-
Notifications
You must be signed in to change notification settings - Fork 4.2k
Pull requests: verl-project/verl
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix/support-checkpointsave_cpuoffload
#7077
opened Jul 16, 2026 by
Lzw-lukeskywalker
Loading…
8 tasks
[trainer] fix: exclude the never-routed final response token from rollout MoE load-balance metrics
#7075
opened Jul 16, 2026 by
zsnoob
Loading…
5 of 6 tasks
[doc] fix: repair broken Ascend tutorial links and shell filename
#7074
opened Jul 16, 2026 by
chengminhua
Contributor
Loading…
8 tasks
[rollout, trainer] feat: add lightweight GPU stall diagnostics
#7073
opened Jul 16, 2026 by
chengcuiping
Contributor
Loading…
[rollout, vllm] fix: strip .base_layer suffix from param names in vLLM weight update
#7070
opened Jul 16, 2026 by
Secbone
Loading…
5 of 8 tasks
[trainer] feat: early stopping via early_stop_patience config
#7069
opened Jul 16, 2026 by
Secbone
Loading…
3 of 8 tasks
[ckpt, model] fix: prevent key corruption during HuggingFace export
#7068
opened Jul 16, 2026 by
rongkunxue
Contributor
Loading…
2 tasks done
[env] fix: Pin the PyArrow version to <=24.0.0 to avoid issues with version 25.0.0.
#7065
opened Jul 16, 2026 by
ruanhao566
Contributor
Loading…
8 tasks
[reward] feat: expose valid_response_length and max_resp_len in extra_info
#7063
opened Jul 16, 2026 by
Secbone
Loading…
3 of 8 tasks
[utils] fix: remove CPU offloading in collect_lora_params
#7062
opened Jul 16, 2026 by
Secbone
Loading…
3 of 8 tasks
[ckpt] feat: add save_lora_only checkpoint support
#7061
opened Jul 16, 2026 by
Secbone
Loading…
5 of 8 tasks
[RFC] [ROCm] feat: bring-up deepseek v4 grpo on AMD GPUs
#7050
opened Jul 15, 2026 by
PeterYang12
Contributor
Loading…
4 of 8 tasks
[trainer, perf] fix: include standalone rollout GPUs in throughput denominator for separate async
#7049
opened Jul 15, 2026 by
mikequan0425
Contributor
Loading…
7 of 8 tasks
[tool] fix: remove circular fallback in BaseTool.__init__
#7046
opened Jul 15, 2026 by
cyyueyang
Loading…
2 tasks done
[checkpoint_engine] fail fast on hung first NCCL group init (#6967)
#7045
opened Jul 15, 2026 by
aryanyadav0402
Loading…
[rollout] fix: handle malformed Qwen3 XML tool calls
#7044
opened Jul 15, 2026 by
zzzzzzzxh
Loading…
[data] feat: auto-download TinyLLaVA-Video-R1 dataset when --data_dir is not given
#7043
opened Jul 14, 2026 by
ArdalanM
Loading…
[misc] handle flash attention 3.0 with recent transformers version (>= 4.56.0)
#7033
opened Jul 13, 2026 by
ArdalanM
Loading…
[reward] feat: support deterministic reward for user-defined generative RM paths
#7027
opened Jul 13, 2026 by
KaisennHu
Contributor
Loading…
[algo, doc] feat: add UP-GRPO unbounded positive asymmetric policy loss
#7022
opened Jul 13, 2026 by
a-F1
Loading…
5 of 8 tasks
[trainer] feat: queue TransferQueue agent batches
#7021
opened Jul 12, 2026 by
luzy99
Loading…
8 tasks done
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.