Upload quantize_config.json with huggingface_hub
Browse files- quantize_config.json +34 -0
quantize_config.json
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"per_channel": true,
|
3 |
+
"reduce_range": true,
|
4 |
+
"per_model_config": {
|
5 |
+
"model": {
|
6 |
+
"op_types": [
|
7 |
+
"Erf",
|
8 |
+
"Constant",
|
9 |
+
"Shape",
|
10 |
+
"Cast",
|
11 |
+
"Add",
|
12 |
+
"Unsqueeze",
|
13 |
+
"Equal",
|
14 |
+
"ConstantOfShape",
|
15 |
+
"Sub",
|
16 |
+
"MatMul",
|
17 |
+
"Gather",
|
18 |
+
"Where",
|
19 |
+
"ReduceMean",
|
20 |
+
"Mul",
|
21 |
+
"Div",
|
22 |
+
"Softmax",
|
23 |
+
"Sqrt",
|
24 |
+
"Slice",
|
25 |
+
"Concat",
|
26 |
+
"Reshape",
|
27 |
+
"Pow",
|
28 |
+
"Transpose",
|
29 |
+
"Expand"
|
30 |
+
],
|
31 |
+
"weight_type": "QInt8"
|
32 |
+
}
|
33 |
+
}
|
34 |
+
}
|