Makkoen/whisper-large-v3-cit-do015-wd0-lr2e-06-HOU-1000

./947

This model is a fine-tuned version of openai/whisper-large-v3 on the 947 SF 1000 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5972
• Wer Ortho: 30.6410
• Wer: 23.2591

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: 2e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• 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: 200
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.8695	1.8692	100	0.5967	31.9007	25.0619
0.5034	3.7383	200	0.5483	29.7518	23.1177
0.3543	5.6075	300	0.5625	30.4187	23.3651
0.2641	7.4766	400	0.5928	31.1226	23.1884
0.2226	9.3458	500	0.5972	30.6410	23.2591

Framework versions

• Transformers 4.44.0
• Pytorch 1.13.1+cu117
• Datasets 2.21.0
• Tokenizers 0.19.1