Popular repositories Loading

1. gptq gptq Public

   Code for the ICLR 2023 paper "GPTQ: Accurate Post-training Quantization of Generative Pretrained Transformers".

   Python 2.3k 202

2. marlin marlin Public

   FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.

   Python 1.1k 90

3. sparsegpt sparsegpt Public

   Code for the ICML 2023 paper "SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot".

   Python 889 124

4. PanzaMail PanzaMail Public

   Python 304 23

5. qmoe qmoe Public

   Code for the paper "QMoE: Practical Sub-1-Bit Compression of Trillion-Parameter Models".

   Python 278 25

6. llmq llmq Public

   Quantized LLM training in pure CUDA/C++.

   C++ 250 15