wav2vec2-base-960h-no-softmax-quality-daps

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: 0.0422
• Accuracy: 0.9925

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: 3e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 512
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6897	0.9630	13	0.6386	0.9318
0.6402	2.0	27	0.3482	0.9943
0.2867	2.9630	40	0.1576	0.9938
0.1729	4.0	54	0.0806	0.9960
0.1058	4.9630	67	0.0539	0.9952
0.0542	6.0	81	0.0446	0.9947
0.0393	6.9630	94	0.0409	0.9943
0.0398	8.0	108	0.0401	0.9938
0.0325	8.9630	121	0.0459	0.9921
0.0251	10.0	135	0.0444	0.9925
0.0331	10.9630	148	0.0418	0.9930
0.0185	12.0	162	0.0373	0.9930
0.0254	12.9630	175	0.0367	0.9930
0.0238	14.0	189	0.0423	0.9925
0.023	14.4444	195	0.0422	0.9925

Framework versions

• Transformers 4.40.1
• Pytorch 2.3.0
• Datasets 2.12.0
• Tokenizers 0.19.1