Hugging Face's logo

Sagicc
/
whisper-base-sr-onnx

Automatic Speech Recognition
Transformers.js
ONNX
Serbian
whisper
Model card Files Community
Sagicc commited on
Commit
3a574b9
•
1 Parent(s): 8873979

Upload 19 files

Browse files
Files changed (12) hide show
  1. config.json +11 -9
  2. generation_config.json +15 -23
  3. onnx/decoder_model.onnx +3 -0
  4. onnx/decoder_model_merged.onnx +3 -0
  5. onnx/decoder_model_merged_quantized.onnx +3 -0
  6. onnx/decoder_model_quantized.onnx +3 -0
  7. onnx/decoder_with_past_model.onnx +3 -0
  8. onnx/decoder_with_past_model_quantized.onnx +3 -0
  9. onnx/encoder_model.onnx +3 -0
  10. onnx/encoder_model_quantized.onnx +3 -0
  11. quantize_config.json +65 -65
  12. tokenizer_config.json +1 -1
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "Sagicc/whisper-small-sr-yodas-v2",
3
  "activation_dropout": 0.0,
4
  "activation_function": "gelu",
5
  "apply_spec_augment": false,
@@ -13,17 +13,17 @@
13
  ],
14
  "bos_token_id": 50257,
15
  "classifier_proj_size": 256,
16
- "d_model": 768,
17
- "decoder_attention_heads": 12,
18
- "decoder_ffn_dim": 3072,
19
  "decoder_layerdrop": 0.0,
20
- "decoder_layers": 12,
21
  "decoder_start_token_id": 50258,
22
  "dropout": 0.0,
23
- "encoder_attention_heads": 12,
24
- "encoder_ffn_dim": 3072,
25
  "encoder_layerdrop": 0.0,
26
- "encoder_layers": 12,
27
  "eos_token_id": 50257,
28
  "forced_decoder_ids": [
29
  [
@@ -52,7 +52,7 @@
52
  "max_target_positions": 448,
53
  "median_filter_width": 7,
54
  "model_type": "whisper",
55
- "num_hidden_layers": 12,
56
  "num_mel_bins": 80,
57
  "pad_token_id": 50257,
58
  "scale_embedding": false,
@@ -140,6 +140,8 @@
140
  49870,
141
  50254,
142
  50258,
 
 
143
  50360,
144
  50361,
145
  50362
 
1
  {
2
+ "_name_or_path": "Sagicc/whisper-base-sr-yodas",
3
  "activation_dropout": 0.0,
4
  "activation_function": "gelu",
5
  "apply_spec_augment": false,
 
13
  ],
14
  "bos_token_id": 50257,
15
  "classifier_proj_size": 256,
16
+ "d_model": 512,
17
+ "decoder_attention_heads": 8,
18
+ "decoder_ffn_dim": 2048,
19
  "decoder_layerdrop": 0.0,
20
+ "decoder_layers": 6,
21
  "decoder_start_token_id": 50258,
22
  "dropout": 0.0,
23
+ "encoder_attention_heads": 8,
24
+ "encoder_ffn_dim": 2048,
25
  "encoder_layerdrop": 0.0,
26
+ "encoder_layers": 6,
27
  "eos_token_id": 50257,
28
  "forced_decoder_ids": [
29
  [
 
52
  "max_target_positions": 448,
53
  "median_filter_width": 7,
54
  "model_type": "whisper",
55
+ "num_hidden_layers": 6,
56
  "num_mel_bins": 80,
57
  "pad_token_id": 50257,
58
  "scale_embedding": false,
 
140
  49870,
141
  50254,
142
  50258,
143
+ 50358,
144
+ 50359,
145
  50360,
146
  50361,
147
  50362
generation_config.json CHANGED
@@ -1,44 +1,36 @@
1
  {
2
  "alignment_heads": [
3
  [
4
- 5,
5
- 3
6
- ],
7
- [
8
- 5,
9
- 9
10
  ],
11
  [
12
- 8,
13
- 0
14
  ],
15
  [
16
- 8,
17
- 4
18
  ],
19
  [
20
- 8,
21
  7
22
  ],
23
  [
24
- 8,
25
- 8
26
- ],
27
- [
28
- 9,
29
- 0
30
  ],
31
  [
32
- 9,
33
- 7
34
  ],
35
  [
36
- 9,
37
- 9
38
  ],
39
  [
40
- 10,
41
- 5
42
  ]
43
  ],
44
  "begin_suppress_tokens": [
 
