Skip to content

Actions: verl-project/verl

Actions

e2e_one_step_off_policy

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
e2e_one_step_off_policy.yml
4,322 workflow runs
4,322 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

[megatron] feat: enhance model offloading and loading for frozen parameters e2e_one_step_off_policy #5946: Pull request #5412 synchronize by RobotGF
6m 7s RobotGF:fix_lora_offload
RobotGF:fix_lora_offload
6m 7s
[ckpt] feat: implement large tensor slicing in vllm rollout and CheckpointEngine for weight updating e2e_one_step_off_policy #5945: Pull request #5378 synchronize by jianjunzhong
5m 25s jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
5m 25s
[algo] fix: seq mean and default scale factor loss_mask.shape[-1] as in seq-mean-token-sum-norm e2e_one_step_off_policy #5944: Pull request #5417 opened by tongyx361
26m 19s tongyx361:tyx/fix/seq-mean-in-seq-mean-token-sum-norm
tongyx361:tyx/fix/seq-mean-in-seq-mean-token-sum-norm
26m 19s
[tool] feature: scheduling analysis based on profiling data for torch profiler e2e_one_step_off_policy #5943: Pull request #5367 synchronize by Rhetee
Action required Rhetee:main
Rhetee:main
Action required
[perf, trtllm] feat: Add Nsight support for rollout server mode (trtllm) e2e_one_step_off_policy #5942: Pull request #5391 synchronize by davidmlw
13m 34s joyang-nv:liweim/nsys
joyang-nv:liweim/nsys
13m 34s
[Megatron] feat: Support routing replay on NPU with performance and compatibility enhancements e2e_one_step_off_policy #5941: Pull request #5298 synchronize by 755651978
14m 37s 755651978:main-0212
755651978:main-0212
14m 37s
why are there multiple settings for actor_rollout_ref.model.enable_gradient_checkpointing? Is this a deliberate design choice? e2e_one_step_off_policy #5940: Pull request #4263 synchronize by khazic
Action required khazic:main
khazic:main
Action required
[algo] feat: add DPPO with binary TV or binary KL implementation (#5397) e2e_one_step_off_policy #5939: Commit 182383b pushed by tongyx361
1h 15m 45s main
main
1h 15m 45s
[Megatron] feat: Support routing replay on NPU with performance and compatibility enhancements e2e_one_step_off_policy #5938: Pull request #5298 synchronize by 755651978
39m 14s 755651978:main-0212
755651978:main-0212
39m 14s
[Megatron] feat: Support routing replay on NPU with performance and compatibility enhancements e2e_one_step_off_policy #5936: Pull request #5298 synchronize by 755651978
14m 46s 755651978:main-0212
755651978:main-0212
14m 46s
why are there multiple settings for actor_rollout_ref.model.enable_gradient_checkpointing? Is this a deliberate design choice? e2e_one_step_off_policy #5935: Pull request #4263 synchronize by khazic
Action required khazic:main
khazic:main
Action required
[fsdp,algo] feat: Support QAT (NVFP4) in FSDPEngine for the unified engine_workers architecture e2e_one_step_off_policy #5934: Pull request #5411 synchronize by zhangyimi
Action required zhangyimi:qat-core-v2
zhangyimi:qat-core-v2
Action required
[fsdp,algo] feat: Support QAT (NVFP4) in FSDPEngine for the unified engine_workers architecture e2e_one_step_off_policy #5933: Pull request #5411 synchronize by zhangyimi
Action required zhangyimi:qat-core-v2
zhangyimi:qat-core-v2
Action required
[fsdp,algo] feat: Support QAT (NVFP4) in FSDPEngine for the unified engine_workers architecture e2e_one_step_off_policy #5932: Pull request #5411 opened by zhangyimi
Action required zhangyimi:qat-core-v2
zhangyimi:qat-core-v2
Action required
[misc,trainer,rollout] feat: add Prometheus metrics logging to experiment tracking e2e_one_step_off_policy #5930: Pull request #5291 synchronize by guillemgt
Action required guillemgt:guillem.tarrach/upstream-prometheus-metrics
guillemgt:guillem.tarrach/upstream-prometheus-metrics
Action required
[misc,trainer,rollout] feat: add Prometheus metrics logging to experiment tracking e2e_one_step_off_policy #5929: Pull request #5291 synchronize by guillemgt
Action required guillemgt:guillem.tarrach/upstream-prometheus-metrics
guillemgt:guillem.tarrach/upstream-prometheus-metrics
Action required
[trainer] feat: add support for the GDPO algorithm e2e_one_step_off_policy #5928: Pull request #5409 opened by yue-zeng-yue
Action required yue-zeng-yue:feat-gdpo
yue-zeng-yue:feat-gdpo
Action required
[trainer,misc] fix: fix multiple bugs in fully async trainings e2e_one_step_off_policy #5927: Pull request #5373 synchronize by guillemgt
Action required guillemgt:upstream/fix-fully-async-bugs
guillemgt:upstream/fix-fully-async-bugs
Action required
[trainer] feat: Add Nemo-Automodel as alternative training engine e2e_one_step_off_policy #5926: Pull request #5407 opened by HuiyingLi
59m 44s HuiyingLi:add_automodel_sft_backend
HuiyingLi:add_automodel_sft_backend
59m 44s
why are there multiple settings for actor_rollout_ref.model.enable_gradient_checkpointing? Is this a deliberate design choice? e2e_one_step_off_policy #5925: Pull request #4263 synchronize by khazic
Action required khazic:main
khazic:main
Action required
why are there multiple settings for actor_rollout_ref.model.enable_gradient_checkpointing? Is this a deliberate design choice? e2e_one_step_off_policy #5924: Pull request #4263 synchronize by khazic
Action required khazic:main
khazic:main
Action required
[perf, trtllm] feat: Add Nsight support for rollout server mode (trtllm) e2e_one_step_off_policy #5923: Pull request #5391 synchronize by davidmlw
14m 52s joyang-nv:liweim/nsys
joyang-nv:liweim/nsys
14m 52s
[ckpt] feat: implement large tensor slicing in vllm rollout and CheckpointEngine for weight updating e2e_one_step_off_policy #5922: Pull request #5378 synchronize by jianjunzhong
14m 45s jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
14m 45s
[megatron] fix: missing model offload to CPU for forward_only mode e2e_one_step_off_policy #5921: Pull request #5406 opened by xhx1022
14m 58s xhx1022:xhx/offload
xhx1022:xhx/offload
14m 58s
[ckpt] feat: implement large tensor slicing in vllm rollout and CheckpointEngine for weight updating e2e_one_step_off_policy #5920: Pull request #5378 synchronize by jianjunzhong
Action required jianjunzhong:feat/chunked_weight_update
jianjunzhong:feat/chunked_weight_update
Action required