Hugging Face's logo

fxmarty
/
bugged-quantized-model-DedicatedQDQPair

Text Classification
Transformers
ONNX
distilbert
text-embeddings-inference
Model card Files
xet
 Community
Felix Marty commited on
Commit
9b3b1bc
·
1 Parent(s): 145d3ad

add model

Browse files
Files changed (8) hide show
  1. config.json +33 -0
  2. eval_results.json +3 -0
  3. model_quantized.onnx +3 -0
  4. ort_config.json +191 -0
  5. special_tokens_map.json +7 -0
  6. tokenizer.json +0 -0
  7. tokenizer_config.json +16 -0
  8. vocab.txt +0 -0
config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "distilbert_onnx",
3
+ "activation": "gelu",
4
+ "architectures": [
5
+ "DistilBertForSequenceClassification"
6
+ ],
7
+ "attention_dropout": 0.1,
8
+ "dim": 768,
9
+ "dropout": 0.1,
10
+ "finetuning_task": "sst-2",
11
+ "hidden_dim": 3072,
12
+ "id2label": {
13
+ "0": "NEGATIVE",
14
+ "1": "POSITIVE"
15
+ },
16
+ "initializer_range": 0.02,
17
+ "label2id": {
18
+ "NEGATIVE": 0,
19
+ "POSITIVE": 1
20
+ },
21
+ "max_position_embeddings": 512,
22
+ "model_type": "distilbert",
23
+ "n_heads": 12,
24
+ "n_layers": 6,
25
+ "output_past": true,
26
+ "pad_token_id": 0,
27
+ "qa_dropout": 0.1,
28
+ "seq_classif_dropout": 0.2,
29
+ "sinusoidal_pos_embds": false,
30
+ "tie_weights_": true,
31
+ "transformers_version": "4.26.0.dev0",
32
+ "vocab_size": 30522
33
+ }
eval_results.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "accuracy": 0.8807339449541285
3
+ }
model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66c47853bbe37bf01ac7f95b8e97a4ee4b9eba2d46b554b515a30f352d91c5f6
3
+ size 268396551
ort_config.json ADDED
@@ -0,0 +1,191 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "one_external_file": true,
3
+ "opset": null,
4
+ "optimization": {},
5
+ "optimum_version": "1.6.2.dev0",
6
+ "quantization": {
7
+ "activations_dtype": "QInt8",
8
+ "activations_symmetric": true,
9
+ "format": "QDQ",
10
+ "is_static": true,
11
+ "mode": "QLinearOps",
12
+ "nodes_to_exclude": [
13
+ "/distilbert/transformer/layer.1/output_layer_norm/ReduceMean",
14
+ "/distilbert/transformer/layer.3/output_layer_norm/Div",
15
+ "/distilbert/transformer/layer.3/Add_1",
16
+ "/distilbert/transformer/layer.1/output_layer_norm/Add",
17
+ "/distilbert/transformer/layer.4/sa_layer_norm/ReduceMean_1",
18
+ "/distilbert/transformer/layer.3/sa_layer_norm/Mul",
19
+ "/distilbert/transformer/layer.5/sa_layer_norm/Div",
20
+ "/distilbert/transformer/layer.5/output_layer_norm/ReduceMean_1",
21
+ "/distilbert/transformer/layer.0/Add",
22
+ "/distilbert/transformer/layer.5/sa_layer_norm/Mul",
23
+ "/distilbert/transformer/layer.2/ffn/activation/Add",
24
+ "/distilbert/transformer/layer.0/output_layer_norm/Mul",
25
+ "/distilbert/transformer/layer.5/Add",
26
+ "/distilbert/transformer/layer.3/output_layer_norm/ReduceMean",
27
+ "/distilbert/transformer/layer.0/sa_layer_norm/Sqrt",
28
+ "/distilbert/transformer/layer.5/sa_layer_norm/ReduceMean",
29
+ "/distilbert/transformer/layer.2/Add_1",
30
+ "/distilbert/transformer/layer.1/sa_layer_norm/Sqrt",
31
+ "/distilbert/transformer/layer.5/output_layer_norm/Add",
32
+ "/distilbert/embeddings/LayerNorm/Div",
33
+ "/distilbert/transformer/layer.1/output_layer_norm/Div",
