UDA-LIDI/openai-whisper-large-fullFT-es_ecu911_V2martin_win30s

Whisper openai-whisper-large

This model is a fine-tuned version of openai/whisper-large on the llamadas ecu911 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2752
• Wer: 21.9973

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: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5213	0.9980	379	0.3177	28.9193
0.1902	1.9987	759	0.2807	26.8399
0.1043	2.9993	1139	0.2605	20.3420
0.0604	4.0	1519	0.2653	17.9070
0.0424	4.9980	1898	0.2632	16.4979
0.2015	5.9987	2278	0.2869	18.2353
0.0257	6.9993	2658	0.2714	16.4979
0.0236	8.0	3038	0.2759	18.3584
0.0198	8.9980	3417	0.2700	15.5267
0.0156	9.9803	3790	0.2752	21.9973

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1