Skip to content

Actions: deepspeedai/DeepSpeed

Actions

python

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
python.yml
2,500+ workflow runs
2,500+ 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
Fuse mixed-precision gradients and parameter writeback into FusedAdam python #17804: Pull request #8673 opened by CharlesXu-HQ
Action required CharlesXu-HQ:perf/fused-adam-mixed-precision
CharlesXu-HQ:perf/fused-adam-mixed-precision
Action required
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17803: Pull request #8602 synchronize by delock
1m 24s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 24s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17802: Pull request #8602 synchronize by delock
1m 33s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 33s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17801: Pull request #8602 synchronize by delock
1m 56s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 56s
Fix: Replace assert statements with ValueError in runtime config validation python #17800: Pull request #8585 synchronize by mayuriphad
Action required mayuriphad:fix-config-asserts
mayuriphad:fix-config-asserts
Action required
Fix: handle None buffers in AutoTP checkpoint load python #17799: Pull request #8584 synchronize by mayuriphad
Action required mayuriphad:fix-autotp-buffer-none
mayuriphad:fix-autotp-buffer-none
Action required
[DON'T MERGE] Run multi-rank CPU unit tests in CI via LOCAL_SIZE python #17798: Pull request #8381 synchronize by delock
1m 43s delock:ci/cpu-multi-rank-local-size
delock:ci/cpu-multi-rank-local-size
1m 43s
AutoEP: reorder expert rows with gathers on the collective path python #17797: Pull request #8671 opened by yh0903
Action required yh0903:yh0903/autoep-gather-permute
yh0903:yh0903/autoep-gather-permute
Action required
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17796: Pull request #8602 synchronize by delock
1m 30s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 30s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17795: Pull request #8602 synchronize by delock
1m 29s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 29s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17794: Pull request #8602 synchronize by delock
1m 31s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 31s
python python #17793: Scheduled
1m 20s master
master
1m 20s
Harden ZeRO-1/2 offload gradient storage lifetime and stream ordering python #17792: Pull request #8632 synchronize by jeffra
1m 21s jeffra/913-fixes-master-sync-r2
jeffra/913-fixes-master-sync-r2
1m 21s
AutoEP + AutoTP folding under ZeRO-3 python #17791: Pull request #8670 opened by pengdurice
1m 36s pengdurice:peng-add-autoep-autotp-zero3-v1
pengdurice:peng-add-autoep-autotp-zero3-v1
1m 36s
Add an opt-in fused Triton RoPE for Hugging Face split-half models python #17790: Pull request #8663 synchronize by yh0903
Action required yh0903:yh0903/fused-rope
yh0903:yh0903/fused-rope
Action required
Fused RMSNorm: multi-row programs and a one-pass backward for narrow norms (stacked on #8645) python #17789: Pull request #8664 synchronize by yh0903
Action required yh0903:yh0903/fused-rmsnorm-narrow-rows
yh0903:yh0903/fused-rmsnorm-narrow-rows
Action required
python python #17788: Merge group checks requested
1m 42s
1m 42s
Replay AutoEP's cached DeepEP dispatch in the layout it was built with python #17787: Pull request #8636 synchronize by pengdurice
1m 42s yh0903:yh0903/autoep-cached-dispatch-do-expand
yh0903:yh0903/autoep-cached-dispatch-do-expand
1m 42s
Apply weight decay to the CPU Adagrad update python #17786: Pull request #8669 opened by vineethsaivs
Action required vineethsaivs:fix/cpu-adagrad-weight-decay
vineethsaivs:fix/cpu-adagrad-weight-decay
Action required
Apply the ZeRO-3 predivide postscale when the factor equals world size python #17785: Pull request #8668 opened by vineethsaivs
Action required vineethsaivs:fix/zero3-predivide-postscale
vineethsaivs:fix/zero3-predivide-postscale
Action required
Preserve convolution shapes and scaling in Muon updates python #17784: Pull request #8552 synchronize by vineethsaivs
Action required vineethsaivs:fix-muon-convolution-shape-20260916
vineethsaivs:fix-muon-convolution-shape-20260916
Action required
Skip ZeRO DeepCompile backward hooks for AutoSP python #17783: Pull request #8667 opened by jinyouzhi
Action required jinyouzhi:fix-autosp-deepcompile-hooks
jinyouzhi:fix-autosp-deepcompile-hooks
Action required
fix: attention_unfused alpha missing norm_factor squaring python #17782: Pull request #8666 opened by delock
1m 33s fix/pt-binding-alpha-squaring
fix/pt-binding-alpha-squaring
1m 33s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17781: Pull request #8602 synchronize by delock
2m 13s gma/qwen_ki_expr
gma/qwen_ki_expr
2m 13s
feat(rollout): segment-based kernel injection for hybrid-engine greedy decode python #17780: Pull request #8602 synchronize by delock
1m 32s gma/qwen_ki_expr
gma/qwen_ki_expr
1m 32s