arun100/whisper-base-tl-1

Whisper Base Tagalog

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

• Loss: 0.7222
• Wer: 30.8106

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-06
• 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
0.5804	38.0	500	0.7836	36.0478
0.1934	76.0	1000	0.6861	31.5220
0.0589	115.0	1500	0.7040	32.4415
0.0251	153.0	2000	0.7222	30.8106
0.0154	192.0	2500	0.7362	31.3593
0.0109	230.0	3000	0.7470	31.7604
0.0085	269.0	3500	0.7562	31.7112
0.0071	307.0	4000	0.7630	31.9874
0.0064	346.0	4500	0.7675	32.0064
0.0061	384.0	5000	0.7692	32.0669

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0