metadata
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: training_45k
results: []
training_45k
This model is a fine-tuned version of Sameen53/cv_bn_bestModel_1 on the None dataset. It achieves the following results on the evaluation set:
- Loss: inf
- Wer: 0.1497
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: 5e-07
- 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: 7
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
0.2549 | 1.25 | 1500 | inf | 0.1495 |
0.2482 | 2.51 | 3000 | inf | 0.1496 |
0.2504 | 3.76 | 4500 | inf | 0.1498 |
0.2479 | 5.02 | 6000 | inf | 0.1495 |
0.2493 | 6.27 | 7500 | inf | 0.1497 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.1+cu118
- Datasets 2.13.1
- Tokenizers 0.12.1