闭外部缓存比较推理本身,再单独比较 DRAM/SSD/peer 命中收益。 - -至少报告 TTFT(首 token 时间)、ITL(后续 token 间隔)的 p50/p99、实际输出 tokens/s、CPU 时间、HBM 峰值、缓存占用与传输退休延迟。首请求的编译/装载与稳态分开。speculation 报真实接受率;不以强制接受草稿、强制均匀专家路由或投影吞吐作为实际服务性能。 - -背景 Restore 的 DMA/SM-kernel 选择沿用 OrbiTE 的 intent、priority、CTA cap。必须在模型计算同时运行时测 ITL 干扰;单独 copy GB/s 不证明 decode 收益。NCCL/DeepGEMM 的持久核与 stream 依赖也纳入测试。 - -本设计没有声称比 DeepInfer/Resin 更快,也没有把源码支持视为硬件验证。GPU、RDMA、GDS、NVLink 等安装结果分别记录。 - -## 10. 第一批明确不做的事 - -- Ascend/其他加速器适配、通用跨硬件 runtime、Rust 重写所有 CUDA kernels。 -- 多模型热切换、任意动态 plugin ABI、自己的 collective 或第二套缓存 Manager。 -- 首版多模态、跨节点 EP/P-D、自创集群发现、CPU expert/Engram offload。 -- 为未来 Qwen/Kimi/GLM 提前抽象所有 attention/state 类型。 -- 修改 OrbitKV/OrbiTE 的产品边界,或把它们当前未运行的硬件门槛算作 OrbitInfer 已通过。 - -建议首先实施 P0:验证同一个 Resin EngineCore 能持有 native OrbitKV adapter 和真实 GPU arena owner,并把异常销毁的证明补齐。这个小步骤决定后续能否保持复用,而不是从一开始承担维护完整 DeepInfer 分叉的成本。 - -[d-repo]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/README.md -[d-engine]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/deepinfer/engine/llm_engine.py -[d-worker]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/deepinfer/worker/worker.py -[d-scheduler]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/csrc/engine/scheduler.hpp -[d-cache]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/deepinfer/worker/cache_engine.py -[d-qa]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/doc/inference_deployment_qa.md -[d-state-tests]: https://github.com/deepseek-ai/DeepInfer/blob/a679bb86ce36e883a98a97733492b686e5b9a08e/tests/worker/test_state_cache_ops.py -[d-tests]: https://github.com/deepseek-ai/DeepInfer/tree/a679bb86ce36e883a98a97733492b686e5b9a08e/tests -[r-repo]: https://github.com/ekzhang/resin/tree/a1989594d81921be3bffe43804eb0550b216ca64 -[r-arch]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/docs/architecture.md -[r-engine]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/crates/resin-core/src/engine.rs -[r-cache]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/crates/resin-core/src/resources/paged_kv.rs -[r-model-readme]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/resin_models_cuda/deepseek_v41/README.md -[r-model-cuda]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/resin_models_cuda/deepseek_v41/cuda.py -[r-model-cache]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/resin_models_cuda/deepseek_v41/cache.py -[r-py]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/crates/resin-py/src/lib.rs -[r-failure]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/resin/backend/cuda/failures.py -[r-storage-adapter]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/resin/_cache_storage.py -[r-qwen]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/resin_models_cuda/qwen3_5/__init__.py -[r-glm]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/resin_models_cuda/glm5_3_flash/__init__.py -[r-core-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/crates/resin-core/tests/engine.rs -[r-overlap-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/tests/test_engine_overlap.py -[r-storage-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/tests/test_cache_storage.py -[r-failure-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/tests/test_cuda_failures.py -[r-contract-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/tests/test_deepseek_v41_contracts.py -[r-yoco-tests]: https://github.com/ekzhang/resin/blob/a1989594d81921be3bffe43804eb0550b216ca64/models/cuda/tests/test_deepseek_v41_yoco.py -[k-arch]: https://github.com/OrbitInfra/OrbitKV/blob/01d9e36c6eed2910368e86cff3cd0f6b3df75ed1/docs/architecture.md -[k-restore]: https://github.com/OrbitInfra/OrbitKV/blob/01d9e36c6eed2910368e86cff3cd0f6b3df75ed1/docs/engine-local-restore.md -[k-identity]: https://github.com/OrbitInfra/OrbitKV/blob/01d9e36c6eed2910368e86cff3cd0f6b3df75ed1/docs/state-identity.md -[k-hybrid]: https://github.com/OrbitInfra/OrbitKV/blob/01d9e36c6eed2910368e86cff3cd0f6b3df75ed1/docs/hybrid-recovery.md -[t-copy]: https://github.com/OrbitInfra/OrbiTE/blob/fbf21da8bfd7f2aec1614bc71126f09297f0cffa/docs/local-copy.md