arun100/whisper-small-tl-1

Whisper Small Tagalog

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

• Loss: 0.6250
• Wer: 20.6653

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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0759	76.0	1000	0.5043	22.2622
0.0099	153.0	2000	0.5464	21.3653
0.0043	230.0	3000	0.5707	21.2215
0.0024	307.0	4000	0.5909	20.9377
0.0015	384.0	5000	0.6090	20.6728
0.001	461.0	6000	0.6250	20.6653
0.0007	538.0	7000	0.6395	20.8582
0.0005	615.0	8000	0.6519	20.9415
0.0004	692.0	9000	0.6613	20.9112
0.0004	769.0	10000	0.6653	20.9377

Framework versions

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