Hugging Face's logo

cahya
/
output

Automatic Speech Recognition
Transformers
PyTorch
Turkish
wav2vec2
common_voice
generated_from_trainer
Inference Endpoints
Model card Files Community
output / README.md
cahya's picture
cahya
update model card README.md
593378d
preview code
|
raw history blame contribute delete
No virus
1.34 kB
metadata
language:
  - tr
license: apache-2.0
tags:
  - automatic-speech-recognition
  - common_voice
  - generated_from_trainer
datasets:
  - common_voice
model-index:
  - name: output
    results: []

output

This model is a fine-tuned version of cahya/wav2vec2-base-turkish-artificial-cv on the COMMON_VOICE - TR dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1822
  • Wer: 0.1423

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: 7.5e-07
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 2000
  • num_epochs: 1.0

Training results

Framework versions

  • Transformers 4.17.0.dev0
  • Pytorch 1.10.1+cu102
  • Datasets 1.18.2
  • Tokenizers 0.10.3