Search results

Google TPU support in verl with Ray

Status: Open.

#5192 In verl-project/verl;

· andrewsykim opened on Feb 3, 2026

[BUG] 2月3日的最新代码跑recipe/dapo/run_dapo_qwen2.5_7b_npu.sh 失败

bugSomething isn't workingSomething isn't working

Status: Open.

#5188 In verl-project/verl;

· daikang6 opened on Feb 3, 2026

[BUG] Mean reward and acc increasing steadily on Qwen-VL-MoE-32B-A3B with vanilla mbridge setup but not with Megatron-Bridge on example scripts (8×H200)

bugSomething isn't workingSomething isn't working

Status: Open.

#5187 In verl-project/verl;

· sensimintel-master opened on Feb 3, 2026

Why GPU OOM when save ckpt？

Status: Open.

#5183 In verl-project/verl;

· qbei-i opened on Feb 3, 2026

[Feature] Reproducibility in training actor

Status: Open.

#5180 In verl-project/verl;

· victordion opened on Feb 2, 2026

[RFC] Support openYuanrong as an optional distributed framework for veRL

Status: Open.

#5169 In verl-project/verl;

· Hellowujie2009 opened on Feb 2, 2026

[FullAsync] 运行官方recipe/fully_async_policy/shell/geo3k_qwen25vl_7b_megatron_4_4.sh 失败

bugSomething isn't workingSomething isn't working

Status: Open.

#5165 In verl-project/verl;

· qq1243196045 opened on Feb 2, 2026

[RFC] Refactor Reward Components

Status: Open.

#5158 In verl-project/verl;

· yyDing1 opened on Feb 1, 2026

[Feature] PARL from Kimi K2.5

Status: Open.

#5140 In verl-project/verl;

· axion66 opened on Jan 31, 2026

Feature Request: Support for Joint Training with Visual Frameworks + RL Gradient Backpropagation

Status: Open.

#5129 In verl-project/verl;

· t2nc5vbrsc-spec opened on Jan 30, 2026

Host memory unecessary usage in FSDP rank 0

bugSomething isn't workingSomething isn't working

Status: Open.

#5125 In verl-project/verl;

· flishwang opened on Jan 30, 2026

0.7.0 npu安装教程配套版本有问题

bugSomething isn't workingSomething isn't working

Status: Open.

#5123 In verl-project/verl;

· ziliangok opened on Jan 30, 2026