34
+ "/distilbert/transformer/layer.1/ffn/activation/Add",
35
+ "/distilbert/transformer/layer.5/ffn/activation/Mul_1",
36
+ "/distilbert/transformer/layer.2/output_layer_norm/Add",
37
+ "/distilbert/transformer/layer.4/sa_layer_norm/Add_1",
38
+ "/distilbert/transformer/layer.5/sa_layer_norm/Pow",
39
+ "/distilbert/transformer/layer.1/sa_layer_norm/ReduceMean_1",
40
+ "/distilbert/transformer/layer.4/output_layer_norm/Mul",
41
+ "/distilbert/transformer/layer.4/sa_layer_norm/ReduceMean",
42
+ "/distilbert/transformer/layer.2/sa_layer_norm/Sub",
43
+ "/distilbert/transformer/layer.3/sa_layer_norm/Sqrt",
44
+ "/distilbert/transformer/layer.2/ffn/activation/Mul",
45
+ "/distilbert/transformer/layer.4/sa_layer_norm/Mul",
46
+ "/distilbert/transformer/layer.5/output_layer_norm/Div",
47
+ "/distilbert/transformer/layer.5/sa_layer_norm/Add",
48
+ "/distilbert/transformer/layer.0/output_layer_norm/Add_1",
49
+ "/distilbert/transformer/layer.2/Add",
50
+ "/distilbert/transformer/layer.3/ffn/activation/Add",
51
+ "/distilbert/transformer/layer.0/Add_1",
52
+ "/distilbert/transformer/layer.4/output_layer_norm/Sqrt",
53
+ "/distilbert/transformer/layer.4/sa_layer_norm/Sub",
54
+ "/distilbert/transformer/layer.5/output_layer_norm/Add_1",
55
+ "/distilbert/transformer/layer.1/ffn/activation/Mul",
56
+ "/distilbert/transformer/layer.0/ffn/activation/Add",
57
+ "/distilbert/embeddings/LayerNorm/Sqrt",
58
+ "/distilbert/transformer/layer.2/sa_layer_norm/Mul",
59
+ "/distilbert/transformer/layer.0/sa_layer_norm/Pow",
60
+ "/distilbert/transformer/layer.5/ffn/activation/Erf",
61
+ "/distilbert/transformer/layer.2/sa_layer_norm/ReduceMean_1",
62
+ "/distilbert/transformer/layer.4/output_layer_norm/Add_1",
63
+ "/distilbert/transformer/layer.3/output_layer_norm/Add",
64
+ "/distilbert/transformer/layer.2/ffn/activation/Mul_1",
65
+ "/distilbert/transformer/layer.5/ffn/activation/Add",
66
+ "/distilbert/transformer/layer.1/output_layer_norm/Pow",
67
+ "/distilbert/transformer/layer.4/output_layer_norm/ReduceMean",
68
+ "/distilbert/embeddings/LayerNorm/Mul",
69
+ "/distilbert/transformer/layer.5/Add_1",
70
+ "/distilbert/transformer/layer.1/output_layer_norm/Sub",
71
+ "/distilbert/transformer/layer.3/output_layer_norm/Pow",
72
+ "/distilbert/transformer/layer.2/output_layer_norm/Sqrt",
73
+ "/distilbert/transformer/layer.0/output_layer_norm/Sub",
74
+ "/distilbert/transformer/layer.3/ffn/activation/Div",
75
+ "/distilbert/transformer/layer.5/sa_layer_norm/Add_1",
76
+ "/distilbert/transformer/layer.3/output_layer_norm/ReduceMean_1",
77
+ "/distilbert/transformer/layer.0/ffn/activation/Mul",
78
+ "/distilbert/transformer/layer.1/output_layer_norm/Add_1",
79
+ "/distilbert/embeddings/LayerNorm/ReduceMean",
80
+ "/distilbert/transformer/layer.4/ffn/activation/Mul_1",
81
+ "/distilbert/transformer/layer.1/output_layer_norm/ReduceMean_1",
82
+ "/distilbert/transformer/layer.4/output_layer_norm/ReduceMean_1",
83
+ "/distilbert/transformer/layer.5/output_layer_norm/Sqrt",
84
+ "/distilbert/transformer/layer.3/sa_layer_norm/Pow",
85
+ "/distilbert/transformer/layer.0/output_layer_norm/Pow",
86
+ "/distilbert/transformer/layer.2/sa_layer_norm/Add_1",
