wav2vec2-base_lr_3e-4

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0682
• Accuracy: 0.9784

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: 0.0003
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• 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
1.7893	0.9851	33	1.5529	0.4602
0.9637	2.0	67	0.8562	0.7563
0.5758	2.9851	100	0.4980	0.8276
0.5401	4.0	134	0.3442	0.8875
0.3908	4.9851	167	0.4630	0.8322
0.348	6.0	201	0.2102	0.9260
0.309	6.9851	234	0.1996	0.9391
0.305	8.0	268	0.3001	0.9185
0.2311	8.9851	301	0.2150	0.9335
0.2362	10.0	335	0.1218	0.9550
0.1929	10.9851	368	0.1334	0.9550
0.1781	12.0	402	0.1077	0.9597
0.15	12.9851	435	0.0749	0.9719
0.1437	14.0	469	0.0710	0.9756
0.1135	14.7761	495	0.0682	0.9784

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1