d-l-dl

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: 3.4495
• 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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
42.4143	49.8	100	21.5116	1.0
5.9884	99.8	200	31.7976	1.0
4.0043	149.8	300	3.4829	1.0
3.653	199.8	400	3.6417	1.0
3.5207	249.8	500	3.5081	1.0
3.63	299.8	600	3.4836	1.0
3.648	349.8	700	3.4515	1.0
3.6448	399.8	800	3.4647	1.0
3.6872	449.8	900	3.4371	1.0
3.6892	499.8	1000	3.4337	1.0
3.684	549.8	1100	3.4375	1.0
3.6843	599.8	1200	3.4452	1.0
3.6842	649.8	1300	3.4416	1.0
3.6819	699.8	1400	3.4498	1.0
3.6832	749.8	1500	3.4524	1.0
3.6828	799.8	1600	3.4495	1.0

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 1.18.3
• Tokenizers 0.10.3