wav2vec2-finetuned-on-shEMO_speech

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6265
• Accuracy: 0.8367

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.3533	1.0	75	1.2104	0.6167
1.0603	2.0	150	0.8963	0.7167
0.7558	3.0	225	0.6868	0.8233
0.5619	4.0	300	0.6482	0.8333
0.5089	5.0	375	0.5713	0.8467
0.333	6.0	450	0.6046	0.83
0.2968	7.0	525	0.5426	0.8633
0.2957	8.0	600	0.5003	0.87
0.1713	9.0	675	0.5812	0.87
0.1744	10.0	750	0.6211	0.8533
0.0821	11.0	825	0.6015	0.87
0.0961	12.0	900	0.5906	0.87
0.0969	13.0	975	0.6339	0.8633
0.1125	14.0	1050	0.6203	0.86
0.0496	15.0	1125	0.6204	0.8667

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1