README.md

metadata

base_model:
  - meta-llama/Llama-2-70b-hf
base_model_relation: quantized
license: llama2

Model Card

• Base model: meta-llama/Llama-2-70b-hf
• Quantization method: LNQ
• Target bit-width: 4
• Backend kernel: Any-Precision-LLM kernel (ap-gemv)
• Calibration data: RedPajama (1024 sentences / 4096 tokens)
• Calibration objective: Next-token prediction

How to run

• Follow the instruction in https://github.com/snu-mllab/GuidedQuant.

References

• Model Paper