whisper-tiny-finetuned-bmd-V8-fp16-20241111_172124-LOSO-section-out1

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

• Loss: 3.1208
• Accuracy: 0.3793
• F1: 0.3576

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: 16
• eval_batch_size: 16
• seed: 1968
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
No log	0.8571	3	1.1408	0.1724	0.0507
No log	2.0	7	1.2357	0.3103	0.2039
1.0639	2.8571	10	1.2708	0.3103	0.2019
1.0639	4.0	14	1.3175	0.3448	0.2309
1.0639	4.8571	17	1.4515	0.3103	0.2075
0.6963	6.0	21	1.4752	0.3103	0.2540
0.6963	6.8571	24	1.8468	0.3448	0.3213
0.6963	8.0	28	1.7103	0.3448	0.3289
0.1964	8.8571	31	2.1595	0.4138	0.3928
0.1964	10.0	35	1.8965	0.4483	0.4416
0.1964	10.8571	38	1.9849	0.4138	0.4139
0.0496	12.0	42	2.8377	0.3793	0.3470
0.0496	12.8571	45	2.7553	0.3448	0.2756
0.0496	14.0	49	2.6326	0.4138	0.4108
0.0119	14.8571	52	2.6776	0.4138	0.4139
0.0119	16.0	56	2.9911	0.3448	0.3362
0.0119	16.8571	59	3.1073	0.3793	0.3576
0.0057	17.1429	60	3.1208	0.3793	0.3576

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3