Add W2A16 AutoRound quantized: quantization_config.json
Browse files
W2A16-AutoRound/quantization_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bits": 2,
|
| 3 |
+
"group_size": 64,
|
| 4 |
+
"sym": true,
|
| 5 |
+
"data_type": "int",
|
| 6 |
+
"lr": 0.002,
|
| 7 |
+
"minmax_lr": 0.002,
|
| 8 |
+
"iters": 1000,
|
| 9 |
+
"nsamples": 1024,
|
| 10 |
+
"autoround_version": "0.9.5",
|
| 11 |
+
"quant_method": "auto-round",
|
| 12 |
+
"packing_format": "auto_round:auto_gptq"
|
| 13 |
+
}
|