Wav2vec2-xlsr-Shemo-Ravdess-4EMO

This model is a fine-tuned version of makhataei/Wav2vec2-xlsr-Shemo-Ravdess-4EMO on the minoosh/shEMO dataset. It achieves the following results on the evaluation set:

• Loss: 0.9089
• Accuracy: 0.6712

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 35

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.9106	1.0	250	1.0288	0.6259
0.8205	2.0	500	0.9426	0.6576
0.7767	3.0	750	0.9707	0.6553
0.803	4.0	1000	0.9698	0.6644
0.7489	5.0	1250	0.9583	0.6463
0.7734	6.0	1500	0.9138	0.6757
0.7603	7.0	1750	0.8905	0.6712
0.7741	8.0	2000	0.9169	0.6599
0.7569	9.0	2250	0.9369	0.6417
0.7854	10.0	2500	0.9256	0.6599
0.7572	11.0	2750	0.9320	0.6621
0.7537	12.0	3000	0.8960	0.6825
0.745	13.0	3250	0.9495	0.6599
0.7598	14.0	3500	0.9196	0.6667
0.7536	15.0	3750	0.9464	0.6599
0.7428	16.0	4000	0.9407	0.6485
0.757	17.0	4250	0.9251	0.6689
0.7694	18.0	4500	0.9246	0.6576
0.7501	19.0	4750	0.9283	0.6621
0.7464	20.0	5000	0.9333	0.6531
0.7569	21.0	5250	0.9062	0.6667
0.745	22.0	5500	0.9569	0.6485
0.7404	23.0	5750	0.9062	0.6667
0.7384	24.0	6000	0.8948	0.6780
0.7524	25.0	6250	0.9296	0.6599
0.7574	26.0	6500	0.8925	0.6825
0.7876	27.0	6750	0.9061	0.6712
0.7692	28.0	7000	0.9319	0.6508
0.7352	29.0	7250	0.9145	0.6644
0.7496	30.0	7500	0.9068	0.6735
0.7406	31.0	7750	0.9024	0.6735
0.7334	32.0	8000	0.9231	0.6576
0.761	33.0	8250	0.9073	0.6712
0.7476	34.0	8500	0.9097	0.6667
0.7868	35.0	8750	0.9089	0.6712

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3