clef-flash-FP8 / clef_quant_recipe.json
aleksblago's picture
Add clef-flash-FP8
7ae79af verified
Raw History Blame
833 Bytes
{
"scheme": "FP8_DYNAMIC",
"format": "fp8",
"release": "/hf/hub/models--Cloudflare--clef-flash/snapshots/8b2e5fd17c09fd49bc2880b805d5323ec7ff4ff3",
"ignore": [
"lm_head",
"re:.*visual.*",
"re:.*linear_attn.in_proj_a$",
"re:.*linear_attn.in_proj_b$"
],
"extra_fused_layer_names": [
[
"in_proj_qkv",
"in_proj_z"
]
],
"calibration": {
"used": false
},
"head_sha256": {
"joint_head.safetensors": "19cdcec8c81dc9212be320fff47462ab342fbc1278be4368fb3da71241cf5ba0",
"joint_head_config.json": "77efe959a38b5b17b241543e129e695f3c77465ece55a25985279bd8176279a0"
},
"versions": {
"llmcompressor": "0.14.0",
"compressed_tensors": "0.19.0",
"transformers": "5.16.1",
"torch": "2.14.0a0+b2c75dd062.nv26.09"
},
"pipeline": "datafree",
"seconds": 117.8
}