Skip to content

Actions: pytorch/executorch

Actions

MLX

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
mlx.yml
1,695 workflow runs
1,695 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Skip LpaiPartitionFallbackSupport when compiler_specs is None (forwar… MLX #1676: Commit 07b9feb pushed by Gasoonjia
42s main
main
42s
portable: accumulate in fp32 for Half/BFloat16 in softmax, log_softma… MLX #1675: Commit 8bb71cf pushed by GregoryComer
39s main
main
39s
Extend CPPCHECK scope to portable kernels MLX #1674: Commit e93a285 pushed by perheld
37s main
main
37s
[MLX] Reduce physical footprint memory in RingBufferKVCache for chunked prefill MLX #1673: Pull request #20341 synchronize by metascroy
50m 19s reduce-mem-sliding-kv
reduce-mem-sliding-kv
50m 19s
Split of "[re] Drop mig capability from tests" (#20330) MLX #1672: Commit fba18e2 pushed by meta-codesync Bot
53m 35s main
main
53m 35s
qwen3_5_moe: add OpenAI serving entrypoint MLX #1671: Pull request #20313 synchronize by mergennachin
1h 0m 4s llm-qwen35-moe-serving
llm-qwen35-moe-serving
1h 0m 4s
[cuda backend] reduce memory consumption on gemma4_31b by running embedding in int8 MLX #1670: Pull request #20351 opened by Gasoonjia
49m 46s g4-gguf-export-mem
g4-gguf-export-mem
49m 46s
Add missing evalue dep to test_cuda_mutable_state (forward fix for D1… MLX #1669: Commit 4576f8f pushed by Gasoonjia
38s main
main
38s
Fix pthreadpool subset corruption + executor deadlock; de-nest ExecuT… MLX #1668: Commit 68f0768 pushed by meta-codesync Bot
38s main
main
38s
Arm backend: Fix test_resnet18.py for Arm64 (#20335) MLX #1667: Commit 15b0d73 pushed by zingo
37s main
main
37s
Arm backend: Lower MXFP Linear to TOSA (#20065) MLX #1666: Commit 6dab6b1 pushed by zingo
1h 13m 3s main
main
1h 13m 3s
Fix sliding-window chunked prefill in the gemma4-31B runner MLX #1665: Pull request #20346 opened by digantdesai
43m 28s gh/digantdesai/65/head
gh/digantdesai/65/head
43m 28s
Make EAGLE-3 chain length K a runtime parameter MLX #1664: Pull request #20345 opened by digantdesai
1h 11m 35s gh/digantdesai/64/head
gh/digantdesai/64/head
1h 11m 35s
Add split-K mid-M SDPA Triton kernel for EAGLE-3 target_verify at long context MLX #1663: Pull request #20344 opened by digantdesai
1h 6m 48s gh/digantdesai/63/head
gh/digantdesai/63/head
1h 6m 48s
Update executorch-ubuntu-26.04-gcc15 docker image to ubuntu-26.04-gcc… MLX #1662: Commit 315a0d8 pushed by rascani
43s main
main
43s
[MLX] Reduce physical footprint memory in MLX MLX #1661: Pull request #20342 opened by metascroy
57m 26s reduce-mem-slot-recycling-fix
reduce-mem-slot-recycling-fix
57m 26s
[MLX] Reduce physical footprint memory in RingBufferKVCache for chunked prefill MLX #1660: Pull request #20341 opened by metascroy
59m 34s reduce-mem-sliding-kv
reduce-mem-sliding-kv
59m 34s
Extract _lora_call into a shared free function in lora.py (#20306) MLX #1659: Commit 0eb8247 pushed by meta-codesync Bot
42s main
main
42s
qwen3_5_moe: add OpenAI serving entrypoint MLX #1658: Pull request #20313 synchronize by mergennachin
42m 27s llm-qwen35-moe-serving
llm-qwen35-moe-serving
42m 27s
NXP backend: Extend unittest-nxp-neutron job timeout to 120 minutes… MLX #1657: Commit ef5c8a7 pushed by MartinPavella
45s main
main
45s
Arm backend: Support sampler-backed VGF grid_sample for RIFE (#20329) MLX #1656: Commit cf59749 pushed by bdemirb
40m 24s main
main
40m 24s
Arm backend: Reduce Qwen3 VL E2E footprint (#20328) MLX #1655: Commit 21a0191 pushed by bdemirb
43s main
main
43s
Arm backend: Add FP8 support for gather/scatter-based composite ops (… MLX #1654: Commit c37b6c3 pushed by YufengShi-dudu
36s main
main
36s
skip h2d copies between forward functions in gemma4-31b MLX #1653: Pull request #20286 synchronize by Gasoonjia
56m 13s export-D108661628
export-D108661628
56m 13s
Enable 128k context for Gemma4-31B CUDA MLX #1652: Pull request #20316 synchronize by Gasoonjia
40m 45s g4-128k-context
g4-128k-context
40m 45s