1
  {
2
  "alignment_heads": [
3
  [
4
+ 3,
5
+ 1
 
 
 
 
6
  ],
7
  [
8
+ 4,
9
+ 2
10
  ],
11
  [
12
+ 4,
13
+ 3
14
  ],
15
  [
16
+ 4,
17
  7
18
  ],
19
  [
20
+ 5,
21
+ 1
 
 
 
 
22
  ],
23
  [
24
+ 5,
25
+ 2
26
  ],
27
  [
28
+ 5,
29
+ 4
30
  ],
31
  [
32
+ 5,
33
+ 6
34
  ]
35
  ],
36
  "begin_suppress_tokens": [
onnx/decoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74b221e15fc84ac646dc5f85df8945a9ef9753bb19706de8ec70be01c32cdd95
3
+ size 208271694
onnx/decoder_model_merged.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a39a68964d0173d925299dcd90220daf0d4f47eb25ca04804c0f4991d98f0b4e
3
+ size 208491287
onnx/decoder_model_merged_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c2b9d923d0ca5133b257a76ce1f1fbe24add643d8fa223076f8feee021fec92
3
+ size 53617060
onnx/decoder_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30f029a380dbd74b8813c706b1b7acece27660aa64832eeadad1634cf67ee93c
3
+ size 53267561
onnx/decoder_with_past_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db76c161a5211025eac23111ab6c82ffdb30cf4402dc1c69c07784c05f0153d6
3
+ size 195635283
onnx/decoder_with_past_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:beb9868434fead466b56b49dcee2ac4543d602b38a1fa8bab4f23fb0e5b2ee8d
3
+ size 50041643
onnx/encoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9af1f15492d1e4cd1bf7e71912e498e2ccbfd5630908e5ff73b2ea5f32fa6e02
3
+ size 82461622
onnx/encoder_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9d80a0831dee24e9a0a145883bb4ac097ccad7dce606f7902413445f403e0a4
3
+ size 23183301
quantize_config.json CHANGED
@@ -4,110 +4,110 @@
4
  "per_model_config": {
5
  "encoder_model": {
6
  "op_types": [
7
- "Transpose",
8
- "Div",
9
- "Shape",
10
- "Concat",
11
- "Softmax",
12
- "Sqrt",
13
- "Pow",
14
- "ReduceMean",
15
  "Sub",
16
- "Unsqueeze",
17
  "Mul",
 
18
  "Add",
19
  "Conv",
20
- "MatMul",
 
21
  "Constant",
22
- "Reshape",
23
- "Erf",
24
- "Gather"
 
 
 
 
 
 
25
  ],
26
  "weight_type": "QUInt8"
27
  },
28
  "decoder_model_merged": {
29
  "op_types": [
30
- "ReduceMean",
31
- "Unsqueeze",
32
- "Less",
33
- "Gather",
34
- "Concat",
35
- "Cast",
36
  "Range",
37
- "Mul",
38
- "Equal",
39
- "Slice",
40
- "Expand",
41
- "Constant",
42
- "Erf",
43
- "Transpose",
44
  "Div",
 
 
 
45
  "Softmax",
46
- "Squeeze",
47
- "Reshape",
48
- "If",
49
  "Shape",
50
- "Sqrt",
51
  "Pow",
52
- "ConstantOfShape",
53
- "Add",
54
  "MatMul",
 
 
 
 
 
 
 
 
55
  "Where",
56
- "Sub"
 
 
 
57
  ],
58
  "weight_type": "QInt8"
59
  },
60
  "decoder_model": {
61
  "op_types": [
62
- "ReduceMean",
63
- "Unsqueeze",
64
- "Less",
65
- "Gather",
66
- "Concat",
67
- "Cast",
68
  "Range",
69
- "Mul",
70
- "Equal",
71
- "Slice",
72
- "Expand",
73
- "Constant",
74
- "Erf",
75
- "Transpose",
76
  "Div",
 
 
 
77
  "Softmax",
78
- "Squeeze",
79
- "Reshape",
80
  "Shape",
81
- "Sqrt",
82
  "Pow",
83
- "ConstantOfShape",
84
- "Add",
85
  "MatMul",
 
 
 
