End of training
Browse files
README.md
CHANGED
|
@@ -6,10 +6,11 @@ base_model: openai/whisper-base
|
|
| 6 |
tags:
|
| 7 |
- generated_from_trainer
|
| 8 |
datasets:
|
|
|
|
|
|
|
| 9 |
- UBC-NLP/Casablanca
|
| 10 |
- google/fleurs
|
| 11 |
-
-
|
| 12 |
-
- mozilla-foundation/common_voice_17_0
|
| 13 |
metrics:
|
| 14 |
- wer
|
| 15 |
model-index:
|
|
@@ -20,15 +21,13 @@ model-index:
|
|
| 20 |
type: automatic-speech-recognition
|
| 21 |
dataset:
|
| 22 |
name: Common Voice 17.0
|
| 23 |
-
type:
|
| 24 |
-
config: ar
|
| 25 |
-
split: test
|
| 26 |
-
args: ar
|
| 27 |
metrics:
|
| 28 |
- name: Wer
|
| 29 |
type: wer
|
| 30 |
-
value:
|
| 31 |
---
|
|
|
|
| 32 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 33 |
should probably proofread and complete it, then remove this comment. -->
|
| 34 |
|
|
@@ -36,8 +35,9 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 36 |
|
| 37 |
This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Common Voice 17.0 dataset.
|
| 38 |
It achieves the following results on the evaluation set:
|
| 39 |
-
- Loss:
|
| 40 |
-
- Wer:
|
|
|
|
| 41 |
|
| 42 |
## Model description
|
| 43 |
|
|
@@ -58,22 +58,25 @@ More information needed
|
|
| 58 |
The following hyperparameters were used during training:
|
| 59 |
- learning_rate: 1e-05
|
| 60 |
- train_batch_size: 64
|
| 61 |
-
- eval_batch_size:
|
| 62 |
- seed: 42
|
| 63 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 64 |
- lr_scheduler_type: linear
|
| 65 |
-
-
|
| 66 |
-
- training_steps:
|
| 67 |
|
| 68 |
### Training results
|
| 69 |
|
| 70 |
-
| Training Loss | Epoch
|
| 71 |
-
|:-------------:|:-----:|:----:|:---------------:|:-------:|
|
| 72 |
-
| 0.
|
| 73 |
-
| 0.
|
| 74 |
-
| 0.
|
| 75 |
-
| 0.
|
| 76 |
-
| 0.
|
|
|
|
|
|
|
|
|
|
| 77 |
|
| 78 |
|
| 79 |
### Framework versions
|
|
@@ -82,14 +85,3 @@ The following hyperparameters were used during training:
|
|
| 82 |
- Pytorch 2.3.0+cu121
|
| 83 |
- Datasets 2.19.1
|
| 84 |
- Tokenizers 0.19.1
|
| 85 |
-
|
| 86 |
-
## Citation
|
| 87 |
-
|
| 88 |
-
```bibtex
|
| 89 |
-
@misc{deepdml/whisper-base-ar-mix-norm,
|
| 90 |
-
title={Fine-tuned Whisper base ASR model for speech recognition in Arabic},
|
| 91 |
-
author={Jimenez, David},
|
| 92 |
-
howpublished={\url{https://huggingface.co/deepdml/whisper-base-ar-mix-norm}},
|
| 93 |
-
year={2025}
|
| 94 |
-
}
|
| 95 |
-
```
|
|
|
|
| 6 |
tags:
|
| 7 |
- generated_from_trainer
|
| 8 |
datasets:
|
| 9 |
+
- ymoslem/MediaSpeech
|
| 10 |
+
- fixie-ai/common_voice_17_0
|
| 11 |
- UBC-NLP/Casablanca
|
| 12 |
- google/fleurs
|
| 13 |
+
- deepdml/Tunisian_MSA
|
|
|
|
| 14 |
metrics:
|
| 15 |
- wer
|
| 16 |
model-index:
|
|
|
|
| 21 |
type: automatic-speech-recognition
|
| 22 |
dataset:
|
| 23 |
name: Common Voice 17.0
|
| 24 |
+
type: ymoslem/MediaSpeech
|
|
|
|
|
|
|
|
|
|
| 25 |
metrics:
|
| 26 |
- name: Wer
|
| 27 |
type: wer
|
| 28 |
+
value: 45.035897247571654
|
| 29 |
---
|
| 30 |
+
|
| 31 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 32 |
should probably proofread and complete it, then remove this comment. -->
|
| 33 |
|
|
|
|
| 35 |
|
| 36 |
This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Common Voice 17.0 dataset.
|
| 37 |
It achieves the following results on the evaluation set:
|
| 38 |
+
- Loss: 0.5226
|
| 39 |
+
- Wer: 45.0359
|
| 40 |
+
- Cer: 14.7675
|
| 41 |
|
| 42 |
## Model description
|
| 43 |
|
|
|
|
| 58 |
The following hyperparameters were used during training:
|
| 59 |
- learning_rate: 1e-05
|
| 60 |
- train_batch_size: 64
|
| 61 |
+
- eval_batch_size: 64
|
| 62 |
- seed: 42
|
| 63 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 64 |
- lr_scheduler_type: linear
|
| 65 |
+
- lr_scheduler_warmup_ratio: 0.04
|
| 66 |
+
- training_steps: 8000
|
| 67 |
|
| 68 |
### Training results
|
| 69 |
|
| 70 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
| 71 |
+
|:-------------:|:------:|:----:|:---------------:|:-------:|:-------:|
|
| 72 |
+
| 0.6613 | 0.125 | 1000 | 0.6130 | 53.4217 | 18.1546 |
|
| 73 |
+
| 0.3701 | 0.25 | 2000 | 0.5791 | 50.5885 | 17.1229 |
|
| 74 |
+
| 0.1948 | 0.375 | 3000 | 0.5610 | 49.0241 | 16.4269 |
|
| 75 |
+
| 0.1173 | 0.5 | 4000 | 0.5481 | 47.8526 | 16.0193 |
|
| 76 |
+
| 0.089 | 0.625 | 5000 | 0.5337 | 47.0171 | 15.6262 |
|
| 77 |
+
| 0.0606 | 0.75 | 6000 | 0.5255 | 45.5317 | 15.0271 |
|
| 78 |
+
| 0.046 | 1.0442 | 7000 | 0.5235 | 45.1791 | 14.8553 |
|
| 79 |
+
| 0.0458 | 1.1692 | 8000 | 0.5226 | 45.0359 | 14.7675 |
|
| 80 |
|
| 81 |
|
| 82 |
### Framework versions
|
|
|
|
| 85 |
- Pytorch 2.3.0+cu121
|
| 86 |
- Datasets 2.19.1
|
| 87 |
- Tokenizers 0.19.1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|