MIT HAN Lab — original authors of the AWQ W4A16 GEMM kernel.

Primary authors of the AWQ paper + reference implementation:

  Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Xingyu Dang, Song Han.

The paper:

  AWQ: Activation-aware Weight Quantization for On-Device LLM
  Compression and Acceleration
  Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen,
  Wei-Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, Song Han
  MLSys 2024, arXiv:2306.00978

  https://arxiv.org/abs/2306.00978

Note: AWQ's dequantize kernel is itself derived from NVIDIA's
FasterTransformer (`cutlass_extensions/interleaved_numeric_conversion.h`)
per the per-file header comment in `src/dequantize.cuh`.

Upstream repository: https://github.com/mit-han-lab/llm-awq
