UDA-LIDI/openai-whisper-large-v3-fullFT-es_ecu911_V2martin_win30s15s_samples

Whisper openai-whisper-large-v3

This model is a fine-tuned version of openai/whisper-large-v3 on the llamadas ecu911 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7087
• Wer: 44.4219

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
0.8627	1.0	72	0.6537	54.2258
0.4133	2.0	144	0.5968	66.1934
0.2131	3.0	216	0.5635	73.5632
0.1184	4.0	288	0.6221	45.9770
0.067	5.0	360	0.6224	44.5571
0.0452	6.0	432	0.6335	50.5747
0.0333	7.0	504	0.6728	44.3543
0.0308	8.0	576	0.7232	44.7600
0.0244	9.0	648	0.7012	43.4753
0.0207	10.0	720	0.7087	44.4219

Framework versions

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