 
 
 
 
86
  "Where",
87
- "Sub"
 
 
 
88
  ],
89
  "weight_type": "QInt8"
90
  },
91
  "decoder_with_past_model": {
92
  "op_types": [
93
- "Concat",
94
- "Div",
95
- "Shape",
 
 
 
96
  "Transpose",
97
  "Softmax",
98
- "Sqrt",
99
- "Pow",
100
  "ReduceMean",
101
- "Erf",
 
102
  "Unsqueeze",
103
- "Mul",
104
- "Add",
105
- "Slice",
106
  "MatMul",
107
- "Constant",
108
- "Reshape",
109
- "Sub",
110
- "Gather"
111
  ],
112
  "weight_type": "QInt8"
113
  }
 
4
  "per_model_config": {
5
  "encoder_model": {
6
  "op_types": [
 
 
 
 
 
 
 
 
7
  "Sub",
8
+ "Gather",
9
  "Mul",
10
+ "Reshape",
11
  "Add",
12
  "Conv",
13
+ "Transpose",
14
+ "Softmax",
15
  "Constant",
16
+ "Concat",
17
+ "ReduceMean",
18
+ "Div",
19
+ "Sqrt",
20
+ "Unsqueeze",
21
+ "MatMul",
22
+ "Shape",
23
+ "Pow",
24
+ "Erf"
25
  ],
26
  "weight_type": "QUInt8"
27
  },
28
  "decoder_model_merged": {
29
  "op_types": [
 
 
 
 
 
 
30
  "Range",
 
 
 
 
 
 
 
31
  "Div",
32
+ "Equal",
33
+ "Gather",
34
+ "Cast",
35
  "Softmax",
36
+ "ReduceMean",
37
+ "Concat",
 
38
  "Shape",
39
+ "Less",
40
  "Pow",
41
+ "Reshape",
42
+ "Sub",
43
  "MatMul",
44
+ "If",
45
+ "Expand",
46
+ "Squeeze",
47
+ "Mul",
48
+ "Slice",
49
+ "Add",
50
+ "Transpose",
51
+ "Constant",
52
  "Where",
53
+ "ConstantOfShape",
54
+ "Unsqueeze",
55
+ "Sqrt",
56
+ "Erf"
57
  ],
58
  "weight_type": "QInt8"
59
  },
60
  "decoder_model": {
61
  "op_types": [
 
 
 
 
 
 
62
  "Range",
 
 
 
 
 
 
 
63
  "Div",
64
+ "Equal",
65
+ "Gather",
66
+ "Cast",
67
  "Softmax",
68
+ "ReduceMean",
69
+ "Concat",
70
  "Shape",
71
+ "Less",
72
  "Pow",
73
+ "Reshape",
74
+ "Sub",
75
  "MatMul",
76
+ "Expand",
77
+ "Squeeze",
78
+ "Mul",
79
+ "Slice",
80
+ "Add",
81
+ "Transpose",
82
+ "Constant",
83
  "Where",
84
+ "ConstantOfShape",
85
+ "Unsqueeze",
86
+ "Sqrt",
87
+ "Erf"
88
  ],
89
  "weight_type": "QInt8"
90
  },
91
  "decoder_with_past_model": {
92
  "op_types": [
93
+ "Gather",
94
+ "Reshape",
95
+ "Sub",
96
+ "Slice",
97
+ "Mul",
98
+ "Add",
99
  "Transpose",
100
  "Softmax",
101
+ "Constant",
 
102
  "ReduceMean",
103
+ "Pow",
104
+ "Div",
105
  "Unsqueeze",
106
+ "Sqrt",
 
 
107
  "MatMul",
108
+ "Concat",
109
+ "Shape",
110
+ "Erf"
 
111
  ],
112
  "weight_type": "QInt8"
113
  }
tokenizer_config.json CHANGED
@@ -12994,7 +12994,7 @@
12994
  "single_word": false
12995
  },
12996
  "errors": "replace",
12997
- "model_max_length": 1024,
12998
  "pad_token": {
12999
  "__type": "AddedToken",
13000
  "content": "<|endoftext|>",
 
12994
  "single_word": false
12995
  },
12996
  "errors": "replace",
12997
+ "model_max_length": 448,
12998
  "pad_token": {
12999
  "__type": "AddedToken",
13000
  "content": "<|endoftext|>",