87
+ "/distilbert/transformer/layer.3/output_layer_norm/Sqrt",
88
+ "/distilbert/transformer/layer.5/sa_layer_norm/Sqrt",
89
+ "/distilbert/transformer/layer.0/sa_layer_norm/ReduceMean",
90
+ "/distilbert/embeddings/LayerNorm/ReduceMean_1",
91
+ "/distilbert/transformer/layer.5/output_layer_norm/ReduceMean",
92
+ "/distilbert/transformer/layer.2/sa_layer_norm/Sqrt",
93
+ "/distilbert/embeddings/LayerNorm/Sub",
94
+ "/distilbert/transformer/layer.0/sa_layer_norm/Mul",
95
+ "/distilbert/transformer/layer.1/output_layer_norm/Mul",
96
+ "/distilbert/transformer/layer.1/sa_layer_norm/ReduceMean",
97
+ "/distilbert/transformer/layer.4/sa_layer_norm/Div",
98
+ "/distilbert/transformer/layer.1/ffn/activation/Erf",
99
+ "/distilbert/transformer/layer.0/output_layer_norm/ReduceMean_1",
100
+ "/distilbert/transformer/layer.0/sa_layer_norm/Sub",
101
+ "/distilbert/transformer/layer.0/output_layer_norm/ReduceMean",
102
+ "/distilbert/transformer/layer.4/ffn/activation/Mul",
103
+ "/distilbert/embeddings/LayerNorm/Pow",
104
+ "/distilbert/transformer/layer.1/sa_layer_norm/Add_1",
105
+ "/distilbert/transformer/layer.2/sa_layer_norm/Div",
106
+ "/distilbert/transformer/layer.1/ffn/activation/Div",
107
+ "/distilbert/transformer/layer.0/ffn/activation/Div",
108
+ "/distilbert/transformer/layer.1/sa_layer_norm/Pow",
109
+ "/distilbert/transformer/layer.4/sa_layer_norm/Add",
110
+ "/distilbert/transformer/layer.2/output_layer_norm/Add_1",
111
+ "/distilbert/transformer/layer.3/ffn/activation/Erf",
112
+ "/distilbert/transformer/layer.0/ffn/activation/Erf",
113
+ "/distilbert/transformer/layer.0/sa_layer_norm/Add",
114
+ "/distilbert/transformer/layer.3/sa_layer_norm/ReduceMean_1",
115
+ "/distilbert/transformer/layer.0/sa_layer_norm/Div",
116
+ "/distilbert/transformer/layer.3/output_layer_norm/Sub",
117
+ "/distilbert/transformer/layer.5/output_layer_norm/Pow",
118
+ "/distilbert/transformer/layer.1/sa_layer_norm/Div",
119
+ "/distilbert/embeddings/Add",
120
+ "/distilbert/transformer/layer.1/sa_layer_norm/Sub",
121
+ "/distilbert/transformer/layer.2/sa_layer_norm/Add",
122
+ "/distilbert/transformer/layer.1/sa_layer_norm/Add",
123
+ "/distilbert/transformer/layer.2/ffn/activation/Div",
124
+ "/distilbert/transformer/layer.1/output_layer_norm/Sqrt",
125
+ "/distilbert/transformer/layer.2/output_layer_norm/ReduceMean",
126
+ "/distilbert/embeddings/LayerNorm/Add",
127
+ "/distilbert/transformer/layer.1/Add_1",
128
+ "/distilbert/transformer/layer.2/ffn/activation/Erf",
129
+ "/distilbert/transformer/layer.0/output_layer_norm/Sqrt",
130
+ "/distilbert/transformer/layer.1/sa_layer_norm/Mul",
131
+ "/distilbert/transformer/layer.3/sa_layer_norm/Add",
132
+ "/distilbert/transformer/layer.4/ffn/activation/Erf",
133
+ "/distilbert/transformer/layer.5/sa_layer_norm/ReduceMean_1",
134
+ "/distilbert/transformer/layer.3/sa_layer_norm/Add_1",
135
+ "/distilbert/transformer/layer.0/sa_layer_norm/Add_1",
136
+ "/distilbert/transformer/layer.2/output_layer_norm/ReduceMean_1",
137
+ "/distilbert/transformer/layer.3/ffn/activation/Mul_1",
138
+ "/distilbert/embeddings/LayerNorm/Add_1",
139
+ "/distilbert/transformer/layer.0/output_layer_norm/Div",
