ras0k/whisper-rap-queb-v5

Whisper Quebec Rap - ras0k

This model is a fine-tuned version of openai/whisper-large-v3 on the Genius-Quebec-Rap-Top-300 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9359
• Wer: 32.0698

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Github: https://github.com/ras0k/whisper-rap-queb

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 4e-05
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 125
• training_steps: 1250

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0	0	0.8411	61.5404
0.0196	13.1579	250	0.7737	32.6363
0.0033	26.3158	500	0.9160	32.2323
0.0029	39.4737	750	0.9310	32.2281
0.0028	52.6316	1000	0.9361	32.2198
0.0028	65.7895	1250	0.9359	32.0698

Framework versions

• Transformers 5.2.0
• Pytorch 2.10.0+cu128
• Datasets 4.6.1
• Tokenizers 0.22.2