metadata
license: apache-2.0
base_model: infinitejoy/wav2vec2-large-xls-r-300m-sakha
tags:
- generated_from_trainer
datasets:
- common_voice_13_0
metrics:
- wer
model-index:
- name: wav2vec2-large-xls-r-300m-sakha-yaksaha
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: common_voice_13_0
type: common_voice_13_0
config: sah
split: test
args: sah
metrics:
- name: Wer
type: wer
value: 0.5587892898719441
wav2vec2-large-xls-r-300m-sakha-yaksaha
This model is a fine-tuned version of infinitejoy/wav2vec2-large-xls-r-300m-sakha on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5156
- Wer: 0.5588
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: 50
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 1.4 | 40 | 0.6200 | 0.6082 |
| 1.0278 | 2.81 | 80 | 0.5563 | 0.6026 |
| 0.648 | 4.21 | 120 | 0.5401 | 0.5763 |
| 0.5401 | 5.61 | 160 | 0.5424 | 0.5824 |
| 0.4972 | 7.02 | 200 | 0.5232 | 0.5638 |
| 0.4972 | 8.42 | 240 | 0.5022 | 0.5528 |
| 0.467 | 9.82 | 280 | 0.5156 | 0.5588 |
Framework versions
- Transformers 4.37.0.dev0
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0