update model card README.md
Browse files
README.md
ADDED
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- tr
|
4 |
+
license: apache-2.0
|
5 |
+
tags:
|
6 |
+
- hf-asr-leaderboard
|
7 |
+
- generated_from_trainer
|
8 |
+
metrics:
|
9 |
+
- wer
|
10 |
+
model-index:
|
11 |
+
- name: base Turkish Whisper (bTW)
|
12 |
+
results: []
|
13 |
+
---
|
14 |
+
|
15 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
16 |
+
should probably proofread and complete it, then remove this comment. -->
|
17 |
+
|
18 |
+
# base Turkish Whisper (bTW)
|
19 |
+
|
20 |
+
This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Ermetal Meetings dataset.
|
21 |
+
It achieves the following results on the evaluation set:
|
22 |
+
- Loss: 1.1836
|
23 |
+
- Wer: 1.7109
|
24 |
+
- Cer: 1.2860
|
25 |
+
|
26 |
+
## Model description
|
27 |
+
|
28 |
+
More information needed
|
29 |
+
|
30 |
+
## Intended uses & limitations
|
31 |
+
|
32 |
+
More information needed
|
33 |
+
|
34 |
+
## Training and evaluation data
|
35 |
+
|
36 |
+
More information needed
|
37 |
+
|
38 |
+
## Training procedure
|
39 |
+
|
40 |
+
### Training hyperparameters
|
41 |
+
|
42 |
+
The following hyperparameters were used during training:
|
43 |
+
- learning_rate: 1e-05
|
44 |
+
- train_batch_size: 16
|
45 |
+
- eval_batch_size: 16
|
46 |
+
- seed: 42
|
47 |
+
- gradient_accumulation_steps: 4
|
48 |
+
- total_train_batch_size: 64
|
49 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
50 |
+
- lr_scheduler_type: linear
|
51 |
+
- lr_scheduler_warmup_steps: 500
|
52 |
+
- training_steps: 1000
|
53 |
+
- mixed_precision_training: Native AMP
|
54 |
+
|
55 |
+
### Training results
|
56 |
+
|
57 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
58 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
|
59 |
+
| 1.7878 | 4.74 | 100 | 1.4516 | 0.8560 | 0.5525 |
|
60 |
+
| 0.6701 | 9.51 | 200 | 0.9194 | 0.8543 | 0.6112 |
|
61 |
+
| 0.3364 | 14.28 | 300 | 0.8871 | 0.7415 | 0.4992 |
|
62 |
+
| 0.1228 | 19.05 | 400 | 0.9671 | 0.9052 | 0.6678 |
|
63 |
+
| 0.0355 | 23.78 | 500 | 1.0515 | 0.8961 | 0.6208 |
|
64 |
+
| 0.0148 | 28.55 | 600 | 1.0684 | 0.6644 | 0.3694 |
|
65 |
+
| 0.0056 | 33.32 | 700 | 1.1488 | 1.3315 | 0.8732 |
|
66 |
+
| 0.0041 | 38.09 | 800 | 1.1700 | 1.7415 | 1.1934 |
|
67 |
+
| 0.0034 | 42.83 | 900 | 1.1801 | 1.7745 | 1.2643 |
|
68 |
+
| 0.0032 | 47.6 | 1000 | 1.1836 | 1.7109 | 1.2860 |
|
69 |
+
|
70 |
+
|
71 |
+
### Framework versions
|
72 |
+
|
73 |
+
- Transformers 4.26.0
|
74 |
+
- Pytorch 1.12.0+cu102
|
75 |
+
- Datasets 2.9.0
|
76 |
+
- Tokenizers 0.13.2
|