enakilci/whisper-large-v3-fi-800steps-8batch-8grad_steps-1e-05lr

Whisper Large v3 Fine-Tuned Finnish - CommonVoice13

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3976
• Wer: 21.4246

It achieves the following results on the Test set:

• Eval_Wer: 21.378612184796612
• Eval_NormalizedWer: 18.415004137170175

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• 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
0.0001	0.84	50	0.4009	21.6363
0.0013	1.68	100	0.3801	22.5014
0.0013	2.53	150	0.3852	23.2192
0.0009	3.37	200	0.3738	23.1824
0.0007	4.21	250	0.3697	23.2100
0.0001	5.05	300	0.3777	21.9032
0.0001	5.89	350	0.3825	21.8388
0.0001	6.74	400	0.3864	21.7651
0.0	7.58	450	0.3895	21.6455
0.0	8.42	500	0.3917	21.5351
0.0	9.26	550	0.3936	21.4983
0.0	10.11	600	0.3951	21.4338
0.0	10.95	650	0.3962	21.4338
0.0	11.79	700	0.3970	21.4614
0.0	12.63	750	0.3975	21.4338
0.0	13.47	800	0.3976	21.4246

Framework versions

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