Search results

Qwen implementation has no direct-align

Status: Open.

#42 In Tencent-Hunyuan/SRPO;

· avramdj opened on Feb 28, 2026

Question about the Direct Align part

Status: Open.

#40 In Tencent-Hunyuan/SRPO;

· yangzhou24 opened on Dec 6, 2025

Why make all the device generate from the same latent (noise)?

Status: Open.

#39 In Tencent-Hunyuan/SRPO;

· mchong6 opened on Dec 2, 2025

Flux.2已经发布，是否会搁置基于Flux.1的量化版而直接从Flux.2开始重新训练？

Status: Open.

#38 In Tencent-Hunyuan/SRPO;

· Dengyinsong opened on Nov 27, 2025

Training problem about QWEN-srpo

Status: Open.

#34 In Tencent-Hunyuan/SRPO;

· tianqyun111 opened on Oct 30, 2025

support flux kontext srpo

Status: Open.

#33 In Tencent-Hunyuan/SRPO;

· akk-123 opened on Oct 21, 2025

have u compared it with flow-grpo?

Status: Open.

#32 In Tencent-Hunyuan/SRPO;

· MengHao666 opened on Oct 16, 2025

Unstable behavior when changing batch size (train_batch_size, sp_size, train_sp_batch_size)

Status: Open.

#31 In Tencent-Hunyuan/SRPO;

· Borntowarn opened on Oct 15, 2025

是否还有发布qwen-image的SRPO版本计划？

Status: Open.

#30 In Tencent-Hunyuan/SRPO;

· TaylorChaos opened on Oct 15, 2025

Support LoRA training

Status: Open.

#29 In Tencent-Hunyuan/SRPO;

· yogurt7771 opened on Oct 11, 2025

Does it support inpainting model?

Status: Open.

#24 In Tencent-Hunyuan/SRPO;

· hdjsjyl opened on Sep 28, 2025

听说star到1000, 发布qwen版本的srpo?

Status: Open.

#23 In Tencent-Hunyuan/SRPO;

· tool2d opened on Sep 24, 2025