Sparse LLM Inference on CPU: 75% fewer parameters(huggingface.co)

58 points by freediver 3 years ago | 1 comments

Qwen3.8-27B at 256K on a 24GB RTX PRO 4000 SFF (432 GB/s): 50 tok/s with MTP(piszczek.pl)

31 points by pich 2 months ago | 24 comments

Uncensor any LLM with abliteration(huggingface.co)

450 points by mizzao 2 years ago | 228 comments

Hugging Face community blog post by Maxime Labonne explaining abliteration, a technique to remove refusal behavior from LLMs without retraining.

LeoLM: German-Language LLM Research(laion.ai)

92 points by doubtfuluser 3 years ago | 82 comments

Fine-Tuning LLMs to 1.58bit(huggingface.co)

24 points by galeos 2 years ago | 1 comments

Making Gaussian Splats more smaller(aras-p.info)

82 points by ibobev 3 years ago | 13 comments