small_finetune_M01

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: 3.2363
• Wer: 1.0

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: 0.0002
• train_batch_size: 20
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 800
• num_epochs: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
121.7217	200.0	800	3.1742	1.0
2.066	400.0	1600	2.8390	1.0
1.7019	600.0	2400	2.8359	1.0
1.5282	800.0	3200	2.8655	1.0
1.4089	1000.0	4000	2.8933	1.0
1.3123	1200.0	4800	2.9047	1.0
1.2361	1400.0	5600	2.9677	1.0
1.1758	1600.0	6400	3.0008	1.0
1.1241	1800.0	7200	3.0795	1.0
1.0816	2000.0	8000	3.1214	1.0
1.0497	2200.0	8800	3.1518	1.0
1.0349	2400.0	9600	3.1584	1.0
1.0058	2600.0	10400	3.1876	1.0
0.9983	2800.0	11200	3.1843	1.0
0.9863	3000.0	12000	3.1914	1.0
0.9776	3200.0	12800	3.2005	1.0
0.9647	3400.0	13600	3.2245	1.0
0.9586	3600.0	14400	3.2352	1.0
0.9521	3800.0	15200	3.2398	1.0
0.9537	4000.0	16000	3.2363	1.0

Framework versions

• Transformers 4.18.0
• Pytorch 1.10.2+cu102
• Datasets 2.3.2
• Tokenizers 0.12.1