|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
base_model: facebook/wav2vec2-xls-r-300m |
|
datasets: |
|
- common_voice_17_0 |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: xls-r-300m-hbs-phoneme-unfrozen-batch16 |
|
results: |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Automatic Speech Recognition |
|
dataset: |
|
name: common_voice_17_0 |
|
type: common_voice_17_0 |
|
config: hsb |
|
split: test |
|
args: hsb |
|
metrics: |
|
- type: wer |
|
value: 0.4111996251171509 |
|
name: Wer |
|
--- |
|
|
|
<!-- 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/7invqf4p) |
|
# xls-r-300m-hbs-phoneme-unfrozen-batch16 |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice_17_0 dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.7105 |
|
- Wer: 0.4112 |
|
- Cer: 0.0948 |
|
|
|
## 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 | |
|
|:-------------:|:-------:|:----:|:---------------:|:------:|:------:| |
|
| 3.6184 | 3.2258 | 100 | 3.4215 | 1.0 | 1.0 | |
|
| 3.2927 | 6.4516 | 200 | 3.2247 | 1.0 | 1.0 | |
|
| 3.2291 | 9.6774 | 300 | 3.2021 | 1.0 | 1.0000 | |
|
| 1.4844 | 12.9032 | 400 | 1.3507 | 0.9857 | 0.2837 | |
|
| 0.4136 | 16.1290 | 500 | 0.6982 | 0.6567 | 0.1608 | |
|
| 0.2346 | 19.3548 | 600 | 0.6496 | 0.5956 | 0.1466 | |
|
| 0.1401 | 22.5806 | 700 | 0.6680 | 0.5565 | 0.1314 | |
|
| 0.1535 | 25.8065 | 800 | 0.6597 | 0.5026 | 0.1190 | |
|
| 0.1165 | 29.0323 | 900 | 0.7085 | 0.5112 | 0.1224 | |
|
| 0.076 | 32.2581 | 1000 | 0.7359 | 0.5026 | 0.1195 | |
|
| 0.083 | 35.4839 | 1100 | 0.7144 | 0.4991 | 0.1205 | |
|
| 0.0985 | 38.7097 | 1200 | 0.6907 | 0.4756 | 0.1120 | |
|
| 0.052 | 41.9355 | 1300 | 0.6806 | 0.4700 | 0.1105 | |
|
| 0.0347 | 45.1613 | 1400 | 0.7097 | 0.4588 | 0.1091 | |
|
| 0.0432 | 48.3871 | 1500 | 0.7086 | 0.4649 | 0.1093 | |
|
| 0.0626 | 51.6129 | 1600 | 0.6947 | 0.4393 | 0.1029 | |
|
| 0.0474 | 54.8387 | 1700 | 0.6915 | 0.4468 | 0.1058 | |
|
| 0.057 | 58.0645 | 1800 | 0.7068 | 0.4358 | 0.1020 | |
|
| 0.0373 | 61.2903 | 1900 | 0.7140 | 0.4419 | 0.1037 | |
|
| 0.0994 | 64.5161 | 2000 | 0.6966 | 0.4208 | 0.0987 | |
|
| 0.0503 | 67.7419 | 2100 | 0.6997 | 0.4306 | 0.0988 | |
|
| 0.0418 | 70.9677 | 2200 | 0.7105 | 0.4353 | 0.1006 | |
|
| 0.036 | 74.1935 | 2300 | 0.7320 | 0.4356 | 0.1024 | |
|
| 0.0171 | 77.4194 | 2400 | 0.7132 | 0.4257 | 0.0994 | |
|
| 0.0234 | 80.6452 | 2500 | 0.7059 | 0.4171 | 0.0967 | |
|
| 0.0335 | 83.8710 | 2600 | 0.7449 | 0.4140 | 0.0973 | |
|
| 0.0288 | 87.0968 | 2700 | 0.7028 | 0.4157 | 0.0964 | |
|
| 0.0344 | 90.3226 | 2800 | 0.7181 | 0.4112 | 0.0960 | |
|
| 0.0298 | 93.5484 | 2900 | 0.7150 | 0.4105 | 0.0951 | |
|
| 0.0532 | 96.7742 | 3000 | 0.7164 | 0.4119 | 0.0950 | |
|
| 0.0058 | 100.0 | 3100 | 0.7105 | 0.4112 | 0.0948 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.42.0.dev0 |
|
- Pytorch 2.3.1+cu121 |
|
- Datasets 2.19.2 |
|
- Tokenizers 0.19.1 |
|
|