arun100/whisper-base-th-2

Whisper Base Thai (2)

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

• Loss: 0.5628
• Wer: 53.6628

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
0.5011	35.0	500	0.5963	59.8868
0.3648	71.0	1000	0.5613	55.9542
0.2732	107.0	1500	0.5504	54.4585
0.2081	142.0	2000	0.5502	53.6705
0.1627	178.0	2500	0.5558	53.8273
0.133	214.0	3000	0.5628	53.6628
0.1112	249.0	3500	0.5696	54.0798
0.0973	285.0	4000	0.5749	53.9995
0.0906	321.0	4500	0.5783	54.1487
0.0874	357.0	5000	0.5793	54.2290

Framework versions

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