Sep26-Mixat-whisper-lg-3-translation

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: 0.7932
• Wer: 42.6353

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
0.8768	0.4292	100	0.4849	38.3799
0.5884	0.8584	200	0.4886	37.6625
0.4802	1.2876	300	0.4899	42.7189
0.4519	1.7167	400	0.5002	42.3724
0.4173	2.1459	500	0.5083	43.9228
0.3271	2.5751	600	0.5200	41.2447
0.3292	3.0043	700	0.5020	41.7533
0.1963	3.4335	800	0.5670	43.8933
0.2076	3.8627	900	0.5536	42.9842
0.1413	4.2918	1000	0.5866	42.1439
0.1194	4.7210	1100	0.6091	43.5739
0.0994	5.1502	1200	0.6991	42.6722
0.067	5.5794	1300	0.6573	44.6869
0.0699	6.0086	1400	0.6579	44.4363
0.0386	6.4378	1500	0.7268	46.2249
0.0414	6.8670	1600	0.7219	44.3527
0.0334	7.2961	1700	0.7521	45.5763
0.0308	7.7253	1800	0.7932	42.6353

Framework versions

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