Divyasreepat
commited on
Commit
•
a2d4c24
1
Parent(s):
2b6815e
Upload folder using huggingface_hub
Browse files- README.md +18 -0
- assets/tokenizer/merges.txt +0 -0
- assets/tokenizer/vocabulary.json +0 -0
- audio_converter.json +22 -0
- config.json +20 -0
- metadata.json +6 -0
- model.weights.h5 +3 -0
- tokenizer.json +25 -0
README.md
ADDED
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
library_name: keras-hub
|
3 |
+
---
|
4 |
+
This is a [`Whisper` model](https://keras.io/api/keras_hub/models/whisper) uploaded using the KerasHub library and can be used with JAX, TensorFlow, and PyTorch backends.
|
5 |
+
Model config:
|
6 |
+
* **name:** whisper_backbone
|
7 |
+
* **trainable:** True
|
8 |
+
* **vocabulary_size:** 51864
|
9 |
+
* **num_layers:** 6
|
10 |
+
* **num_heads:** 8
|
11 |
+
* **hidden_dim:** 512
|
12 |
+
* **intermediate_dim:** 2048
|
13 |
+
* **num_mels:** 80
|
14 |
+
* **dropout:** 0.0
|
15 |
+
* **max_encoder_sequence_length:** 3000
|
16 |
+
* **max_decoder_sequence_length:** 448
|
17 |
+
|
18 |
+
This model card has been generated automatically and should be completed by the model author. See [Model Cards documentation](https://huggingface.co/docs/hub/model-cards) for more information.
|
assets/tokenizer/merges.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
assets/tokenizer/vocabulary.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
audio_converter.json
ADDED
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"module": "keras_nlp.src.models.whisper.whisper_audio_converter",
|
3 |
+
"class_name": "WhisperAudioConverter",
|
4 |
+
"config": {
|
5 |
+
"name": "whisper_audio_converter",
|
6 |
+
"trainable": true,
|
7 |
+
"dtype": {
|
8 |
+
"module": "keras",
|
9 |
+
"class_name": "DTypePolicy",
|
10 |
+
"config": {
|
11 |
+
"name": "float32"
|
12 |
+
},
|
13 |
+
"registered_name": null
|
14 |
+
},
|
15 |
+
"num_mels": 80,
|
16 |
+
"num_fft_bins": 400,
|
17 |
+
"stride": 160,
|
18 |
+
"sampling_rate": 16000,
|
19 |
+
"max_audio_length": 30
|
20 |
+
},
|
21 |
+
"registered_name": "keras_nlp>WhisperAudioConverter"
|
22 |
+
}
|
config.json
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"module": "keras_nlp.src.models.whisper.whisper_backbone",
|
3 |
+
"class_name": "WhisperBackbone",
|
4 |
+
"config": {
|
5 |
+
"name": "whisper_backbone",
|
6 |
+
"trainable": true,
|
7 |
+
"vocabulary_size": 51864,
|
8 |
+
"num_layers": 6,
|
9 |
+
"num_heads": 8,
|
10 |
+
"hidden_dim": 512,
|
11 |
+
"intermediate_dim": 2048,
|
12 |
+
"num_mels": 80,
|
13 |
+
"dropout": 0.0,
|
14 |
+
"max_encoder_sequence_length": 3000,
|
15 |
+
"max_decoder_sequence_length": 448
|
16 |
+
},
|
17 |
+
"registered_name": "keras_nlp>WhisperBackbone",
|
18 |
+
"assets": [],
|
19 |
+
"weights": "model.weights.h5"
|
20 |
+
}
|
metadata.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"keras_version": "3.0.1",
|
3 |
+
"keras_nlp_version": "0.7.0",
|
4 |
+
"parameter_count": 72593408,
|
5 |
+
"date_saved": "2023-12-27@02:32:06"
|
6 |
+
}
|
model.weights.h5
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:08fc5e7cfebc9247c6b451166ce25b8392c03606210e75bca5e10439f7bc6483
|
3 |
+
size 290909920
|
tokenizer.json
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"module": "keras_nlp.src.models.whisper.whisper_tokenizer",
|
3 |
+
"class_name": "WhisperTokenizer",
|
4 |
+
"config": {
|
5 |
+
"name": "whisper_tokenizer",
|
6 |
+
"trainable": true,
|
7 |
+
"dtype": "int32",
|
8 |
+
"sequence_length": null,
|
9 |
+
"add_prefix_space": false,
|
10 |
+
"special_tokens": {
|
11 |
+
"<|startoftranscript|>": 50257,
|
12 |
+
"<|endoftext|>": 50256,
|
13 |
+
"<|notimestamps|>": 50362,
|
14 |
+
"<|translate|>": 50358,
|
15 |
+
"<|transcribe|>": 50357
|
16 |
+
},
|
17 |
+
"language_tokens": null
|
18 |
+
},
|
19 |
+
"registered_name": "keras_nlp>WhisperTokenizer",
|
20 |
+
"assets": [
|
21 |
+
"assets/tokenizer/merges.txt",
|
22 |
+
"assets/tokenizer/vocabulary.json"
|
23 |
+
],
|
24 |
+
"weights": null
|
25 |
+
}
|