CI (self-hosted WebGPU backend)

Workflow runs for CI (self-hosted WebGPU backend)

• hexagon: vectorize softplus with HVX (#30435)

  queued:
  CI (self-hosted WebGPU backend)#33:Commit27c7a76pushedbyEAddario
  master

  Oct 16, 2026Queued
  More actions for workflow run 33

• opencl: Add MxFP4 MoE bin kernel for Adreno X1 devices (#30691)

  queued:
  CI (self-hosted WebGPU backend)#32:Commit2f2763dpushedbyEAddario
  master

  Oct 16, 2026Queued
  More actions for workflow run 32

• model : Granite Switch Decision (#30676)

  queued:
  CI (self-hosted WebGPU backend)#31:Commit23f7aabpushedbyEAddario
  master

  Oct 16, 2026Queued
  More actions for workflow run 31

• common: honor HTTP_PROXY/HTTPS_PROXY env vars in http client (#21752)

  queued:
  CI (self-hosted WebGPU backend)#30:Commit7182ef7pushedbyEAddario
  master

  Oct 16, 2026Queued
  More actions for workflow run 30

• model : MTP support for Mellum2.1 (#30612)

  cancelled:
  CI (self-hosted WebGPU backend)#29:Commit9a16f1dpushedbyEAddario
  master

  Oct 16, 20261d 0h 0m 2s
  More actions for workflow run 29

• vulkan: Optimize pool_2d when the pooling window is large (#30119)

  cancelled:
  CI (self-hosted WebGPU backend)#28:Commit77c4540pushedbyEAddario
  master

  Oct 13, 20261d 0h 0m 11s
  More actions for workflow run 28

• cuda : fix compiler warnings in mmq-load-tiles.cuh (#30443)

  cancelled:
  CI (self-hosted WebGPU backend)#27:Commitbfe65e1pushedbyEAddario
  master

  Oct 13, 20261d 0h 0m 2s
  More actions for workflow run 27

• convert : fix LoRA conversion with --outtype auto (#30433)

  cancelled:
  CI (self-hosted WebGPU backend)#26:Commit9d7d504pushedbyEAddario
  master

  Oct 13, 20261d 0h 0m 2s
  More actions for workflow run 26

• cuda: always use MMVQ for MUL_MAT_ID on sm_60 (#27828)

  cancelled:
  CI (self-hosted WebGPU backend)#25:Commit1623d8cpushedbyEAddario
  master

  Oct 10, 20261d 0h 0m 2s
  More actions for workflow run 25

• vendor: apply deep nested json patch from upstream (#30253)

  cancelled:
  CI (self-hosted WebGPU backend)#24:Commit10a60cfpushedbyEAddario
  master

  Oct 9, 20261d 0h 0m 2s
  More actions for workflow run 24

• CUDA: fix round issue, under MSVC the CPU and GPU agree (#30229)

  cancelled:
  CI (self-hosted WebGPU backend)#23:Commit79e2e74pushedbyEAddario
  master

  Oct 9, 20261d 0h 0m 9s
  More actions for workflow run 23

• CUDA: pass src1 precision to host MMQ config helpers (#30168)

  cancelled:
  CI (self-hosted WebGPU backend)#22:Commit8a1a9b5pushedbyEAddario
  master

  Oct 9, 20261d 0h 0m 9s
  More actions for workflow run 22

• sycl: FWHT optimizations (#29605)

  cancelled:
  CI (self-hosted WebGPU backend)#21:Commit46baf1fpushedbyEAddario
  master

  Oct 8, 20261d 0h 0m 2s
  More actions for workflow run 21

• chat: fix jinja parser for TranslateGemma (#30096)

  cancelled:
  CI (self-hosted WebGPU backend)#20:Commitbd4eeaapushedbyEAddario
  master

  Oct 7, 20261d 0h 0m 2s
  More actions for workflow run 20

• CI (self-hosted WebGPU backend)

  cancelled:
  CI (self-hosted WebGPU backend)#19:byEAddario
  master

  Oct 7, 20261d 0h 1m 55s
  More actions for workflow run 19

• feat: add GLM5Next MTP, optimize (#29928)

  cancelled:
  CI (self-hosted WebGPU backend)#18:Commitb9acf13pushedbyEAddario
  master

  Oct 7, 20261d 0h 0m 2s
  More actions for workflow run 18

• ggml-cuda: use per-thread stream for buffer-init padding memset (#28782)

  cancelled:
  CI (self-hosted WebGPU backend)#17:Commit51ce9c1pushedbyEAddario
  master

  Oct 6, 20261d 0h 0m 3s
  More actions for workflow run 17

• ci : winget urls must be separate strings (#29978)

  cancelled:
  CI (self-hosted WebGPU backend)#16:Commite5983d6pushedbyEAddario
  master

  Oct 5, 20261d 0h 0m 1s
  More actions for workflow run 16

• ggml-cpu: support BF16/FP16/FP32 K tails in tinyBLAS on x86 (#29806)

  cancelled:
  CI (self-hosted WebGPU backend)#15:Commita7b94dfpushedbyEAddario
  master

  Oct 4, 20261d 0h 0m 2s
  More actions for workflow run 15

• CUDA: fix MMQ memory fault if n_expert >> n_ubatch (#29941)

  cancelled:
  CI (self-hosted WebGPU backend)#14:Commitdd26678pushedbyEAddario
  master

  Oct 4, 20261d 0h 0m 2s
  More actions for workflow run 14

• imatrix: calculate activation-based statistics for new format (GGUF) …

  cancelled:
  CI (self-hosted WebGPU backend)#13:Commit0504396pushedbyEAddario
  master

  Oct 4, 20261d 0h 0m 2s
  More actions for workflow run 13

• model : Add LFM2.5-Encoder-350M and LFM2.5-Encoder-230M (#29862)

  cancelled:
  CI (self-hosted WebGPU backend)#12:Commitcb7934cpushedbyEAddario
  master

  Oct 3, 20261d 0h 0m 2s
  More actions for workflow run 12

• CUDA: fuse shared experts into MMVQ (#29184)

  cancelled:
  CI (self-hosted WebGPU backend)#11:Commitbed0a85pushedbyEAddario
  master

  Oct 2, 20261d 0h 0m 2s
  More actions for workflow run 11

• metal : add tensor API flash attention kernel for F16 KV (#29570)

  cancelled:
  CI (self-hosted WebGPU backend)#10:Commit926862epushedbyEAddario
  master

  Oct 2, 20261d 0h 0m 2s
  More actions for workflow run 10

• llama : silence unused-result warnings (#29839)

  cancelled:
  CI (self-hosted WebGPU backend)#9:Commit8d81559pushedbyEAddario
  master

  Oct 2, 20261d 0h 0m 22s
  More actions for workflow run 9