File size: 2,565 Bytes
cce5311
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
---
library_name: peft
language:
- it
license: apache-2.0
base_model: openai/whisper-medium
tags:
- generated_from_trainer
datasets:
- b-brave-clean
metrics:
- wer
model-index:
- name: Whisper Medium
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: b-brave-clean
      type: b-brave-clean
      config: default
      split: test
      args: default
    metrics:
    - type: wer
      value: 56.73352435530086
      name: Wer
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# Whisper Medium

This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the b-brave-clean dataset.
It achieves the following results on the evaluation set:
- Loss: 0.6333
- Wer: 56.7335
- Cer: 39.2960
- Lr: 0.0000

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.3
- num_epochs: 8
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer     | Cer     | Lr     |
|:-------------:|:------:|:----:|:---------------:|:-------:|:-------:|:------:|
| 4.1253        | 1.0    | 502  | 3.7472          | 83.6676 | 51.0638 | 0.0000 |
| 1.2287        | 2.0    | 1004 | 1.0543          | 73.3524 | 47.8329 | 0.0000 |
| 0.9543        | 3.0    | 1506 | 0.8523          | 80.3725 | 49.8555 | 0.0000 |
| 0.7365        | 4.0    | 2008 | 0.7315          | 79.6562 | 58.7602 | 0.0000 |
| 0.4663        | 5.0    | 2510 | 0.6675          | 57.3066 | 39.2960 | 0.0000 |
| 0.5042        | 6.0    | 3012 | 0.6423          | 54.8711 | 39.2960 | 0.0000 |
| 0.386         | 7.0    | 3514 | 0.6314          | 55.1576 | 38.2453 | 0.0000 |
| 0.2509        | 7.9850 | 4008 | 0.6333          | 56.7335 | 39.2960 | 0.0000 |


### Framework versions

- PEFT 0.14.0
- Transformers 4.48.3
- Pytorch 2.2.0
- Datasets 3.2.0
- Tokenizers 0.21.0