output_model_shunyalabs_data_base_model_proper_feature_extractor_more

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

• Loss: 0.6676
• Wer: 25.3729

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 4
• 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: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 40000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6306	1.9531	2500	0.8088	52.2670
0.2802	3.9062	5000	0.7200	39.6101
0.0919	5.8594	7500	0.7014	41.1018
0.0711	7.8125	10000	0.7334	36.9960
0.0522	9.7656	12500	0.6971	34.1161
0.0440	11.7188	15000	0.7129	32.5949
0.0159	13.6719	17500	0.7492	33.7912
0.0174	15.625	20000	0.7170	31.4577
0.0093	17.5781	22500	0.7107	30.3205
0.0035	19.5312	25000	0.7096	28.4301
0.0032	21.4844	27500	0.6960	28.2824
0.0004	23.4375	30000	0.6746	28.6516
0.0001	25.3906	32500	0.6863	26.5987
0.0000	27.3438	35000	0.6776	26.3624
0.0000	29.2969	37500	0.6740	25.7421
0.0000	31.25	40000	0.6676	25.3729

Framework versions

• Transformers 5.0.0
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.2