outs

This model is a fine-tuned version of biodatlab/whisper-th-medium-combined on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0874
• Cer: 5.5267

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2.0
• total_train_batch_size: 32.0
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
No log	0.9926	67	0.0899	4.9847
0.2506	2.0	135	0.0949	4.8051
0.085	2.9926	202	0.0822	4.3363
0.085	4.0	270	0.0841	5.2042
0.0456	4.9926	337	0.0896	9.6136
0.0392	6.0	405	0.0875	4.7087
0.0392	6.9926	472	0.0879	4.7320
0.0296	8.0	540	0.0874	5.3804
0.0276	8.9926	607	0.0860	5.2873
0.0276	9.9259	670	0.0874	5.5267

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1