Repository navigation
Validate expression values and compile liveness-scheduled GPU readouts - #50
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Expression bindings containing opaque, Boolean, complex, object or text values could leak Python/NumPy arithmetic errors, and affine broadcasting failed outside the shared validation boundary. Validate numeric types and shapes before evaluation while retaining native Torch/JAX arrays and gradients, including user tensor subclasses.
Lower Pauli and factored readout DAGs once into immutable last-use register programs. Release dead array references without mutating tensor data, preserve output/fanout lifetimes, and account for live registers and owned bindings in primal workspace admission. Add an explicit CUDABackend compile_mode using checked TorchInductor programs and optional CUDA Graph replay. Transfer only active packed parameter columns, isolate compiled caches by policy/device/shape, and lock replay through the host output snapshot. Preserve the caller's selected CUDA index across scheduler workers.
Validation: numeric/gradient/compilation checks cover NumPy, Torch and JAX, dead-array lifetime, duplicate operands, higher derivatives, packed inputs and cache isolation. Physical CUDA tests now cover eager/default/reduce-overhead with repeated fresh batches. This environment has no CUDA device, so physical throughput and CUDA Graph replay remain hardware checks. CPU compilation checks do not establish GPU performance. Version: 0.25.0.