Luandrie/_Whisper_Call_Center_NamesAdded_1000

Distill Whisper Call Center NER 1000

This model is a fine-tuned version of lelapa/distill_whisper_call_center_en_merged on the lelapa/Names_Accents dataset. It achieves the following results on the evaluation set:

• Loss: 0.1828
• Wer: 13.1183

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: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 256
• total_eval_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 125
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3331	2.9412	100	0.4857	39.1828
0.1568	5.8824	200	0.2184	17.1613
0.0243	8.8235	300	0.1941	15.2258
0.0059	11.7647	400	0.1830	13.9355
0.0019	14.7059	500	0.1823	13.8065
0.001	17.6471	600	0.1819	13.3763
0.0008	20.5882	700	0.1823	13.2043
0.0007	23.5294	800	0.1825	13.2043
0.0006	26.4706	900	0.1828	13.2043
0.0006	29.4118	1000	0.1828	13.1183

Framework versions

• Transformers 4.45.2
• Pytorch 2.7.0+cu126
• Datasets 3.6.0
• Tokenizers 0.20.3