clef-FP8 / clef_quant_recipe.json
aleksblago's picture
Add clef-FP8
122e4f1 verified
Raw History Blame
827 Bytes
{
"scheme": "FP8_DYNAMIC",
"format": "fp8",
"release": "/hf/hub/models--Cloudflare--clef/snapshots/0b331204bb13fbd2ca93a64df1956f5b55478ce5",
"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": "a010ac04f078e699988e4049cbea5e62c962393f59fec366640b64e8d69a4953",
"joint_head_config.json": "890be585d75b981201eb96a35f98a9967afe37bc8d220cfdea2e72d56507534f"
},
"versions": {
"llmcompressor": "0.14.0",
"compressed_tensors": "0.19.0",
"transformers": "5.16.1",
"torch": "2.14.0a0+b2c75dd062.nv26.09"
},
"pipeline": "datafree",
"seconds": 473.5
}