miosipof/whisper_medium_bbrave_12-11-2024_v3.1

Whisper Medium

This model is a fine-tuned version of b-brave/asr_double_training_15-10-2024_merged on the b-brave/speech_disorders_voice_edit dataset. It achieves the following results on the evaluation set:

• Loss: 0.4433
• Wer: 37.0508

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 150
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1148	1.3245	100	0.4660	36.0595
0.0889	2.6490	200	0.4470	39.0335
0.0689	3.9735	300	0.4346	36.1834
0.0424	5.2980	400	0.4367	36.0595
0.0288	6.6225	500	0.4420	36.3073
0.0273	7.9470	600	0.4433	37.0508

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.2.0
• Datasets 3.1.0
• Tokenizers 0.20.3