Li Xin comments

Repositories
Issues
Comments

Results 4 comments of


                                            Li Xin

How did this job achieve such excellent performance?

So, can we regard this code as a subset of vLLM? Are the speed-ups purely due to stripping out vLLM’s redundant logic (since it only need to handle Qwen)?

[Bug] Wan2.1-14B-I2V VSA acceleration is not satisfactory

Which script are you using? As far as I remember, training DMD + VSA on Wan2.1-14B-I2V-720P isn’t supported at the moment.

[Feature] Thanks for adding VSA support - question about compatibility with larger models

Does it currently support only a fixed resolution, or can finetune_v1_VSA.sh handle arbitrary resolutions such as 1280 × 720?

Release Sparse-vDiT on Hugging Face

have any recent updates?