Search results

[Question]

questionFurther information is requestedFurther information is requested

Status: Open.

#1872 In THUDM/slime;

· BitCalSaul opened on Apr 28, 2026

[Bug] # [Bug] --allgather-cp silently scrambles token order in hf_attention.py CP reshuffle

bugSomething isn't workingSomething isn't working

Status: Open.

#1871 In THUDM/slime;

· sunzx8 opened on Apr 27, 2026

[Question] GLM5----MTP training support

questionFurther information is requestedFurther information is requested

Status: Open.

#1870 In THUDM/slime;

· Hevans123 opened on Apr 27, 2026

[Question] qwen3-vl转化

questionFurther information is requestedFurther information is requested

Status: Open.

#1863 In THUDM/slime;

· showshout opened on Apr 26, 2026

[Question] GLM5模型sglang rollout有概率出现首字token乱码/没有意义

questionFurther information is requestedFurther information is requested

Status: Open.

#1853 In THUDM/slime;

· iamweiliu opened on Apr 23, 2026

Qwen3.5-397B-A17B 全异步 H20（96 Actor，64 Rollout）运行正常，但 Rollout 生成的 Response 不正常

questionFurther information is requestedFurther information is requested

Status: Open.

#1852 In THUDM/slime;

· YaxinFAN1 opened on Apr 23, 2026

[Question] RAM requirement for Qwen3-30B-A3B with 8xH100 with --optimizer-cpu-offload

questionFurther information is requestedFurther information is requested

Status: Open.

#1851 In THUDM/slime;

· Moskenstraumen opened on Apr 22, 2026

[Bug] Wrong inference using docker container's sglang server with Qwen3-VL models

bugSomething isn't workingSomething isn't working

Status: Open.

#1850 In THUDM/slime;

· YWenxi opened on Apr 21, 2026

[Bug] VLM multi-turn rollout: model cannot see images on multi-turns

bugSomething isn't workingSomething isn't working

Status: Open.

#1847 In THUDM/slime;

· TongkunGuan opened on Apr 20, 2026

Ray OOM when saving checkpoint unless both --no-save-optim and --disable-weights-backuper are enabled

questionFurther information is requestedFurther information is requested

Status: Open.

#1845 In THUDM/slime;

· samaritan1998 opened on Apr 20, 2026

[Question] How to add a custom auxiliary SFT loss to the GRPO training loop （Chord）?

questionFurther information is requestedFurther information is requested

Status: Open.

#1843 In THUDM/slime;

· canlin03 opened on Apr 20, 2026

[Bug] GPT-OSS raw converter emits incorrect expert weight format, causing gibberish output with --megatron-to-hf-mode raw

bugSomething isn't workingSomething isn't working

Status: Open.

#1840 In THUDM/slime;

· qqqlm opened on Apr 16, 2026