Jixiang Hong

Results 4 comments of Jixiang Hong

Code for finetuning Janus Pro and Janus Flow

> Our multimodal alignment repository, [Align Anything](https://github.com/PKU-Alignment/align-anything), supports the fine-tuning of the Janus model series. We invite you to try it out! Does `Align Anything` support fine-tuning image generation? Thanks.

Add an option to disable flash attention for older GPU's

A simple way is to modify the `config.json`

[Usage]: why speculate decoding is slower than normal decoding？

same problem when using qwen2.5 7b with 0.5b as the draft model and I inspect that the avg draft accept rate is over 0.7

[Bug] vllm deploy InternVL3_5-241B-A28B error

> > > > same err here > > > > > > > > > 官方一直都没回，估计是还没支持这个模型，我试了下InternVL3_5-1B是可以部署的 > > > > > > 似乎MoE都不太行，另外 5B ViT 的那个38B是不是也有点小问题？加载不了 > > 我这边38B可以跑...