leideng/QCFuse / srt /layers /modelopt_utils.py
leideng's picture
download
raw
335 Bytes
"""
ModelOpt related constants
"""
QUANT_CFG_CHOICES = {
"fp8": "FP8_DEFAULT_CFG",
"int4_awq": "INT4_AWQ_CFG", # TODO: add support for int4_awq
"w4a8_awq": "W4A8_AWQ_BETA_CFG", # TODO: add support for w4a8_awq
"nvfp4": "NVFP4_DEFAULT_CFG",
"nvfp4_awq": "NVFP4_AWQ_LITE_CFG", # TODO: add support for nvfp4_awq
}

Xet Storage Details

Size:
335 Bytes
·
Xet hash:
81987bc3f39b4c5b23112c6b4b3028b726b83ace81f0d9b01a6f8ccc8e082aaf

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.