checkpoints2

This model is a fine-tuned version of ttop324/wav2vec2-live-japanese on the extracted and cleaned transcript of Holo No Graffiti dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Evaluated from common_voice | train | ja :

  • WER: 32.940524
  • CER: 15.251746

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0003
  • train_batch_size: 3
  • eval_batch_size: 2
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 6
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 50
  • mixed_precision_training: Native AMP

Framework versions

  • Transformers 4.17.0
  • Pytorch 1.11.0+cu113
  • Datasets 1.18.3
  • Tokenizers 0.12.1
Downloads last month
10
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.