spark-servitor / quant_metadata.json
wxsys's picture
config: add quant_metadata.json
eb69e29 verified
Raw History Blame Contribute Delete
278 Bytes
{
"quant_algorithm": "AWQ_INT4",
"group_size": 128,
"total_tensors": 444,
"outlier_preserved_tensors": 219,
"quantized_gemm_tensors": 225,
"compression_ratio": 2.34,
"format": "INT4",
"binary_file": "spark-servitor-awq_int4.servitor",
"file_size_mb": 1608.27
}