moore_small_V1

This model is a fine-tuned version of openai/whisper-small on the abdouaziiz/moore_clean dataset. It achieves the following results on the evaluation set:

• Loss: 0.0515
• Wer: 0.1004

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 48000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8667	0.5770	500	0.2067	0.2110
0.3227	1.1541	1000	0.1456	0.1656
0.2201	1.7311	1500	0.1188	0.1238
0.1564	2.3081	2000	0.1009	0.1232
0.1213	2.8852	2500	0.0778	0.1123
0.0779	3.4622	3000	0.0749	0.1027
0.0694	4.0392	3500	0.0689	0.1069
0.0434	4.6163	4000	0.0639	0.1254
0.0413	5.1933	4500	0.0611	0.1055
0.0332	5.7703	5000	0.0570	0.1110
0.0287	6.3474	5500	0.0583	0.1050
0.0257	6.9244	6000	0.0515	0.1004
0.0217	7.5014	6500	0.0538	0.0964
0.021	8.0785	7000	0.0537	0.0902
0.018	8.6555	7500	0.0567	0.0850
0.0172	9.2325	8000	0.0533	0.0792

Framework versions

• Transformers 4.46.0
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.20.3