Jenalea/_Whisper_Call_Center_en_merged_v2

Distill Whisper Call Center

This model is a fine-tuned version of distil-whisper/distil-large-v3 on the www_call_center_eng_merged_v2 dataset. It achieves the following results on the evaluation set:

• Loss: 2.0608
• Wer: 83.0947

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.1093	0.1511	50	3.2757	120.7289
2.2517	0.3021	100	2.9028	103.9791
2.6498	0.4532	150	2.6369	99.3625
2.3836	0.6042	200	2.4538	91.9658
2.3113	0.7553	250	2.3202	89.5940
2.1796	0.9063	300	2.2170	87.2175
1.9552	1.0574	350	2.1468	84.1168
1.8708	1.2085	400	2.0990	81.8186
1.8367	1.3595	450	2.0710	83.7168
1.8655	1.5106	500	2.0608	83.0947

Framework versions

• Transformers 4.45.2
• Pytorch 2.6.0+cu124
• Datasets 3.4.0
• Tokenizers 0.20.3