Makkoen/whisper-medium.en-cit-do015-wd0-lr1e-06-SF-600

./600

This model is a fine-tuned version of openai/whisper-medium.en on the 600 SF 1000 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6509
• Wer Ortho: 30.9402
• Wer: 20.0933

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-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: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
1.6008	2.6667	100	1.0986	41.0350	29.9605
0.8522	5.3333	200	0.7925	32.0335	21.0621
0.6516	8.0	300	0.7207	30.5029	19.9856
0.5337	10.6667	400	0.6885	30.5758	20.3803
0.4489	13.3333	500	0.6709	31.0496	20.3086
0.4003	16.0	600	0.6577	31.0496	20.2727
0.3588	18.6667	700	0.6533	31.0496	20.0933
0.3499	21.3333	800	0.6509	30.9402	20.0933

Framework versions

• Transformers 4.44.0
• Pytorch 1.13.1+cu117
• Datasets 2.20.0
• Tokenizers 0.19.1