Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- ja
|
| 5 |
+
metrics:
|
| 6 |
+
- cer
|
| 7 |
+
- wer
|
| 8 |
+
base_model:
|
| 9 |
+
- openai/whisper-medium
|
| 10 |
+
tags:
|
| 11 |
+
- ctranslate2
|
| 12 |
+
- faster-whisper
|
| 13 |
+
- whisper
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
> **This repository contains the CTranslate2 export of the fine-tuned model.**
|
| 17 |
+
>
|
| 18 |
+
> • Base Transformers model: [drepic/whisper-medium-jp](https://huggingface.co/drepic/whisper-medium-jp)
|
| 19 |
+
> • Use with `faster-whisper`:
|
| 20 |
+
>
|
| 21 |
+
> ```python
|
| 22 |
+
> from faster_whisper import WhisperModel
|
| 23 |
+
> model = WhisperModel("drepic/whisper-medium-jp-ct2", device="cuda", compute_type="float16")
|
| 24 |
+
> ```
|
| 25 |
+
|
| 26 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 27 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 28 |
+
|
| 29 |
+
# whisper-medium-jp
|
| 30 |
+
|
| 31 |
+
This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on an unknown dataset.
|
| 32 |
+
It achieves the following results on the evaluation set:
|
| 33 |
+
- Loss: 0.4828
|
| 34 |
+
- Wer: 0.2254
|
| 35 |
+
- Cer: 0.2254
|
| 36 |
+
|
| 37 |
+
## Model description
|
| 38 |
+
|
| 39 |
+
Better suited for transcribing japanese youtube content.
|
| 40 |
+
|
| 41 |
+
## Intended uses & limitations
|
| 42 |
+
|
| 43 |
+
More information needed
|
| 44 |
+
|
| 45 |
+
## Training and evaluation data
|
| 46 |
+
|
| 47 |
+
More information needed
|
| 48 |
+
|
| 49 |
+
## Training procedure
|
| 50 |
+
|
| 51 |
+
### Training hyperparameters
|
| 52 |
+
|
| 53 |
+
The following hyperparameters were used during training:
|
| 54 |
+
- learning_rate: 4e-06
|
| 55 |
+
- train_batch_size: 4
|
| 56 |
+
- eval_batch_size: 2
|
| 57 |
+
- seed: 42
|
| 58 |
+
- distributed_type: multi-GPU
|
| 59 |
+
- num_devices: 2
|
| 60 |
+
- gradient_accumulation_steps: 2
|
| 61 |
+
- total_train_batch_size: 16
|
| 62 |
+
- total_eval_batch_size: 4
|
| 63 |
+
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
|
| 64 |
+
- lr_scheduler_type: linear
|
| 65 |
+
- lr_scheduler_warmup_steps: 400
|
| 66 |
+
- num_epochs: 15
|
| 67 |
+
- mixed_precision_training: Native AMP
|
| 68 |
+
|
| 69 |
+
### Training results
|
| 70 |
+
|
| 71 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
| 72 |
+
|:-------------:|:-----:|:-----:|:---------------:|:------:|:------:|
|
| 73 |
+
| 0.5341 | 1.0 | 7155 | 0.5321 | 0.2416 | 0.2416 |
|
| 74 |
+
| 0.5023 | 2.0 | 14310 | 0.5143 | 0.2369 | 0.2369 |
|
| 75 |
+
| 0.499 | 3.0 | 21465 | 0.5063 | 0.2337 | 0.2337 |
|
| 76 |
+
| 0.4773 | 4.0 | 28620 | 0.5010 | 0.2310 | 0.2310 |
|
| 77 |
+
| 0.4775 | 5.0 | 35775 | 0.4944 | 0.2289 | 0.2289 |
|
| 78 |
+
| 0.4709 | 6.0 | 42930 | 0.4886 | 0.2288 | 0.2288 |
|
| 79 |
+
| 0.4907 | 7.0 | 50085 | 0.4870 | 0.2271 | 0.2271 |
|
| 80 |
+
| 0.4855 | 8.0 | 57240 | 0.4868 | 0.2261 | 0.2261 |
|
| 81 |
+
| 0.4487 | 9.0 | 64395 | 0.4828 | 0.2254 | 0.2254 |
|
| 82 |
+
|
| 83 |
+
|
| 84 |
+
### Framework versions
|
| 85 |
+
|
| 86 |
+
- Transformers 4.56.1
|
| 87 |
+
- Pytorch 2.8.0+cu128
|
| 88 |
+
- Datasets 4.0.0
|
| 89 |
+
- Tokenizers 0.22.0
|