wav2vec2-large-xls-r-300m-ja-colab-3

This model is a fine-tuned version of pinot/wav2vec2-large-xls-r-300m-ja-colab-2 on the common_voice_10_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.2696
• Wer: 0.2299

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	637	1.4666	0.2862
No log	2.0	1274	1.4405	0.2866
No log	3.0	1911	1.4162	0.2762
No log	4.0	2548	1.4128	0.2709
0.2814	5.0	3185	1.3927	0.2613
0.2814	6.0	3822	1.3629	0.2536
0.2814	7.0	4459	1.3349	0.2429
0.2814	8.0	5096	1.3116	0.2356
0.1624	9.0	5733	1.2774	0.2307
0.1624	10.0	6370	1.2696	0.2299

Framework versions

• Transformers 4.21.2
• Pytorch 1.10.0+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1