arun100/whisper-base-vi-2

Whisper Base Vietnamese

This model is a fine-tuned version of arun100/whisper-base-vi-1 on the google/fleurs vi_vn dataset. It achieves the following results on the evaluation set:

• Loss: 0.6949
• Wer: 31.0338

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5823	43.0	500	0.7964	37.8978
0.3312	86.0	1000	0.6997	33.7125
0.2009	130.0	1500	0.6784	32.7479
0.1271	173.0	2000	0.6760	31.9985
0.0815	217.0	2500	0.6799	31.3028
0.0561	260.0	3000	0.6851	31.2337
0.0438	304.0	3500	0.6896	31.7256
0.0367	347.0	4000	0.6928	31.5949
0.0331	391.0	4500	0.6949	31.0338
0.0317	434.0	5000	0.6957	31.0453

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0