sew-timit-upsample-1

This model is a fine-tuned version of asapp/sew-tiny-100k-ft-ls100h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1115
• Wer: 0.1198
• Cer: 0.0289

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.2366	1.2690	500	0.3113	0.2875	0.0846
1.2642	2.5381	1000	0.3129	0.3038	0.0878
1.1885	3.8071	1500	0.2440	0.2468	0.0697
1.036	5.0761	2000	0.1945	0.2014	0.0543
0.9806	6.3452	2500	0.1604	0.1650	0.0428
0.9132	7.6142	3000	0.1288	0.1400	0.0351
0.8964	8.8832	3500	0.1115	0.1198	0.0289

Framework versions

• Transformers 4.51.3
• Pytorch 2.2.1
• Datasets 3.6.0
• Tokenizers 0.21.1