Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
language:
|
| 4 |
+
- ps
|
| 5 |
+
license: apache-2.0
|
| 6 |
+
base_model: openai/whisper-large
|
| 7 |
+
tags:
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
metrics:
|
| 10 |
+
- wer
|
| 11 |
+
model-index:
|
| 12 |
+
- name: Whisper large Ps - ZFA
|
| 13 |
+
results: []
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# Whisper large Ps - ZFA
|
| 17 |
+
|
| 18 |
+
This model is a fine-tuned version of [openai/whisper-large](https://huggingface.co/openai/whisper-large) on a custom Pashto speech dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.7591
|
| 21 |
+
- Wer: 24.1861
|
| 22 |
+
|
| 23 |
+
## Model description
|
| 24 |
+
|
| 25 |
+
More information needed
|
| 26 |
+
|
| 27 |
+
## Intended uses & limitations
|
| 28 |
+
|
| 29 |
+
More information needed
|
| 30 |
+
|
| 31 |
+
## Training and evaluation data
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Training procedure
|
| 36 |
+
|
| 37 |
+
### Training hyperparameters
|
| 38 |
+
|
| 39 |
+
The following hyperparameters were used during training:
|
| 40 |
+
- learning_rate: 2e-05
|
| 41 |
+
- train_batch_size: 1
|
| 42 |
+
- eval_batch_size: 1
|
| 43 |
+
- seed: 42
|
| 44 |
+
- gradient_accumulation_steps: 32
|
| 45 |
+
- total_train_batch_size: 32
|
| 46 |
+
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
|
| 47 |
+
- lr_scheduler_type: SchedulerType.LINEAR
|
| 48 |
+
- lr_scheduler_warmup_steps: 500
|
| 49 |
+
- training_steps: 3000
|
| 50 |
+
- mixed_precision_training: Native AMP
|
| 51 |
+
|
| 52 |
+
### Training results
|
| 53 |
+
|
| 54 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 55 |
+
|:-------------:|:-----:|:----:|:---------------:|:---:|
|
| 56 |
+
| 14.2763 | 12.8298 | 500 | 0.639 | 28.7915 |
|
| 57 |
+
| 0.574 | 25.6483 | 1000 | 0.6891 | 25.7107 |
|
| 58 |
+
| 0.0377 | 38.4668 | 1500 | 0.7166 | 24.8849 |
|
| 59 |
+
| 0.0033 | 51.2853 | 2000 | 0.7313 | 24.3132 |
|
| 60 |
+
| 0.0023 | 64.1037 | 2500 | 0.7548 | 24.202 |
|
| 61 |
+
| 0.0019 | 76.9335 | 3000 | 0.7591 | 24.1861 |
|
| 62 |
+
|
| 63 |
+
### Framework versions
|
| 64 |
+
|
| 65 |
+
- Transformers 5.0.0
|
| 66 |
+
- Pytorch 2.10.0+cpu
|
| 67 |
+
- Datasets 4.0.0
|
| 68 |
+
- Tokenizers 0.22.2
|