Skip to content

Actions: verl-project/verl

Actions

e2e_ppo_trainer_megatron_vllm_2

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
e2e_ppo_trainer_megatron_vllm_2.yml
3,120 workflow runs
3,120 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

[BREAKING][reward] refactor: deprecate batch reward manager e2e_ppo_trainer_megatron_vllm_2 #4130: Pull request #5237 synchronize by yyDing1
2h 27m 45s yyDing1:rm_rm
yyDing1:rm_rm
2h 27m 45s
[reward] fix: preserve input non_tensor_batch in AgentLoopManager whe… e2e_ppo_trainer_megatron_vllm_2 #4129: Commit 28688e7 pushed by yyDing1
2h 26m 3s main
main
2h 26m 3s
[docker] feat: update stable image to vllm==0.15.1 e2e_ppo_trainer_megatron_vllm_2 #4127: Pull request #5221 synchronize by Begunner
27m 58s Begunner:vllm015
Begunner:vllm015
27m 58s
[algo] feat: reduce routed expert padding via NestedTensor and uint8 dtype e2e_ppo_trainer_megatron_vllm_2 #4125: Pull request #5240 opened by xhx1022
2m 30s xhx1022:fixed_r3
xhx1022:fixed_r3
2m 30s
[fsdp, megatron] feat: refactor fully-async training to support multiple checkpoint engine backends e2e_ppo_trainer_megatron_vllm_2 #4122: Pull request #5029 synchronize by Shangwei-Li
20m 9s Shangwei-Li:ckpt_engine
Shangwei-Li:ckpt_engine
20m 9s
[vllm] fix: run post-load weight processing once after async IPC sync e2e_ppo_trainer_megatron_vllm_2 #4121: Pull request #5235 synchronize by zjchenn
32m 49s zjchenn:fix/add-post-process-call
zjchenn:fix/add-post-process-call
32m 49s
[algo] feat: support router replay in MegatronEngine e2e_ppo_trainer_megatron_vllm_2 #4120: Pull request #5219 synchronize by xhx1022
1h 3m 15s xhx1022:r2_engine
xhx1022:r2_engine
1h 3m 15s
[vllm] fix: apply moe weight loader patch for standard wight loading … e2e_ppo_trainer_megatron_vllm_2 #4119: Commit a5087cb pushed by wuxibin89
1h 12m 34s main
main
1h 12m 34s
[BREAKING][reward] refactor: deprecate batch reward manager e2e_ppo_trainer_megatron_vllm_2 #4118: Pull request #5237 synchronize by yyDing1
27m 5s yyDing1:rm_rm
yyDing1:rm_rm
27m 5s
[BREAKING][reward] refactor: deprecate batch reward manager e2e_ppo_trainer_megatron_vllm_2 #4117: Pull request #5237 synchronize by yyDing1
30m 19s yyDing1:rm_rm
yyDing1:rm_rm
30m 19s
[BREAKING][reward] refactor: deprecate batch reward manager e2e_ppo_trainer_megatron_vllm_2 #4116: Pull request #5237 opened by yyDing1
43m 20s yyDing1:rm_rm
yyDing1:rm_rm
43m 20s
[algo] feat: support router replay in MegatronEngine e2e_ppo_trainer_megatron_vllm_2 #4115: Pull request #5219 synchronize by xhx1022
28m 15s xhx1022:r2_engine
xhx1022:r2_engine
28m 15s
[fsdp] feat: add script for qwen3next training on npu platform e2e_ppo_trainer_megatron_vllm_2 #4114: Pull request #5236 synchronize by zjchenn
42m 57s zjchenn:feature/qwen3next-npu
zjchenn:feature/qwen3next-npu
42m 57s
[fsdp] feat: add script for qwen3next training on npu platform e2e_ppo_trainer_megatron_vllm_2 #4113: Pull request #5236 opened by zjchenn
Action required zjchenn:feature/qwen3next-npu
zjchenn:feature/qwen3next-npu
Action required
[algo] feat: support router replay in MegatronEngine e2e_ppo_trainer_megatron_vllm_2 #4112: Pull request #5219 synchronize by xhx1022
40m 38s xhx1022:r2_engine
xhx1022:r2_engine
40m 38s
[vllm] fix: apply moe weight loader patch for standard wight loading e2e_ppo_trainer_megatron_vllm_2 #4111: Pull request #5234 synchronize by zjchenn
46m 7s zjchenn:fix/add-moe-weight-loader
zjchenn:fix/add-moe-weight-loader
46m 7s
[vllm] fix: run post-load weight processing once after async IPC sync e2e_ppo_trainer_megatron_vllm_2 #4110: Pull request #5235 opened by zjchenn
53m 37s zjchenn:fix/add-post-process-call
zjchenn:fix/add-post-process-call
53m 37s
[vllm] fix: apply moe weight loader patch for standard wight loading e2e_ppo_trainer_megatron_vllm_2 #4109: Pull request #5234 opened by zjchenn
Action required zjchenn:fix/add-moe-weight-loader
zjchenn:fix/add-moe-weight-loader
Action required
[rollout,vllm] Fix DP args and local_rank for Ray NOSET_VISIBLE_DEVICES e2e_ppo_trainer_megatron_vllm_2 #4108: Pull request #5233 synchronize by JohnConnor123
Action required JohnConnor123:bugfix/vllm_dp_rank_and_args
JohnConnor123:bugfix/vllm_dp_rank_and_args
Action required
[rollout,vllm] Fix DP args and local_rank for Ray NOSET_VISIBLE_DEVICES e2e_ppo_trainer_megatron_vllm_2 #4107: Pull request #5233 opened by JohnConnor123
Action required JohnConnor123:bugfix/vllm_dp_rank_and_args
JohnConnor123:bugfix/vllm_dp_rank_and_args
Action required
fix(agent_loop): handle batch size smaller than num_workers e2e_ppo_trainer_megatron_vllm_2 #4106: Pull request #5231 opened by aoshen524
40m 53s aoshen524:fix/agent-loop-batch-size-edge-case
aoshen524:fix/agent-loop-batch-size-edge-case
40m 53s
feat(vision): add Vision DP for parallel ViT computation across SP ranks e2e_ppo_trainer_megatron_vllm_2 #4105: Pull request #5230 synchronize by aoshen524
Action required aoshen524:feat/vision-dp-parallel
aoshen524:feat/vision-dp-parallel
Action required
[model] fix: slice rolled_labels for Ulysses Sequence Parallel in all model backends e2e_ppo_trainer_megatron_vllm_2 #4104: Pull request #5227 synchronize by aoshen524
Action required aoshen524:fix/qwen2_vl_rolled_labels_sp_slice
aoshen524:fix/qwen2_vl_rolled_labels_sp_slice
Action required
[trainer] feat: add per-round logprob mismatch metrics for multi-turn training e2e_ppo_trainer_megatron_vllm_2 #4103: Pull request #5229 synchronize by aoshen524
Action required aoshen524:feat/per-round-logprob-mismatch-metrics
aoshen524:feat/per-round-logprob-mismatch-metrics
Action required
[trainer] feat: add per-round logprob mismatch metrics for multi-turn training e2e_ppo_trainer_megatron_vllm_2 #4102: Pull request #5229 synchronize by aoshen524
Action required aoshen524:feat/per-round-logprob-mismatch-metrics
aoshen524:feat/per-round-logprob-mismatch-metrics
Action required