sitwala/whisper-large-v3-turbo-nchlt-sot

Whisper whisper-large-v3-turbo nchlt

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the dsfsi/multilingual-nchlt-dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.0352
• Wer: 1.9939

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: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 200
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1356	0.125	250	0.0931	7.1911
0.3632	0.25	500	0.1767	17.8252
0.079	1.0485	750	0.0476	3.2687
0.155	1.1735	1000	0.0730	6.8751
0.1758	1.2985	1250	0.1010	10.5688
0.0198	2.097	1500	0.0311	1.9503
0.1286	2.222	1750	0.0616	6.3521
0.0518	3.0205	2000	0.0352	1.9939

Framework versions

• Transformers 4.52.0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.4