whisper-20hrs-random

This model is a fine-tuned version of openai/whisper-large-v2 on the JASMIN-CGN dataset. It achieves the following results on the evaluation set:

• Loss: 0.4303
• Wer: 22.4947

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: 48
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.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: 49
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0263	0.1543	25	1.2168	38.0448
1.0661	0.3086	50	1.1739	37.5583
0.9871	0.4630	75	1.1019	36.3337
0.8873	0.6173	100	1.0211	35.1126
0.8572	0.7716	125	0.9299	34.6831
0.7706	0.9259	150	0.8373	32.4219
0.7349	1.0802	175	0.7395	31.9455
0.6256	1.2346	200	0.6531	31.5127
0.5956	1.3889	225	0.5880	28.6141
0.5474	1.5432	250	0.5423	26.7555
0.5441	1.6975	275	0.5079	25.7490
0.5065	1.8519	300	0.4803	23.9541
0.5186	2.0062	325	0.4605	22.4477
0.4526	2.1605	350	0.4491	22.9241
0.4636	2.3148	375	0.4415	22.7229
0.4768	2.4691	400	0.4364	22.6692
0.4721	2.6235	425	0.4332	22.5081
0.4729	2.7778	450	0.4312	22.5115
0.4786	2.9321	475	0.4303	22.4947

Framework versions

• PEFT 0.16.0
• Transformers 4.52.0
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.2