uaspeech-whisper-lg-3-Nov3

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

• Loss: 0.1246
• Wer: 15.4002

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: 8
• eval_batch_size: 8
• 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: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.7741	0.0719	100	0.6537	58.1245
0.4975	0.1437	200	0.4079	39.8141
0.3739	0.2156	300	0.3398	33.3872
0.3037	0.2875	400	0.2941	30.2344
0.2783	0.3593	500	0.2456	26.1116
0.2568	0.4312	600	0.2270	25.1011
0.2012	0.5031	700	0.2372	25.9903
0.2139	0.5749	800	0.1828	21.3015
0.1649	0.6468	900	0.1750	19.7656
0.149	0.7186	1000	0.1640	19.4826
0.146	0.7905	1100	0.1444	17.5829
0.1424	0.8624	1200	0.1305	15.5214
0.116	0.9342	1300	0.1294	16.3703
0.121	1.0061	1400	0.1210	16.1277
0.0751	1.0780	1500	0.1246	15.4002

Framework versions

• Transformers 4.43.4
• Pytorch 2.4.1
• Datasets 3.0.0
• Tokenizers 0.19.1