Skip to content

Actions: verl-project/verl

Actions

e2e_one_step_off_policy_ascend

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
e2e_one_step_off_policy_ascend.yml
540 workflow runs
540 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

[BREAKING][rollout,cfg] refactor: get rid of actor_rollout_ref config from rollout e2e_one_step_off_policy_ascend #554: Pull request #5418 synchronize by wuxibin89
1h 0m 36s wuxibin89:wuxibin/refactor_config
wuxibin89:wuxibin/refactor_config
1h 0m 36s
[tool] refactor: remove tool schema plumbing from SingleTurnAgentLoop… e2e_one_step_off_policy_ascend #553: Commit 48b367b pushed by wuxibin89
57m 17s main
main
57m 17s
[megatron, vllm] feat: NVFP4 (W4A16) QAT training support via ModelOpt e2e_one_step_off_policy_ascend #552: Pull request #5254 synchronize by jQizhang
Action required jQizhang:nvfp4_cleanup
jQizhang:nvfp4_cleanup
Action required
[perf, trtllm] feat: Add Nsight support for rollout server mode (trtllm) e2e_one_step_off_policy_ascend #551: Pull request #5391 synchronize by davidmlw
1h 1m 23s joyang-nv:liweim/nsys
joyang-nv:liweim/nsys
1h 1m 23s
[trainer] feat: Add Nemo-Automodel as alternative training engine e2e_one_step_off_policy_ascend #550: Pull request #5407 synchronize by HuiyingLi
1h 1m 23s HuiyingLi:add_automodel_sft_backend
HuiyingLi:add_automodel_sft_backend
1h 1m 23s
[tool, rollout] refactor: cache tool list initialization with MCP timeout e2e_one_step_off_policy_ascend #549: Pull request #5426 opened by denismegerle
Action required denismegerle:public/tool-list-caching
denismegerle:public/tool-list-caching
Action required
[trainer] feat: Add Nemo-Automodel as alternative training engine e2e_one_step_off_policy_ascend #548: Pull request #5407 synchronize by HuiyingLi
Action required HuiyingLi:add_automodel_sft_backend
HuiyingLi:add_automodel_sft_backend
Action required
[trainer] feat: Add Nemo-Automodel as alternative training engine e2e_one_step_off_policy_ascend #547: Pull request #5407 synchronize by HuiyingLi
Action required HuiyingLi:add_automodel_sft_backend
HuiyingLi:add_automodel_sft_backend
Action required
[WIP][algo] Migrate and implement the GDPO algorithm into the existing framework. e2e_one_step_off_policy_ascend #546: Pull request #5422 synchronize by Rhetee
Action required Rhetee:GDPO
Rhetee:GDPO
Action required
[tool] refactor: remove tool schema plumbing from SingleTurnAgentLoop e2e_one_step_off_policy_ascend #545: Pull request #5425 opened by denismegerle
15m 45s denismegerle:public/single-turn-disable-tools
denismegerle:public/single-turn-disable-tools
15m 45s
[trainer] feat: add support for the GDPO algorithm e2e_one_step_off_policy_ascend #544: Pull request #5409 synchronize by yue-zeng-yue
Action required yue-zeng-yue:feat-gdpo
yue-zeng-yue:feat-gdpo
Action required
[rollout, data] fix: honor train_max_samples/val_max_samples in fully… e2e_one_step_off_policy_ascend #543: Commit c3e3970 pushed by ArronHZG
1h 28m 48s main
main
1h 28m 48s
[tool] fix: handle empty image inputs in ToolAgentLoop (#5420) e2e_one_step_off_policy_ascend #542: Commit 9dd447e pushed by wuxibin89
1h 0m 46s main
main
1h 0m 46s
[trainer] feat: add support for the GDPO algorithm e2e_one_step_off_policy_ascend #541: Pull request #5409 synchronize by yue-zeng-yue
Action required yue-zeng-yue:feat-gdpo
yue-zeng-yue:feat-gdpo
Action required
[BREAKING][rollout,cfg] refactor: get rid of actor_rollout_ref config from rollout e2e_one_step_off_policy_ascend #540: Pull request #5418 synchronize by wuxibin89
1h 0m 48s wuxibin89:wuxibin/refactor_config
wuxibin89:wuxibin/refactor_config
1h 0m 48s
[misc] feat: Add code for data grouping in no-padding scenario e2e_one_step_off_policy_ascend #539: Pull request #5424 synchronize by Kite0011
17m 7s Kite0011:add_rm_support
Kite0011:add_rm_support
17m 7s
[rollout] feat: Fix partial load problem, Add vlm support for trtllm rollout e2e_one_step_off_policy_ascend #536: Pull request #5149 synchronize by SchumiDing
Action required SchumiDing:vlm_trtllm_support
SchumiDing:vlm_trtllm_support
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_ascend #535: Pull request #4263 synchronize by khazic
Action required khazic:main
khazic:main
Action required
[rollout] feat: Fix partial load problem, Add vlm support for trtllm rollout e2e_one_step_off_policy_ascend #534: Pull request #5149 synchronize by SchumiDing
Action required SchumiDing:vlm_trtllm_support
SchumiDing:vlm_trtllm_support
Action required
[fsdp] fix: make GDN work correctly when using ulysses and varlen sequence e2e_one_step_off_policy_ascend #533: Pull request #5346 synchronize by nono-Sang
Action required nono-Sang:gdn_fix
nono-Sang:gdn_fix
Action required
[BREAKING][megatron] feat: support Megatron-FSDP as a new training backend e2e_one_step_off_policy_ascend #532: Pull request #5423 opened by conver334
Action required conver334:megatron-fsdp
conver334:megatron-fsdp
Action required
[WIP][algo] Migrate and implement the GDPO algorithm into the existing framework. e2e_one_step_off_policy_ascend #531: Pull request #5422 synchronize by Rhetee
Action required Rhetee:GDPO
Rhetee:GDPO
Action required
[megatron, vllm] feat: NVFP4 (W4A16) QAT training support via ModelOpt e2e_one_step_off_policy_ascend #530: Pull request #5254 synchronize by jQizhang
Action required jQizhang:nvfp4_cleanup
jQizhang:nvfp4_cleanup
Action required
[WIP][algo] Migrate and implement the GDPO algorithm into the existing framework. e2e_one_step_off_policy_ascend #529: Pull request #5422 opened by Rhetee
Action required Rhetee:GDPO
Rhetee:GDPO
Action required
[tool] feature: scheduling analysis based on profiling data for torch profiler e2e_one_step_off_policy_ascend #528: Pull request #5367 synchronize by Rhetee
Action required Rhetee:main
Rhetee:main
Action required