miosipof/asr_michael_test2.1

Demo Model Whisper Large

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

• Loss: 0.3289

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: 0.0002
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.6981	0.0794	10	3.0679
2.7279	0.1587	20	2.9277
2.4063	0.2381	30	2.5832
2.0624	0.3175	40	1.9876
1.5946	0.3968	50	1.4144
1.032	0.4762	60	0.8543
0.6495	0.5556	70	0.5918
0.4086	0.6349	80	0.5795
0.6702	0.7143	90	1.3734
0.7216	0.7937	100	1.2499
0.7264	0.8730	110	0.4231
0.442	0.9524	120	0.3969
0.2848	1.0317	130	0.3757
0.2054	1.1111	140	0.3697
0.307	1.1905	150	0.3584
0.3914	1.2698	160	0.3579
0.175	1.3492	170	0.3515
0.2399	1.4286	180	0.3414
0.3367	1.5079	190	0.3360
0.2457	1.5873	200	0.3315
0.3358	1.6667	210	0.3351
0.1335	1.7460	220	0.3314
0.1844	1.8254	230	0.3308
0.3878	1.9048	240	0.3282
0.3594	1.9841	250	0.3289

Framework versions

• PEFT 0.11.1
• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1