Update README.md
Browse files
README.md
CHANGED
|
@@ -32,9 +32,6 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 32 |
# Whisper Small Hebrew
|
| 33 |
|
| 34 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the ivrit-ai/whisper-training dataset.
|
| 35 |
-
It achieves the following results on the evaluation set:
|
| 36 |
-
- Loss: 0.4241
|
| 37 |
-
- Wer: 37.8652
|
| 38 |
|
| 39 |
## Model description
|
| 40 |
|
|
@@ -63,32 +60,6 @@ The following hyperparameters were used during training:
|
|
| 63 |
- training_steps: 10000
|
| 64 |
- mixed_precision_training: Native AMP
|
| 65 |
|
| 66 |
-
### Training results
|
| 67 |
-
|
| 68 |
-
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 69 |
-
|:-------------:|:-----:|:-----:|:---------------:|:-------:|
|
| 70 |
-
| 0.4946 | 0.13 | 500 | 0.4572 | 46.4463 |
|
| 71 |
-
| 0.4629 | 0.25 | 1000 | 0.4492 | 43.7663 |
|
| 72 |
-
| 0.4067 | 0.38 | 1500 | 0.4337 | 42.6317 |
|
| 73 |
-
| 0.3663 | 0.5 | 2000 | 0.3892 | 41.8427 |
|
| 74 |
-
| 0.3857 | 0.63 | 2500 | 0.4017 | 40.7473 |
|
| 75 |
-
| 0.3795 | 0.75 | 3000 | 0.4011 | 39.4823 |
|
| 76 |
-
| 0.368 | 0.88 | 3500 | 0.3967 | 39.9778 |
|
| 77 |
-
| 0.2353 | 1.01 | 4000 | 0.3801 | 38.3281 |
|
| 78 |
-
| 0.2405 | 1.13 | 4500 | 0.4062 | 41.5428 |
|
| 79 |
-
| 0.2512 | 1.26 | 5000 | 0.3975 | 38.6215 |
|
| 80 |
-
| 0.2433 | 1.38 | 5500 | 0.4035 | 38.5824 |
|
| 81 |
-
| 0.2368 | 1.51 | 6000 | 0.3983 | 37.8652 |
|
| 82 |
-
| 0.2592 | 1.63 | 6500 | 0.4184 | 39.1041 |
|
| 83 |
-
| 0.2629 | 1.76 | 7000 | 0.4000 | 39.5475 |
|
| 84 |
-
| 0.2318 | 1.88 | 7500 | 0.4012 | 39.1954 |
|
| 85 |
-
| 0.1658 | 2.01 | 8000 | 0.3941 | 39.1367 |
|
| 86 |
-
| 0.1546 | 2.14 | 8500 | 0.4226 | 39.8865 |
|
| 87 |
-
| 0.1665 | 2.26 | 9000 | 0.4295 | 40.9755 |
|
| 88 |
-
| 0.1642 | 2.39 | 9500 | 0.4314 | 41.1255 |
|
| 89 |
-
| 0.1694 | 2.51 | 10000 | 0.4241 | 40.6755 |
|
| 90 |
-
|
| 91 |
-
|
| 92 |
### Framework versions
|
| 93 |
|
| 94 |
- Transformers 4.40.0.dev0
|
|
|
|
| 32 |
# Whisper Small Hebrew
|
| 33 |
|
| 34 |
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the ivrit-ai/whisper-training dataset.
|
|
|
|
|
|
|
|
|
|
| 35 |
|
| 36 |
## Model description
|
| 37 |
|
|
|
|
| 60 |
- training_steps: 10000
|
| 61 |
- mixed_precision_training: Native AMP
|
| 62 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
### Framework versions
|
| 64 |
|
| 65 |
- Transformers 4.40.0.dev0
|