Search results

[roadmap] low precision support

roadmap

Status: Open.

#5972 In verl-project/verl;

· sophiayyya opened on Apr 11, 2026

Aarch64 GB200 container perf regression from vllm 0.17 to vllm 0.18

bugSomething isn't workingSomething isn't working

Status: Open.

#5964 In verl-project/verl;

· kaixih opened on Apr 10, 2026

[Bug] Training fails with vllm 0.13.0 due to removed parameter: disable_mm_preprocessor_cache

bugSomething isn't workingSomething isn't working

Status: Open.

#5959 In verl-project/verl;

· Silas-11 opened on Apr 10, 2026

[NPU] Ulysses Sequence Parallelism 导致 logits_rmpad 与 temperature_rmpad 维度不匹配 (Ascend NPU)

bugSomething isn't workingSomething isn't working

Status: Open.

#5957 In verl-project/verl;

· singing4you opened on Apr 10, 2026

指定使用gpu无效

bugSomething isn't workingSomething isn't working

Status: Open.

#5954 In verl-project/verl;

· cqray1990 opened on Apr 10, 2026

训练Qwen3.5-35B-A3B出现熵坍塌的问题

Status: Open.

#5953 In verl-project/verl;

· panjicaiz opened on Apr 10, 2026

[Bug] AssertionError: Weight too large to fit in bucket during update_weights in vllm_rollout (Qwen3-VL-8B)

bugSomething isn't workingSomething isn't working

Status: Open.

#5952 In verl-project/verl;

· Silas-11 opened on Apr 10, 2026

Feature Request: Adding GenRM to Fully Async

Status: Open.

#5949 In verl-project/verl;

· shayansadeghieh opened on Apr 9, 2026

Qwen3 GRPO Rollout 乱码

bugSomething isn't workingSomething isn't working

Status: Open.

#5946 In verl-project/verl;

· wangsh0111 opened on Apr 9, 2026

Qwen3.5 FSDP SFT Bug

bugSomething isn't workingSomething isn't working

Status: Open.

#5944 In verl-project/verl;

· feirz-tech opened on Apr 9, 2026

[BUG] FSDP engine forward_step hardcodes bf16 autocast and ignores configured mixed-precision dtype

Status: Open.

#5932 In verl-project/verl;

· ChangyiYang opened on Apr 8, 2026

DataLoader worker (pid 10855) is killed by signal: Killed.

Status: Open.

#5922 In verl-project/verl;

· pyy-hotyouth opened on Apr 8, 2026