Search results

[RFC][Feat Proposal][New Contributor] Integrating F-GRPO

Status: Open.

#1565 In THUDM/slime;

· Felix-Zhenghao opened on Feb 10, 2026

Training sglang rollout timeout

Status: Open.

#1562 In THUDM/slime;

· p1k0pan opened on Feb 9, 2026

Resume with --load restores Megatron ckpt but train.py rollout loop still starts from 0

Status: Open.

#1558 In THUDM/slime;

· p1k0pan opened on Feb 7, 2026

MoE training: is there an API/hook to monitor expert activations?

Status: Open.

#1555 In THUDM/slime;

· tiesanguaixia opened on Feb 6, 2026

全异步返回数据是不是有bug? RolloutFnTrainOutput 格式？

Status: Open.

#1553 In THUDM/slime;

· HanelYuki opened on Feb 6, 2026

[feature request] Support multiple rollout worker for CPU-heavy custom <code>generate</code> function

Status: Open.

#1552 In THUDM/slime;

· leoleoasd opened on Feb 5, 2026

Speculative Decoding Standalone mode load draft model

Status: Open.

#1548 In THUDM/slime;

· ruiqiRichard opened on Feb 5, 2026

[BUG] LR scheduler double-counted when resuming from checkpoint

Status: Open.

#1546 In THUDM/slime;

· Surya-Gunukula opened on Feb 5, 2026

Do you have plans to support MiMo-V2-Flash?

Status: Open.

#1544 In THUDM/slime;

· Eisenhower opened on Feb 4, 2026

【模型支持】关于对Qwen3-Coder-Next模型的支持

Status: Open.

#1543 In THUDM/slime;

· Mrkkew opened on Feb 4, 2026

run-qwen3-30B-A3B-int4.sh megatron在加载模型时超时

Status: Open.

#1540 In THUDM/slime;

· 191220042 opened on Feb 3, 2026

[Usage] does update_weights_from_tensor save weights data ?

Status: Open.

#1536 In THUDM/slime;

· xgwang opened on Feb 3, 2026