whisper-base-malayalam-colab-CV17.0

This model is a fine-tuned version of openai/whisper-base on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4369
• Wer: 0.7676

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.15
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0335	1.5748	200	0.4105	0.9504
0.2301	3.1496	400	0.3121	0.8417
0.0954	4.7244	600	0.2964	0.8288
0.0442	6.2992	800	0.3350	0.7843
0.0217	7.8740	1000	0.3740	0.8133
0.0104	9.4488	1200	0.3858	0.7782
0.0048	11.0236	1400	0.4128	0.7747
0.002	12.5984	1600	0.4319	0.7747
0.0006	14.1732	1800	0.4324	0.7701
0.0002	15.7480	2000	0.4369	0.7676

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1