B4_1000_1e-5_hp-myself-2

This model is a fine-tuned version of lnxdx/20_2000_1e-5_hp-mehrdad on the None dataset. It achieves the following results on the evaluation set:

• Loss on ShEMO train set: 0.7516
• Loss on ShEMO dev set: 0.6705
• WER on ShEMO train set: 28.02
• WER on ShEMO dev set: 31.16
• WER on Common Voice 13 test set: 19.34

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8083	0.62	100	0.6766	0.3271
0.8414	1.25	200	0.6774	0.3259
0.8465	1.88	300	0.6686	0.3262
0.7819	2.5	400	0.6749	0.3207
0.7905	3.12	500	0.6848	0.3178
0.8078	3.75	600	0.6571	0.3245
0.7771	4.38	700	0.6683	0.3145
0.7786	5.0	800	0.6688	0.3137
0.7656	5.62	900	0.6703	0.3134
0.7516	6.25	1000	0.6706	0.3131

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0