cantillation/whisper-medium-he-teamim-aviv-bavly-without-nikud-test

he

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

• Loss: 1.3168
• Precision: 0.1010
• Recall: 0.1098
• F1: 0.1051
• Precision Median: 0.0
• Recall Median: 0.0
• F1 Median: 0.0
• Precision Max: 0.5
• Recall Max: 0.5333
• F1 Max: 0.5161
• Precision Min: 0.0
• Recall Min: 0.0
• F1 Min: 0.0

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: 32
• eval_batch_size: 8
• seed: 42
• 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	Precision	Recall	F1	Precision Median	Recall Median	F1 Median	Precision Max	Recall Max	F1 Max	Precision Min	Recall Min	F1 Min
3.5705	0.32	200	2.7306	0.0	0.0	0.0	0.0	0.0	0.0	0	0	0	0	0	0
2.0183	0.63	400	1.6209	0.0341	0.0373	0.0356	0.0	0.0	0.0	0.4375	0.4667	0.4516	0.0	0.0	0.0
1.6646	0.95	600	1.3710	0.1062	0.1151	0.1103	0.0	0.0	0.0	0.5	0.5333	0.5161	0.0	0.0	0.0
1.6004	1.27	800	1.3168	0.1010	0.1098	0.1051	0.0	0.0	0.0	0.5	0.5333	0.5161	0.0	0.0	0.0

Framework versions

• Transformers 4.36.2
• Pytorch 1.13.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.0