enakilci/whisper-large-v3-fi-800steps-8batch-2grad_steps-5e-05lr

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: 0.2128
• Wer: 19.4828

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: 5e-05
• 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
0.6193	0.21	50	0.2905	29.1920
0.3515	0.42	100	0.3581	32.2014
0.3433	0.63	150	0.3497	43.9812
0.3196	0.84	200	0.3080	27.9956
0.2597	1.05	250	0.3213	27.5630
0.1368	1.26	300	0.3088	29.0263
0.1316	1.47	350	0.3018	27.0569
0.1193	1.68	400	0.2948	28.5846
0.1219	1.89	450	0.2608	25.1979
0.0738	2.11	500	0.2645	30.9682
0.042	2.32	550	0.2493	23.2008
0.0406	2.53	600	0.2589	21.6823
0.0317	2.74	650	0.2391	24.9862
0.0336	2.95	700	0.2217	21.6639
0.0127	3.16	750	0.2126	20.3939
0.0085	3.37	800	0.2128	19.4828

Framework versions

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