whisper-big-kclpn

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1280
• Wer: 22.3875

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: 0.0004
• 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: 132
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5531	2.4845	200	0.2010	18.2677
0.0654	4.9689	400	0.1670	21.0812
0.0348	7.4534	600	0.1150	16.9815
0.0203	9.9379	800	0.1049	43.5088
0.0111	12.4224	1000	0.1212	84.6262
0.0053	14.9068	1200	0.1043	21.1616
0.002	17.3913	1400	0.1263	21.8248
0.0008	19.8758	1600	0.1255	22.7492
0.0	22.3602	1800	0.1269	22.7090
0.0	24.8447	2000	0.1276	22.5080
0.0	27.3292	2200	0.1279	22.5281
0.0	29.8137	2400	0.1280	22.3875

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1