140
+ "/distilbert/transformer/layer.3/output_layer_norm/Mul",
141
+ "/distilbert/transformer/layer.1/ffn/activation/Mul_1",
142
+ "/distilbert/transformer/layer.2/output_layer_norm/Pow",
143
+ "/distilbert/transformer/layer.3/output_layer_norm/Add_1",
144
+ "/distilbert/transformer/layer.3/ffn/activation/Mul",
145
+ "/distilbert/transformer/layer.0/sa_layer_norm/ReduceMean_1",
146
+ "/distilbert/transformer/layer.4/sa_layer_norm/Sqrt",
147
+ "/distilbert/transformer/layer.2/output_layer_norm/Mul",
148
+ "/distilbert/transformer/layer.5/output_layer_norm/Sub",
149
+ "/distilbert/transformer/layer.1/Add",
150
+ "/distilbert/transformer/layer.4/ffn/activation/Add",
151
+ "/distilbert/transformer/layer.4/ffn/activation/Div",
152
+ "/distilbert/transformer/layer.2/sa_layer_norm/ReduceMean",
153
+ "/distilbert/transformer/layer.4/Add",
154
+ "/distilbert/transformer/layer.4/output_layer_norm/Sub",
155
+ "/distilbert/transformer/layer.0/ffn/activation/Mul_1",
156
+ "/distilbert/transformer/layer.4/Add_1",
157
+ "/distilbert/transformer/layer.4/sa_layer_norm/Pow",
158
+ "/distilbert/transformer/layer.3/sa_layer_norm/Sub",
159
+ "/distilbert/transformer/layer.2/output_layer_norm/Sub",
160
+ "/distilbert/transformer/layer.0/output_layer_norm/Add",
161
+ "/distilbert/transformer/layer.5/ffn/activation/Mul",
162
+ "/distilbert/transformer/layer.5/output_layer_norm/Mul",
163
+ "/distilbert/transformer/layer.4/output_layer_norm/Pow",
164
+ "/distilbert/transformer/layer.5/ffn/activation/Div",
165
+ "/distilbert/transformer/layer.4/output_layer_norm/Div",
166
+ "/distilbert/transformer/layer.3/Add",
167
+ "/distilbert/transformer/layer.5/sa_layer_norm/Sub",
168
+ "/distilbert/transformer/layer.2/output_layer_norm/Div",
169
+ "/distilbert/transformer/layer.4/output_layer_norm/Add",
170
+ "/distilbert/transformer/layer.3/sa_layer_norm/Div",
171
+ "/distilbert/transformer/layer.2/sa_layer_norm/Pow",
172
+ "/distilbert/transformer/layer.3/sa_layer_norm/ReduceMean"
173
+ ],
174
+ "nodes_to_quantize": [],
175
+ "operators_to_quantize": [
176
+ "MatMul",
177
+ "Add"
178
+ ],
179
+ "per_channel": true,
180
+ "qdq_add_pair_to_weight": true,
181
+ "qdq_dedicated_pair": true,
182
+ "qdq_op_type_per_channel_support_to_axis": {
183
+ "MatMul": 1
184
+ },
185
+ "reduce_range": false,
186
+ "weights_dtype": "QInt8",
187
+ "weights_symmetric": true
188
+ },
189
+ "transformers_version": "4.26.0.dev0",
190
+ "use_external_data_format": false
191
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "mask_token": "[MASK]",
4
+ "pad_token": "[PAD]",
5
+ "sep_token": "[SEP]",
6
+ "unk_token": "[UNK]"
7
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cls_token": "[CLS]",
3
+ "do_basic_tokenize": true,
4
+ "do_lower_case": true,
5
+ "mask_token": "[MASK]",
6
+ "model_max_length": 512,
7
+ "name_or_path": "distilbert_onnx",
8
+ "never_split": null,
9
+ "pad_token": "[PAD]",
10
+ "sep_token": "[SEP]",
11
+ "special_tokens_map_file": null,
12
+ "strip_accents": null,
13
+ "tokenize_chinese_chars": true,
14
+ "tokenizer_class": "DistilBertTokenizer",
15
+ "unk_token": "[UNK]"
16
+ }
vocab.txt ADDED
The diff for this file is too large to render. See raw diff