simpragma/breeze-listen-dsw-base-ta

Breeze DSW Tamil - base

This model is a fine-tuned version of openai/whisper-base on the mozilla-foundation/common_voice_16_0 ta dataset. It achieves the following results on the evaluation set:

• Loss: 0.375
• Wer: 21.4071

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: 32
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1698	0.1	100	0.5723	30.4406
0.3578	0.2	200	0.4302	25.6862
0.2832	0.3	300	0.3967	23.2048
0.2663	0.4	400	0.4038	23.8525
0.5175	0.5	500	0.3962	24.1466
0.2365	0.6	600	0.3850	22.2595
0.1692	0.7	700	0.3960	21.8687
0.1815	0.8	800	0.3823	22.0772
0.1612	0.9	900	0.3701	21.8056
0.1393	1.0	1000	0.375	21.4071

Framework versions

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