wav2vec2-base-960h-EMOPIA-10sec-full

This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2928
• Accuracy: 0.8488

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2057	1.0	2248	1.4522	0.4502
1.3873	2.0	4496	1.4503	0.6423
1.4246	3.0	6744	1.6165	0.6673
1.3335	4.0	8992	1.4786	0.7206
1.251	5.0	11240	1.6414	0.6886
1.1859	6.0	13488	1.3300	0.7544
1.1132	7.0	15736	1.3665	0.7509
1.0189	8.0	17984	1.6665	0.7153
0.9807	9.0	20232	1.1175	0.7794
0.8786	10.0	22480	1.1786	0.7883
0.8677	11.0	24728	1.1295	0.7811
0.7554	12.0	26976	1.1185	0.8185
0.7196	13.0	29224	1.4067	0.7847
0.692	14.0	31472	1.1175	0.8203
0.6276	15.0	33720	1.4490	0.7883
0.6083	16.0	35968	1.0983	0.8345
0.5204	17.0	38216	1.1814	0.8256
0.5197	18.0	40464	1.2945	0.8167
0.488	19.0	42712	1.4494	0.8025
0.4714	20.0	44960	1.3499	0.8114
0.3641	21.0	47208	1.2525	0.8381
0.3877	22.0	49456	1.2610	0.8381
0.3253	23.0	51704	1.3913	0.8274
0.2978	24.0	53952	1.2990	0.8416
0.3238	25.0	56200	1.4328	0.8274
0.2669	26.0	58448	1.3079	0.8327
0.2521	27.0	60696	1.3250	0.8399
0.2632	28.0	62944	1.3357	0.8416
0.2655	29.0	65192	1.2957	0.8434
0.2379	30.0	67440	1.2928	0.8488

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu118
• Datasets 3.0.1
• Tokenizers 0.20.0