Upload WhisperForConditionalGeneration
Browse files- config.json +121 -0
- pytorch_model.bin +3 -0
config.json
ADDED
|
@@ -0,0 +1,121 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "whisper/small",
|
| 3 |
+
"activation_dropout": 0.0,
|
| 4 |
+
"activation_function": "gelu",
|
| 5 |
+
"architectures": [
|
| 6 |
+
"WhisperForConditionalGeneration"
|
| 7 |
+
],
|
| 8 |
+
"attention_dropout": 0.0,
|
| 9 |
+
"bos_token_id": 50258,
|
| 10 |
+
"d_model": 768,
|
| 11 |
+
"decoder_attention_heads": 12,
|
| 12 |
+
"decoder_layerdrop": 0.0,
|
| 13 |
+
"decoder_layers": 12,
|
| 14 |
+
"decoder_start_token_id": 2,
|
| 15 |
+
"dropout": 0.0,
|
| 16 |
+
"encoder_attention_heads": 12,
|
| 17 |
+
"encoder_layerdrop": 0.0,
|
| 18 |
+
"encoder_layers": 12,
|
| 19 |
+
"eos_token_id": 50257,
|
| 20 |
+
"feature_size": 1,
|
| 21 |
+
"init_std": 0.02,
|
| 22 |
+
"input_channels": 1,
|
| 23 |
+
"is_encoder_decoder": true,
|
| 24 |
+
"max_source_positions": 1500,
|
| 25 |
+
"max_target_positions": 448,
|
| 26 |
+
"model_type": "whisper",
|
| 27 |
+
"non_speech_tokens": [
|
| 28 |
+
1,
|
| 29 |
+
2,
|
| 30 |
+
6,
|
| 31 |
+
7,
|
| 32 |
+
8,
|
| 33 |
+
9,
|
| 34 |
+
10,
|
| 35 |
+
12,
|
| 36 |
+
14,
|
| 37 |
+
25,
|
| 38 |
+
26,
|
| 39 |
+
27,
|
| 40 |
+
28,
|
| 41 |
+
29,
|
| 42 |
+
31,
|
| 43 |
+
58,
|
| 44 |
+
59,
|
| 45 |
+
60,
|
| 46 |
+
61,
|
| 47 |
+
62,
|
| 48 |
+
63,
|
| 49 |
+
90,
|
| 50 |
+
91,
|
| 51 |
+
92,
|
| 52 |
+
93,
|
| 53 |
+
359,
|
| 54 |
+
503,
|
| 55 |
+
522,
|
| 56 |
+
542,
|
| 57 |
+
873,
|
| 58 |
+
893,
|
| 59 |
+
902,
|
| 60 |
+
918,
|
| 61 |
+
922,
|
| 62 |
+
931,
|
| 63 |
+
1350,
|
| 64 |
+
1853,
|
| 65 |
+
1982,
|
| 66 |
+
2460,
|
| 67 |
+
2627,
|
| 68 |
+
3246,
|
| 69 |
+
3253,
|
| 70 |
+
3268,
|
| 71 |
+
3536,
|
| 72 |
+
3846,
|
| 73 |
+
3961,
|
| 74 |
+
4183,
|
| 75 |
+
4667,
|
| 76 |
+
6585,
|
| 77 |
+
6647,
|
| 78 |
+
7273,
|
| 79 |
+
9061,
|
| 80 |
+
9383,
|
| 81 |
+
10428,
|
| 82 |
+
10929,
|
| 83 |
+
11938,
|
| 84 |
+
12033,
|
| 85 |
+
12331,
|
| 86 |
+
12562,
|
| 87 |
+
13793,
|
| 88 |
+
14157,
|
| 89 |
+
14635,
|
| 90 |
+
15265,
|
| 91 |
+
15618,
|
| 92 |
+
16553,
|
| 93 |
+
16604,
|
| 94 |
+
18362,
|
| 95 |
+
18956,
|
| 96 |
+
20075,
|
| 97 |
+
21675,
|
| 98 |
+
22520,
|
| 99 |
+
26130,
|
| 100 |
+
26161,
|
| 101 |
+
26435,
|
| 102 |
+
28279,
|
| 103 |
+
29464,
|
| 104 |
+
31650,
|
| 105 |
+
32302,
|
| 106 |
+
32470,
|
| 107 |
+
36865,
|
| 108 |
+
42863,
|
| 109 |
+
47425,
|
| 110 |
+
49870,
|
| 111 |
+
50254
|
| 112 |
+
],
|
| 113 |
+
"num_hidden_layers": 12,
|
| 114 |
+
"num_mel_bins": 80,
|
| 115 |
+
"pad_token_id": 0,
|
| 116 |
+
"scale_embedding": false,
|
| 117 |
+
"torch_dtype": "float32",
|
| 118 |
+
"transformers_version": "4.23.0.dev0",
|
| 119 |
+
"use_cache": true,
|
| 120 |
+
"vocab_size": 51865
|
| 121 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed956b1c073bc45d2cb0c50b6148e8d623f1e545a22dc548d8dd2d1bfe6c9b34
|
| 3 |
+
size 967092547
|