mehmet.erdogan
commited on
Commit
·
6e18f51
1
Parent(s):
1cf7547
- onnx/config.json +1 -1
- onnx/model.onnx +2 -2
- onnx/ort_config.json +37 -0
onnx/config.json
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
{
|
2 |
"_attn_implementation_autoset": true,
|
3 |
-
"_name_or_path": "
|
4 |
"architectures": [
|
5 |
"HF_ColBERT"
|
6 |
],
|
|
|
1 |
{
|
2 |
"_attn_implementation_autoset": true,
|
3 |
+
"_name_or_path": "onnx",
|
4 |
"architectures": [
|
5 |
"HF_ColBERT"
|
6 |
],
|
onnx/model.onnx
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:90a45d3c619f36126f207f319f6242de1bb76372ebc37cf7228b9e9044550d1c
|
3 |
+
size 440234982
|
onnx/ort_config.json
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"one_external_file": true,
|
3 |
+
"opset": null,
|
4 |
+
"optimization": {
|
5 |
+
"disable_attention": null,
|
6 |
+
"disable_attention_fusion": false,
|
7 |
+
"disable_bias_gelu": null,
|
8 |
+
"disable_bias_gelu_fusion": false,
|
9 |
+
"disable_bias_skip_layer_norm": null,
|
10 |
+
"disable_bias_skip_layer_norm_fusion": false,
|
11 |
+
"disable_embed_layer_norm": true,
|
12 |
+
"disable_embed_layer_norm_fusion": true,
|
13 |
+
"disable_gelu": null,
|
14 |
+
"disable_gelu_fusion": false,
|
15 |
+
"disable_group_norm_fusion": true,
|
16 |
+
"disable_layer_norm": null,
|
17 |
+
"disable_layer_norm_fusion": false,
|
18 |
+
"disable_packed_kv": true,
|
19 |
+
"disable_rotary_embeddings": false,
|
20 |
+
"disable_shape_inference": true,
|
21 |
+
"disable_skip_layer_norm": null,
|
22 |
+
"disable_skip_layer_norm_fusion": false,
|
23 |
+
"enable_gelu_approximation": false,
|
24 |
+
"enable_gemm_fast_gelu_fusion": false,
|
25 |
+
"enable_transformers_specific_optimizations": true,
|
26 |
+
"fp16": false,
|
27 |
+
"no_attention_mask": false,
|
28 |
+
"optimization_level": 2,
|
29 |
+
"optimize_for_gpu": false,
|
30 |
+
"optimize_with_onnxruntime_only": null,
|
31 |
+
"use_mask_index": false,
|
32 |
+
"use_multi_head_attention": false,
|
33 |
+
"use_raw_attention_mask": false
|
34 |
+
},
|
35 |
+
"quantization": {},
|
36 |
+
"use_external_data_format": false
|
37 |
+
}
|