Upload quant_config.json with huggingface_hub
Browse files- quant_config.json +10 -0
quant_config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"w_bit": 4,
|
| 3 |
+
"group_size": 128,
|
| 4 |
+
"method": "RTN",
|
| 5 |
+
"kernel": "torchao_int4_weight_only (_weight_int4pack_mm)",
|
| 6 |
+
"skipped": [
|
| 7 |
+
"shape_proj (in=16)",
|
| 8 |
+
"lm_head (out=4099)"
|
| 9 |
+
]
|
| 10 |
+
}
|