Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,70 @@
|
|
1 |
---
|
2 |
-
license:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
license: apache-2.0
|
3 |
+
tags:
|
4 |
+
- whisper-event
|
5 |
+
- generated_from_trainer
|
6 |
+
datasets:
|
7 |
+
- Ranjit/or_in_dataset
|
8 |
+
metrics:
|
9 |
+
- wer
|
10 |
+
model-index:
|
11 |
+
- name: Whisper Small Odia 10k steps
|
12 |
+
results:
|
13 |
+
- task:
|
14 |
+
name: Automatic Speech Recognition
|
15 |
+
type: automatic-speech-recognition
|
16 |
+
dataset:
|
17 |
+
name: Ranjit/or_in_dataset
|
18 |
+
type: Ranjit/or_in_dataset
|
19 |
+
split: test
|
20 |
+
metrics:
|
21 |
+
- name: Wer
|
22 |
+
type: wer
|
23 |
+
value: 19.772043934466467
|
24 |
---
|
25 |
+
|
26 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
27 |
+
should probably proofread and complete it, then remove this comment. -->
|
28 |
+
|
29 |
+
# Whisper Small Odia v6.0
|
30 |
+
|
31 |
+
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the [Ranjit/or_in_dataset](https://huggingface.co/datasets/Ranjit/or_in_dataset) dataset.
|
32 |
+
It achieves the following results on the evaluation set:
|
33 |
+
- Loss: 0.2806
|
34 |
+
- Wer: 19.7720
|
35 |
+
|
36 |
+
### Training hyperparameters
|
37 |
+
|
38 |
+
The following hyperparameters were used during training:
|
39 |
+
- learning_rate: 1e-05
|
40 |
+
- train_batch_size: 32
|
41 |
+
- eval_batch_size: 32
|
42 |
+
- seed: 42
|
43 |
+
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-06
|
44 |
+
- lr_scheduler_type: linear
|
45 |
+
- lr_scheduler_warmup_steps: 200
|
46 |
+
- training_steps: 10000
|
47 |
+
- mixed_precision_training: Native AMP
|
48 |
+
|
49 |
+
### Training results
|
50 |
+
|
51 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
52 |
+
|:-------------:|:-----:|:-----:|:---------------:|:-------:|
|
53 |
+
| 0.0384 | 0.49 | 1000 | 0.1349 | 40.3740 |
|
54 |
+
| 0.0175 | 0.98 | 2000 | 0.1601 | 22.6468 |
|
55 |
+
| 0.0091 | 1.46 | 3000 | 0.1817 | 23.1515 |
|
56 |
+
| 0.0082 | 1.95 | 4000 | 0.2125 | 23.9139 |
|
57 |
+
| 0.0048 | 2.44 | 5000 | 0.2110 | 20.2522 |
|
58 |
+
| 0.0037 | 2.93 | 6000 | 0.2270 | 21.4855 |
|
59 |
+
| 0.0017 | 3.42 | 7000 | 0.2534 | 20.2399 |
|
60 |
+
| 0.0018 | 3.9 | 8000 | 0.2706 | 20.7277 |
|
61 |
+
| 0.0005 | 4.39 | 9000 | 0.2806 | 19.7720 |
|
62 |
+
| 0.0006 | 4.88 | 10000 | 0.2929 | 20.9747 |
|
63 |
+
|
64 |
+
|
65 |
+
### Framework versions
|
66 |
+
|
67 |
+
- Transformers 4.28.0.dev0
|
68 |
+
- Pytorch 1.11.0+cu113
|
69 |
+
- Datasets 2.10.1
|
70 |
+
- Tokenizers 0.13.2
|