End of training
Browse files
README.md
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ru
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
base_model: openai/whisper-large-v2
|
| 6 |
+
tags:
|
| 7 |
+
- generated_from_trainer
|
| 8 |
+
metrics:
|
| 9 |
+
- wer
|
| 10 |
+
library_name: peft
|
| 11 |
+
model-index:
|
| 12 |
+
- name: 'Whisper Large Ru ORD 0.9 Peft PEFT 4-bit Q DoRA - Mizoru '
|
| 13 |
+
results: []
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 17 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 18 |
+
|
| 19 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/mizoru/ORD/runs/jmqewt8c)
|
| 20 |
+
# Whisper Large Ru ORD 0.9 Peft PEFT 4-bit Q DoRA - Mizoru
|
| 21 |
+
|
| 22 |
+
This model is a fine-tuned version of [openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) on the ORD_0.9synth dataset.
|
| 23 |
+
It achieves the following results on the evaluation set:
|
| 24 |
+
- Loss: 0.9568
|
| 25 |
+
- Wer: 40.6051
|
| 26 |
+
- Cer: 24.1709
|
| 27 |
+
- Clean Wer: 30.6324
|
| 28 |
+
- Clean Cer: 18.3212
|
| 29 |
+
|
| 30 |
+
## Model description
|
| 31 |
+
|
| 32 |
+
More information needed
|
| 33 |
+
|
| 34 |
+
## Intended uses & limitations
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Training and evaluation data
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training procedure
|
| 43 |
+
|
| 44 |
+
### Training hyperparameters
|
| 45 |
+
|
| 46 |
+
The following hyperparameters were used during training:
|
| 47 |
+
- learning_rate: 0.001
|
| 48 |
+
- train_batch_size: 16
|
| 49 |
+
- eval_batch_size: 16
|
| 50 |
+
- seed: 42
|
| 51 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 52 |
+
- lr_scheduler_type: linear
|
| 53 |
+
- lr_scheduler_warmup_steps: 50
|
| 54 |
+
- num_epochs: 4
|
| 55 |
+
- mixed_precision_training: Native AMP
|
| 56 |
+
|
| 57 |
+
### Training results
|
| 58 |
+
|
| 59 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | Clean Wer | Clean Cer |
|
| 60 |
+
|:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:---------:|:---------:|
|
| 61 |
+
| 1.2498 | 1.0 | 1162 | 1.0305 | 44.9736 | 26.2819 | 34.8983 | 20.4239 |
|
| 62 |
+
| 1.1376 | 2.0 | 2324 | 0.9974 | 45.1705 | 26.6383 | 33.6258 | 20.2044 |
|
| 63 |
+
| 1.0073 | 3.0 | 3486 | 0.9692 | 41.3758 | 24.8708 | 31.3579 | 19.0546 |
|
| 64 |
+
| 0.9389 | 4.0 | 4648 | 0.9568 | 40.6051 | 24.1709 | 30.6324 | 18.3212 |
|
| 65 |
+
|
| 66 |
+
|
| 67 |
+
### Framework versions
|
| 68 |
+
|
| 69 |
+
- PEFT 0.12.0
|
| 70 |
+
- Transformers 4.41.0.dev0
|
| 71 |
+
- Pytorch 2.3.1
|
| 72 |
+
- Datasets 3.2.0
|
| 73 |
+
- Tokenizers 0.19.1
|