whisper-large-v3-turbo-Tamil-Version2

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

• Loss: 0.3283
• Wer: 51.5481

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3703	5.8309	2000	0.3876	58.7349
0.3127	11.6618	4000	0.3546	55.2184
0.281	17.4927	6000	0.3441	53.8958
0.2749	23.3236	8000	0.3368	53.2807
0.2641	29.1545	10000	0.3326	53.0244
0.2569	34.9854	12000	0.3292	52.6451
0.236	40.8163	14000	0.3284	52.0094
0.2314	46.6472	16000	0.3281	51.9069
0.2302	52.4781	18000	0.3275	51.9787
0.2259	58.3090	20000	0.3283	51.5481

Framework versions

• PEFT 0.14.0
• Transformers 4.46.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.20.1