enakilci/whisper-large-v3-fi-800steps-8batch-2grad_steps-0.0001lr

Whisper Large v3 Fine-Tuned Finnish

This model is a fine-tuned version of openai/whisper-large-v3 on the Common Voice 13.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3108
• Wer: 24.2868

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: 0.0001
• 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: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7157	0.21	50	0.4892	42.8216
0.6314	0.42	100	0.6716	58.7153
0.6187	0.63	150	0.5979	47.1195
0.5396	0.84	200	0.5503	45.8126
0.4857	1.05	250	0.5842	42.9873
0.246	1.26	300	0.5526	43.8984
0.2635	1.47	350	0.4893	39.4994
0.2346	1.68	400	0.4657	36.8489
0.2268	1.89	450	0.4163	34.5113
0.1345	2.11	500	0.4152	30.9590
0.0862	2.32	550	0.4157	32.6063
0.0723	2.53	600	0.3942	29.5785
0.0667	2.74	650	0.3654	28.3913
0.0571	2.95	700	0.3235	25.8513
0.0241	3.16	750	0.3109	25.0874
0.0124	3.37	800	0.3108	24.2868

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.0.1
• Datasets 2.16.1
• Tokenizers 0.15.0