Skip to content

Actions: verl-project/verl

Actions

e2e_ascend

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
e2e_ascend.yml
8,459 workflow runs
8,459 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

[ci] chore: pin version cupy-cuda12x==13.6.0 (#5377) e2e_ascend #11615: Commit 58f38fb pushed by wuxibin89
In progress main
main
In progress
[vllm] feat: implement chunked weight handling in vllm rollout for large tensors e2e_ascend #11614: Pull request #5378 synchronize by jianjunzhong
Action required jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
Action required
[trtllm, rollout] Add support for fp8 refit e2e_ascend #11613: Pull request #5374 synchronize by shuyixiong
49m 58s joyang-nv:fp8_refit
joyang-nv:fp8_refit
49m 58s
[vllm] feat: implement chunked weight handling in vllm rollout for large tensors e2e_ascend #11612: Pull request #5378 synchronize by jianjunzhong
Action required jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
Action required
[vllm] feat: implement chunked weight handling in vllm rollout for large tensors e2e_ascend #11611: Pull request #5378 opened by jianjunzhong
Action required jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
Action required
[fsdp, megatron] feat: refactor fully-async and one-step-off training to support multiple checkpoint engine backends e2e_ascend #11610: Pull request #5029 synchronize by Shangwei-Li
1h 1m 25s Shangwei-Li:ckpt_engine
Shangwei-Li:ckpt_engine
1h 1m 25s
[doc] fix: update recipe link to fix 404 not found e2e_ascend #11609: Pull request #5286 synchronize by tardis-key
33m 4s tardis-key:main
tardis-key:main
33m 4s
[tool] refactor: centralized tool registry and decorator-based registration e2e_ascend #11608: Pull request #5376 opened by ascl1u
Action required ascl1u:roadmap/agentloop
ascl1u:roadmap/agentloop
Action required
[rollout] feat: enable R3 in fully async mode e2e_ascend #11607: Pull request #5344 synchronize by guillemgt
Action required guillemgt:guillemgt/enable-r3-fully-async
guillemgt:guillemgt/enable-r3-fully-async
Action required
[megatron] fix: patch support newer mcore version e2e_ascend #11606: Pull request #5372 synchronize by HollowMan6
37m 6s HollowMan6:mcore_patch
HollowMan6:mcore_patch
37m 6s
[trainer,misc] fix: fix multiple bugs in fully async trainings e2e_ascend #11605: Pull request #5373 synchronize by guillemgt
Action required guillemgt:upstream/fix-fully-async-bugs
guillemgt:upstream/fix-fully-async-bugs
Action required
[megatron] fix: patch support newer mcore version e2e_ascend #11604: Pull request #5372 synchronize by HollowMan6
14m 22s HollowMan6:mcore_patch
HollowMan6:mcore_patch
14m 22s
[trtllm, rollout] Add support for fp8 refit e2e_ascend #11603: Pull request #5374 synchronize by shuyixiong
32m 28s joyang-nv:fp8_refit
joyang-nv:fp8_refit
32m 28s
[config] remove deprecated actor_rollout_ref.rollout.mode config field e2e_ascend #11602: Pull request #5352 synchronize by dhruvnigam93
Action required dhruvnigam93:feat/ENG-4604-remove-rollout-mode-config
dhruvnigam93:feat/ENG-4604-remove-rollout-mode-config
Action required
[trtllm, rollout] Add support for fp8 refit e2e_ascend #11601: Pull request #5374 opened by shuyixiong
32m 8s joyang-nv:fp8_refit
joyang-nv:fp8_refit
32m 8s
[misc] fix: pass torch dtype when init random model (#5370) e2e_ascend #11600: Commit f56c893 pushed by vermouth1992
37m 26s main
main
37m 26s
[trainer,misc] fix: fix multiple bugs in fully async trainings e2e_ascend #11599: Pull request #5373 synchronize by guillemgt
Action required guillemgt:upstream/fix-fully-async-bugs
guillemgt:upstream/fix-fully-async-bugs
Action required
[trainer,misc] fix: fix multiple bugs in fully async trainings e2e_ascend #11598: Pull request #5373 opened by guillemgt
Action required guillemgt:upstream/fix-fully-async-bugs
guillemgt:upstream/fix-fully-async-bugs
Action required
[rollout] fix: Include routed_experts in ToolAgentLoop return value t… e2e_ascend #11597: Commit c6255ae pushed by HollowMan6
32m 12s main
main
32m 12s
[megatron] fix: patch support newer mcore version e2e_ascend #11596: Pull request #5372 opened by HollowMan6
33m 32s HollowMan6:mcore_patch
HollowMan6:mcore_patch
33m 32s
[fsdp,trainer,algo] feat: On-Policy Distillation e2e_ascend #11595: Pull request #4897 synchronize by JacobHelwig
31m 43s JacobHelwig:jhelwig/onPolicyDistillation
JacobHelwig:jhelwig/onPolicyDistillation
31m 43s
[fsdp,trainer,vllm_omni,algo] feat: support FlowGRPO training for QwenImage e2e_ascend #11594: Pull request #5297 synchronize by zhtmike
Action required zhtmike:verl-omni-pr
zhtmike:verl-omni-pr
Action required
[fsdp,trainer,algo] feat: On-Policy Distillation e2e_ascend #11593: Pull request #4897 synchronize by JacobHelwig
8m 10s JacobHelwig:jhelwig/onPolicyDistillation
JacobHelwig:jhelwig/onPolicyDistillation
8m 10s
[megatron] refactor: Refactor policy update failure handling from exception to warning in MegatronPPOActor to support FP16 training e2e_ascend #11592: Pull request #5369 synchronize by mirrorboat
Action required mirrorboat:megatron_fp16_gradient_overflow
mirrorboat:megatron_fp16_gradient_overflow
Action required
[megatron] refactor: Refactor policy update failure handling from exception to warning in MegatronPPOActor to support FP16 training e2e_ascend #11591: Pull request #5369 opened by mirrorboat
Action required mirrorboat:megatron_fp16_gradient_overflow
mirrorboat:megatron_fp16_gradient_overflow
Action required