This model is a fine-tuned version of facebook/wav2vec2-xls-r-2b on the OPENSLR_SLR66 - NA dataset. It achieves the following results on the evaluation set:
- Loss: 0.4253
- Wer: 0.5109
Evaluation metrics
| Metric | Split | Decode with LM | Value |
|---|---|---|---|
| WER | Train | No | |
| CER | Train | No | |
| WER | Test | No | |
| CER | Test | No | |
| WER | Train | Yes | |
| CER | Train | Yes | |
| WER | Test | Yes | |
| CER | Test | Yes |
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: 2e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 12
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- learning_rate: 3e-6
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 2000
- num_epochs: 150.0
- hidden_dropout: 0.15
- mixed_precision_training: Native AMP
Training results
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.11.0
- Downloads last month
- 3
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Dataset used to train chmanoj/xls-r-2B-te
Evaluation results
- Test WER on Open SLRself-reported0.510
- Test CER on Open SLRself-reported0.097