Hugging Face's logo

Musembi
/
whisper-small-kamba-model

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
whisper
Generated from Trainer
Model card Files
xet
 Metrics Community
Musembi commited on
Commit
6676df8
·
verified ·
1 Parent(s): e375036

Fine-tuned Whisper Small on FLEURS Kamba — KambaVoice v1.0

Browse files
Files changed (7) hide show
  1. README.md +88 -0
  2. model.safetensors +1 -1
  3. processor_config.json +17 -0
  4. runs/May13_08-32-28_9263f3ad3166/events.out.tfevents.1778661148.9263f3ad3166.5757.0 +2 -2
  5. runs/May13_08-32-28_9263f3ad3166/events.out.tfevents.1778666336.9263f3ad3166.5757.1 +3 -0
  6. tokenizer.json +0 -0
  7. tokenizer_config.json +21 -0
README.md ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: openai/whisper-small
5
+ tags:
6
+ - generated_from_trainer
7
+ datasets:
8
+ - fleurs
9
+ metrics:
10
+ - wer
11
+ model-index:
12
+ - name: whisper-small-kamba-model
13
+ results:
14
+ - task:
15
+ name: Automatic Speech Recognition
16
+ type: automatic-speech-recognition
17
+ dataset:
18
+ name: fleurs
19
+ type: fleurs
20
+ config: kam_ke
21
+ split: None
22
+ args: kam_ke
23
+ metrics:
24
+ - name: Wer
25
+ type: wer
26
+ value: 77.01
27
+ ---
28
+
29
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
30
+ should probably proofread and complete it, then remove this comment. -->
31
+
32
+ # whisper-small-kamba-model
33
+
34
+ This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the fleurs dataset.
35
+ It achieves the following results on the evaluation set:
36
+ - Loss: 1.1138
37
+ - Wer: 77.01
38
+ - Cer: 25.99
39
+
40
+ ## Model description
41
+
42
+ More information needed
43
+
44
+ ## Intended uses & limitations
45
+
46
+ More information needed
47
+
48
+ ## Training and evaluation data
49
+
50
+ More information needed
51
+
52
+ ## Training procedure
53
+
54
+ ### Training hyperparameters
55
+
56
+ The following hyperparameters were used during training:
57
+ - learning_rate: 1e-05
58
+ - train_batch_size: 8
59
+ - eval_batch_size: 8
60
+ - seed: 42
61
+ - gradient_accumulation_steps: 2
62
+ - total_train_batch_size: 16
63
+ - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
64
+ - lr_scheduler_type: linear
65
+ - lr_scheduler_warmup_steps: 200
66
+ - training_steps: 800
67
+ - mixed_precision_training: Native AMP
68
+
69
+ ### Training results
70
+
71
+ | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
72
+ |:-------------:|:------:|:----:|:---------------:|:------:|:------:|
73
+ | 4.8287 | 0.4785 | 100 | 1.9965 | 176.18 | 112.93 |
74
+ | 3.2071 | 0.9569 | 200 | 1.4722 | 85.75 | 29.63 |
75
+ | 2.4629 | 1.4354 | 300 | 1.2785 | 86.1 | 28.44 |
76
+ | 2.1983 | 1.9139 | 400 | 1.1858 | 77.28 | 26.53 |
77
+ | 1.8547 | 2.3923 | 500 | 1.1541 | 80.92 | 27.11 |
78
+ | 1.7905 | 2.8708 | 600 | 1.1224 | 81.2 | 26.99 |
79
+ | 1.5513 | 3.3493 | 700 | 1.1189 | 78.15 | 25.88 |
80
+ | 1.5585 | 3.8278 | 800 | 1.1130 | 80.33 | 26.31 |
81
+
82
+
83
+ ### Framework versions
84
+
85
+ - Transformers 5.0.0
86
+ - Pytorch 2.10.0+cu128
87
+ - Datasets 2.21.0
88
+ - Tokenizers 0.22.2
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1caed77a569bcbe91f62a92ad847f1de28787a39274680ac5e15def49cdaf7f7
3
  size 966998152
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3290930f519cf54074b62c782728bdbb2e21ba730681e2ab99c49b9c6184e8cc
3
  size 966998152
processor_config.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "feature_extractor": {
3
+ "chunk_length": 30,
4
+ "dither": 0.0,
5
+ "feature_extractor_type": "WhisperFeatureExtractor",
6
+ "feature_size": 80,
7
+ "hop_length": 160,
8
+ "n_fft": 400,
9
+ "n_samples": 480000,
10
+ "nb_max_frames": 3000,
11
+ "padding_side": "right",
12
+ "padding_value": 0.0,
13
+ "return_attention_mask": false,
14
+ "sampling_rate": 16000
15
+ },
16
+ "processor_class": "WhisperProcessor"
17
+ }
runs/May13_08-32-28_9263f3ad3166/events.out.tfevents.1778661148.9263f3ad3166.5757.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:68dd537b81e8b0f00e7acebf8ca090980a74dfb30e6a616d570b56f0097761e0
3
- size 14663
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86cceb1093087cdad895081bee1826f12e1d795e553eef4ce715d1e72014ae01
3
+ size 15017
runs/May13_08-32-28_9263f3ad3166/events.out.tfevents.1778666336.9263f3ad3166.5757.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d499e2d56ef47fb0aa3e67c872e0d262828dba0c24aac2cdca86ff912ae66dc
3
+ size 453
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "clean_up_tokenization_spaces": true,
6
+ "eos_token": "<|endoftext|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|kam|>"
10
+ ],
11
+ "is_local": false,
12
+ "language": null,
13
+ "model_max_length": 1024,
14
+ "pad_token": "<|endoftext|>",
15
+ "predict_timestamps": false,
16
+ "processor_class": "WhisperProcessor",
17
+ "return_attention_mask": false,
18
+ "task": "transcribe",
19
+ "tokenizer_class": "WhisperTokenizer",
20
+ "unk_token": "<|endoftext|>"
21
+ }