whisper-large-v3-MH-fine-tuned

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6099

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: 0.001
• 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.0586	1.0	4	0.8056
1.0643	2.0	8	0.6915
0.7456	3.0	12	0.6105
0.6496	4.0	16	0.5754
0.6013	5.0	20	0.5447
0.443	6.0	24	0.5263
0.3267	7.0	28	0.5356
0.2908	8.0	32	0.5291
0.1509	9.0	36	0.5256
0.1156	10.0	40	0.5158
0.0715	11.0	44	0.5315
0.0516	12.0	48	0.5286
0.0457	13.0	52	0.5434
0.048	14.0	56	0.5550
0.0297	15.0	60	0.5781
0.0213	16.0	64	0.5890
0.0198	17.0	68	0.6000
0.0175	18.0	72	0.6046
0.0158	19.0	76	0.6101
0.0128	20.0	80	0.6099

Framework versions

• PEFT 0.11.2.dev0
• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2