wav2vec2-base-finetuned-ic-slurp-no-pretrain

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

• Loss: 3.1033
• Accuracy: 0.3082

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: 5e-05
• train_batch_size: 24
• eval_batch_size: 24
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.8949	1.0	527	3.8954	0.0717
3.8006	2.0	1055	3.8025	0.0810
3.7439	3.0	1582	3.7616	0.0821
3.7199	4.0	2110	3.6873	0.0925
3.6145	5.0	2637	3.6367	0.0980
3.542	6.0	3165	3.5391	0.1161
3.3305	7.0	3692	3.3986	0.1528
3.171	8.0	4220	3.2162	0.2022
2.9197	9.0	4747	3.0826	0.2344
2.6468	10.0	5275	2.9709	0.2643
2.4813	11.0	5802	2.9282	0.2880
2.1928	12.0	6330	2.9192	0.2943
1.9368	13.0	6857	2.9719	0.2974
1.693	14.0	7385	3.0304	0.3021
1.3964	15.0	7912	3.1033	0.3082
1.3051	16.0	8440	3.2700	0.2945
1.0794	17.0	8967	3.4284	0.3033
0.9993	18.0	9495	3.5327	0.2998
0.7641	19.0	10022	3.6907	0.2978
0.68	20.0	10550	3.8579	0.2984

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2