LoRA

This model is a fine-tuned version of openai/whisper-medium on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.9259

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: 2
• eval_batch_size: 8
• 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: 50
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
5.5649	0.0484	100	5.0972
4.8399	0.0969	200	4.6258
4.4226	0.1453	300	4.1057
3.7549	0.1937	400	3.4928
3.376	0.2421	500	3.2500
3.2956	0.2906	600	3.0540
3.0468	0.3390	700	2.8512
2.8859	0.3874	800	2.7101
2.7545	0.4358	900	2.6127
2.59	0.4843	1000	2.5253
2.541	0.5327	1100	2.4448
2.5204	0.5811	1200	2.3783
2.4348	0.6295	1300	2.3030
2.3142	0.6780	1400	2.2346
2.3304	0.7264	1500	2.1672
2.1671	0.7748	1600	2.1006
2.1628	0.8232	1700	2.0384
2.0887	0.8717	1800	1.9837
2.0299	0.9201	1900	1.9445
2.0743	0.9685	2000	1.9259

Framework versions

• PEFT 0.14.1.dev0
• Transformers 4.48.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.1.dev0
• Tokenizers 0.21.0