|
--- |
|
license: apache-2.0 |
|
base_model: facebook/wav2vec2-large-xlsr-53 |
|
tags: |
|
- generated_from_trainer |
|
datasets: |
|
- common_voice_17_0 |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: xls-r-300m-hbs-phoneme-unfrozen-batch16 |
|
results: |
|
- task: |
|
name: Automatic Speech Recognition |
|
type: automatic-speech-recognition |
|
dataset: |
|
name: common_voice_17_0 |
|
type: common_voice_17_0 |
|
config: hsb |
|
split: test |
|
args: hsb |
|
metrics: |
|
- name: Wer |
|
type: wer |
|
value: 0.5337394564198688 |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
[<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/badr-nlp/xlsr-continual-finetuning/runs/7duhfamy) |
|
# xls-r-300m-hbs-phoneme-unfrozen-batch16 |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the common_voice_17_0 dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.9205 |
|
- Wer: 0.5337 |
|
- Cer: 0.1244 |
|
|
|
## Model description |
|
|
|
More information needed |
|
|
|
## Intended uses & limitations |
|
|
|
More information needed |
|
|
|
## Training and evaluation data |
|
|
|
More information needed |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0003 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 32 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_steps: 500 |
|
- num_epochs: 100 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |
|
|:-------------:|:-------:|:----:|:---------------:|:------:|:------:| |
|
| 4.0877 | 3.2258 | 100 | 3.7799 | 1.0 | 1.0 | |
|
| 3.2643 | 6.4516 | 200 | 3.2338 | 1.0 | 1.0 | |
|
| 3.2182 | 9.6774 | 300 | 3.1963 | 1.0 | 1.0 | |
|
| 0.8009 | 12.9032 | 400 | 0.9289 | 0.8240 | 0.2193 | |
|
| 0.2664 | 16.1290 | 500 | 0.8523 | 0.7381 | 0.1855 | |
|
| 0.1359 | 19.3548 | 600 | 0.8465 | 0.6757 | 0.1676 | |
|
| 0.1022 | 22.5806 | 700 | 0.8537 | 0.6603 | 0.1656 | |
|
| 0.0641 | 25.8065 | 800 | 0.8821 | 0.6664 | 0.1620 | |
|
| 0.0565 | 29.0323 | 900 | 0.9185 | 0.6610 | 0.1608 | |
|
| 0.068 | 32.2581 | 1000 | 0.8839 | 0.6286 | 0.1513 | |
|
| 0.0556 | 35.4839 | 1100 | 0.8898 | 0.6125 | 0.1479 | |
|
| 0.0457 | 38.7097 | 1200 | 0.8840 | 0.6204 | 0.1448 | |
|
| 0.0439 | 41.9355 | 1300 | 0.9207 | 0.6249 | 0.1490 | |
|
| 0.0296 | 45.1613 | 1400 | 0.9572 | 0.6246 | 0.1510 | |
|
| 0.0461 | 48.3871 | 1500 | 0.8875 | 0.5918 | 0.1395 | |
|
| 0.0419 | 51.6129 | 1600 | 0.8967 | 0.5846 | 0.1384 | |
|
| 0.0333 | 54.8387 | 1700 | 0.9827 | 0.5951 | 0.1420 | |
|
| 0.0318 | 58.0645 | 1800 | 0.9055 | 0.5733 | 0.1364 | |
|
| 0.0238 | 61.2903 | 1900 | 0.9497 | 0.5696 | 0.1363 | |
|
| 0.0257 | 64.5161 | 2000 | 0.9268 | 0.5590 | 0.1330 | |
|
| 0.0266 | 67.7419 | 2100 | 0.9374 | 0.5703 | 0.1351 | |
|
| 0.0292 | 70.9677 | 2200 | 0.9304 | 0.5754 | 0.1352 | |
|
| 0.0288 | 74.1935 | 2300 | 0.9419 | 0.5649 | 0.1334 | |
|
| 0.0125 | 77.4194 | 2400 | 0.9625 | 0.5581 | 0.1335 | |
|
| 0.0241 | 80.6452 | 2500 | 0.9449 | 0.5569 | 0.1313 | |
|
| 0.0217 | 83.8710 | 2600 | 0.9315 | 0.5504 | 0.1292 | |
|
| 0.0136 | 87.0968 | 2700 | 0.9079 | 0.5373 | 0.1257 | |
|
| 0.0203 | 90.3226 | 2800 | 0.8935 | 0.5373 | 0.1241 | |
|
| 0.0166 | 93.5484 | 2900 | 0.9169 | 0.5354 | 0.1239 | |
|
| 0.0114 | 96.7742 | 3000 | 0.9245 | 0.5323 | 0.1240 | |
|
| 0.011 | 100.0 | 3100 | 0.9205 | 0.5337 | 0.1244 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.42.0.dev0 |
|
- Pytorch 2.3.1+cu121 |
|
- Datasets 2.19.2 |
|
- Tokenizers 0.19.1 |
|
|
