napatswift/whisper-sm-th-7k

Whisper Tiny Thai

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0, lotus, google/fleurs th,None,th_th dataset. It achieves the following results on the evaluation set:

• Loss: 0.2262
• Wer: 14.69

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: 1e-05
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 7000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0473	0.14	1000	0.2459	21.6283
0.0253	1.07	2000	0.1970	17.2157
0.0181	2.0	3000	0.2017	17.8993
0.0088	2.15	4000	0.2148	16.8428
0.0055	3.07	5000	0.2166	15.8484
0.0048	4.0	6000	0.2261	16.0348
0.0026	4.15	7000	0.2262	15.5998

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2