End of training
Browse files
README.md
CHANGED
|
@@ -22,8 +22,9 @@ model-index:
|
|
| 22 |
metrics:
|
| 23 |
- name: Wer
|
| 24 |
type: wer
|
| 25 |
-
value:
|
| 26 |
---
|
|
|
|
| 27 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 28 |
should probably proofread and complete it, then remove this comment. -->
|
| 29 |
|
|
@@ -31,8 +32,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 31 |
|
| 32 |
This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the Common Voice 17.0 dataset.
|
| 33 |
It achieves the following results on the evaluation set:
|
| 34 |
-
- Loss:
|
| 35 |
-
- Wer:
|
| 36 |
|
| 37 |
## Model description
|
| 38 |
|
|
@@ -57,18 +58,14 @@ The following hyperparameters were used during training:
|
|
| 57 |
- seed: 42
|
| 58 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 59 |
- lr_scheduler_type: linear
|
| 60 |
-
-
|
| 61 |
-
- training_steps:
|
| 62 |
|
| 63 |
### Training results
|
| 64 |
|
| 65 |
-
| Training Loss | Epoch | Step
|
| 66 |
-
|
| 67 |
-
| 0.
|
| 68 |
-
| 0.0288 | 0.4 | 2000 | 0.6834 | 1.5808 |
|
| 69 |
-
| 0.0225 | 0.6 | 3000 | 0.7354 | 11.6002 |
|
| 70 |
-
| 0.0082 | 0.8 | 4000 | 0.9359 | 41.3866 |
|
| 71 |
-
| 0.0438 | 1.0 | 5000 | 0.9379 | 39.6117 |
|
| 72 |
|
| 73 |
|
| 74 |
### Framework versions
|
|
@@ -77,14 +74,3 @@ The following hyperparameters were used during training:
|
|
| 77 |
- Pytorch 2.3.0+cu121
|
| 78 |
- Datasets 2.19.1
|
| 79 |
- Tokenizers 0.19.1
|
| 80 |
-
|
| 81 |
-
## Citation
|
| 82 |
-
|
| 83 |
-
```bibtex
|
| 84 |
-
@misc{deepdml/whisper-medium-ar-quran-mix,
|
| 85 |
-
title={Fine-tuned Whisper medium ASR model for speech recognition in Arabic},
|
| 86 |
-
author={Jimenez, David},
|
| 87 |
-
howpublished={\url{https://huggingface.co/deepdml/whisper-medium-ar-quran-mix}},
|
| 88 |
-
year={2025}
|
| 89 |
-
}
|
| 90 |
-
```
|
|
|
|
| 22 |
metrics:
|
| 23 |
- name: Wer
|
| 24 |
type: wer
|
| 25 |
+
value: 0.15533980582524273
|
| 26 |
---
|
| 27 |
+
|
| 28 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 29 |
should probably proofread and complete it, then remove this comment. -->
|
| 30 |
|
|
|
|
| 32 |
|
| 33 |
This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the Common Voice 17.0 dataset.
|
| 34 |
It achieves the following results on the evaluation set:
|
| 35 |
+
- Loss: 1.0757
|
| 36 |
+
- Wer: 0.1553
|
| 37 |
|
| 38 |
## Model description
|
| 39 |
|
|
|
|
| 58 |
- seed: 42
|
| 59 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 60 |
- lr_scheduler_type: linear
|
| 61 |
+
- lr_scheduler_warmup_ratio: 0.04
|
| 62 |
+
- training_steps: 15000
|
| 63 |
|
| 64 |
### Training results
|
| 65 |
|
| 66 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 67 |
+
|:-------------:|:-----:|:-----:|:---------------:|:------:|
|
| 68 |
+
| 0.0377 | 1.0 | 15000 | 1.0757 | 0.1553 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
|
|
|
| 74 |
- Pytorch 2.3.0+cu121
|
| 75 |
- Datasets 2.19.1
|
| 76 |
- Tokenizers 0.19.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|