JUJORUME/whisper-medium-es

FT-Spanish-openai/whisper-medium

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: 0.0000
• Wer: 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-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0134	4.83	250	0.0090	3.1136
0.0204	9.66	500	0.0153	2.9471
0.0248	14.49	750	0.0164	2.8472
0.0062	19.32	1000	0.0033	0.2581
0.0032	24.15	1250	0.0031	0.0833
0.0023	28.99	1500	0.0004	0.0083
0.0001	33.82	1750	0.0001	0.0
0.0	38.65	2000	0.0000	0.0
0.0	43.48	2250	0.0000	0.0
0.0	48.31	2500	0.0000	0.0
0.0	53.14	2750	0.0000	0.0
0.0	57.97	3000	0.0000	0.0
0.0	62.8	3250	0.0000	0.0
0.0	67.63	3500	0.0000	0.0
0.0	72.46	3750	0.0000	0.0
0.0	77.29	4000	0.0000	0.0

Framework versions

• Transformers 4.35.2
• Pytorch 2.0.1+cu117
• Datasets 2.15.0
• Tokenizers 0.15.0