wav2vec2-base-finetuned-sentiment-mesd

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

• Loss: 0.5729
• Accuracy: 0.8308

Model description

This model was trained to classify underlying sentiment of Spanish audio/speech.

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: 1.25e-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
No log	1.0	7	0.5729	0.8308
No log	2.0	14	0.6577	0.8
0.1602	3.0	21	0.7055	0.8
0.1602	4.0	28	0.8696	0.7615
0.1602	5.0	35	0.6807	0.7923
0.1711	6.0	42	0.7303	0.7923
0.1711	7.0	49	0.7028	0.8077
0.1711	8.0	56	0.7368	0.8
0.1608	9.0	63	0.7190	0.7923
0.1608	10.0	70	0.6913	0.8077
0.1608	11.0	77	0.7047	0.8077
0.1753	12.0	84	0.6801	0.8
0.1753	13.0	91	0.7208	0.7769
0.1753	14.0	98	0.7458	0.7846
0.203	15.0	105	0.6494	0.8077
0.203	16.0	112	0.6256	0.8231
0.203	17.0	119	0.6788	0.8
0.1919	18.0	126	0.6757	0.7846
0.1919	19.0	133	0.6859	0.7846
0.1641	20.0	140	0.6832	0.7846

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu111
• Datasets 2.0.0
• Tokenizers 0.10.3