wav2vec2-large-xlsr-53-th-speech-emotion-recognition-3c

This model is a fine-tuned version of airesearch/wav2vec2-large-xlsr-53-th on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4445
• Accuracy: 0.8492

Model description

three emotion [Anger , Happiness , Neutral]

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.0305	0.9956	57	1.0278	0.4874
0.6947	1.9913	114	0.6649	0.6645
0.622	2.9869	171	0.5644	0.7607
0.5051	4.0	229	0.4936	0.7967
0.4791	4.9956	286	0.4235	0.8328
0.3918	5.9913	343	0.4273	0.8328
0.3399	6.9869	400	0.4316	0.8437
0.3473	8.0	458	0.4013	0.8448
0.3276	8.9956	515	0.4140	0.8437
0.3355	9.9913	572	0.4069	0.8459
0.2958	10.9869	629	0.4440	0.8372
0.2803	12.0	687	0.4381	0.8404
0.2996	12.9956	744	0.4100	0.8492
0.2995	13.9913	801	0.4310	0.8459
0.2645	14.9869	858	0.4590	0.8393
0.279	16.0	916	0.4317	0.8492
0.249	16.9956	973	0.4564	0.8437
0.238	17.9913	1030	0.4473	0.8459
0.209	18.9869	1087	0.4428	0.8492
0.2323	19.9127	1140	0.4445	0.8492

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1