metadata
license: apache-2.0
base_model:
- nvidia/OpenReasoning-Nemotron-7B
datasets:
- mit-han-lab/pile-val-backup
OpenReasoning-Nemotron-7B-AWQ
Method
Quantised using vllm-project/llm-compressor and the following configs:
recipe = [
AWQModifier(ignore=["lm_head"], scheme="W4A16_ASYM", targets=["Linear"]),
]