Skip to content

Actions: deepspeedai/DeepSpeed

Actions

cpu-torch-latest

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
cpu-torch-latest.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
Give Muon's momentum the dtype of the gradient it is combined with cpu-torch-latest #9671: Pull request #8483 opened by alanhuangyoo
Action required alanhuangyoo:fix/muon-momentum-dtype-both-paths
alanhuangyoo:fix/muon-momentum-dtype-both-paths
Action required
Fixed memory leak in loss backward with AC & zero3 & single-model-multi-branch network cpu-torch-latest #9670: Pull request #8482 synchronize by supermeng
Action required supermeng:fix/zero3_ac_mul_branck_mem_leak
supermeng:fix/zero3_ac_mul_branck_mem_leak
Action required
Fixed memory leak in loss backward with AC & zero3 & single-model-multi-branch network cpu-torch-latest #9669: Pull request #8482 opened by supermeng
Action required supermeng:fix/zero3_ac_mul_branck_mem_leak
supermeng:fix/zero3_ac_mul_branck_mem_leak
Action required
cpu-torch-latest cpu-torch-latest #9668: Merge group checks requested
38m 56s
38m 56s
[muon] Reconcile the momentum dtype when a checkpoint is restored cpu-torch-latest #9667: Pull request #8433 synchronize by delock
40m 31s alanhuangyoo:fix/muon-momentum-dtype-on-resume
alanhuangyoo:fix/muon-momentum-dtype-on-resume
40m 31s
cpu-torch-latest cpu-torch-latest #9666: Merge group checks requested
34m 2s
34m 2s
cpu-torch-latest cpu-torch-latest #9665: Merge group checks requested
33m 15s
33m 15s
cpu-torch-latest cpu-torch-latest #9664: Merge group checks requested
38m 48s
38m 48s
fix(inference): reject non-positive max_out_tokens at config validation cpu-torch-latest #9663: Pull request #8454 synchronize by chakshu-dhannawat
Action required chakshu-dhannawat:fix/max-out-tokens-positive-validation
chakshu-dhannawat:fix/max-out-tokens-positive-validation
Action required
Fix ZeRO parameter alignment for grouped_mm cpu-torch-latest #9662: Pull request #8277 synchronize by tohtana
38m 24s fwerkor:fix-8276-grouped-mm-alignment
fwerkor:fix-8276-grouped-mm-alignment
38m 24s
Count each expert once in the fp32 gradient-clipping norm cpu-torch-latest #9661: Pull request #8478 synchronize by alanhuangyoo
Action required alanhuangyoo:fix/fp32-clip-moe-norm
alanhuangyoo:fix/fp32-clip-moe-norm
Action required
cpu-torch-latest cpu-torch-latest #9660: Merge group checks requested
38m 44s
38m 44s
cpu-torch-latest cpu-torch-latest #9659: Merge group checks requested
37m 27s
37m 27s
cpu-torch-latest cpu-torch-latest #9658: Merge group checks requested
38m 4s
38m 4s
cpu-torch-latest cpu-torch-latest #9657: Merge group checks requested
35m 58s
35m 58s
feat(rollout): add continuous batching generation prototype cpu-torch-latest #9656: Pull request #8368 synchronize by delock
31m 21s nathon-lee:perf/opsd-continuous-batching-prototype
nathon-lee:perf/opsd-continuous-batching-prototype
31m 21s
fix(inference): break circular import in ops.transformer.inference cpu-torch-latest #9655: Pull request #8465 synchronize by chakshu-dhannawat
Action required chakshu-dhannawat:fix/inference-circular-import-7159
chakshu-dhannawat:fix/inference-circular-import-7159
Action required
cpu-torch-latest cpu-torch-latest #9654: Scheduled
38m 1s master
master
38m 1s
Carry the affine scale on the replicated map, not the split cpu-torch-latest #9653: Pull request #8477 synchronize by delock
37m 22s Achyuthan-S:affine-scale-on-replicated
Achyuthan-S:affine-scale-on-replicated
37m 22s
cpu-torch-latest cpu-torch-latest #9652: Merge group checks requested
37m 4s
37m 4s
Fix comms logger KeyError when log_name is omitted cpu-torch-latest #9651: Pull request #8267 synchronize by tohtana
39m 1s jinyouzhi:comms_logger
jinyouzhi:comms_logger
39m 1s
Recognize a norm by its shape, not by whether its class name was listed cpu-torch-latest #9650: Pull request #8479 opened by alanhuangyoo
Action required alanhuangyoo:fix/autotp-norm-by-shape
alanhuangyoo:fix/autotp-norm-by-shape
Action required
Count each expert once in the fp32 gradient-clipping norm cpu-torch-latest #9649: Pull request #8478 opened by alanhuangyoo
Action required alanhuangyoo:fix/fp32-clip-moe-norm
alanhuangyoo:fix/fp32-clip-moe-norm
Action required
Fix AutoEP global L2 clipping for sharded expert ownership cpu-torch-latest #9648: Pull request #8476 synchronize by gss10282023
Action required gss10282023:fix-autoep-ownership-aware-clipping
gss10282023:fix-autoep-ownership-aware-clipping
Action required
Fix Muon optimizer under ZeRO CPU offload and bound gather buffers cpu-torch-latest #9647: Pull request #8464 synchronize by jinyouzhi
Action required jinyouzhi:muon-cpu-offload-fix
jinyouzhi:muon-cpu-offload-fix
Action required