Pinned Loading

1. nsa-from-scratchnsa-from-scratchPublic

   From-scratch reimplementation of DeepSeek's Native Sparse Attention (arXiv:2502.11089) in Triton + CUDA Hopper WGMMA. 7.07x faster than FlashAttention-3 at 64k context. Five-model training fleet, p…

   Python 10

2. tinycompresstinycompressPublic

   Implemented and benchmarked LLM inference compression: int4/int8 quantization, GPTQ-like calibration, int8 KV cache, pruning, distillation, speculative decoding, torch.compile, and ONNX. Every numb…

   Python

3. pytorchpytorchPublic

   Forked from pytorch/pytorch

   Tensors and Dynamic neural networks in Python with strong GPU acceleration

   Python

4. transformerstransformersPublic

   Forked from huggingface/transformers

   🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

   Python