FelixK7/whisper-medium-lv-ver1

Whisper medium LV - Felikss Kleins

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2416
• Wer: 11.9387

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• 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: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.02	200	0.2930	22.3576
0.9797	1.0116	400	0.2359	18.2083
0.357	2.0033	600	0.2274	16.4665
0.2582	2.0233	800	0.2111	15.6402
0.1718	3.0149	1000	0.2135	14.9883
0.1718	4.0066	1200	0.2090	14.2294
0.1355	4.0266	1400	0.2193	13.5537
0.1024	5.0183	1600	0.2255	14.5048
0.0836	6.0099	1800	0.2145	12.9751
0.0699	7.0015	2000	0.2232	13.2129
0.0699	7.0216	2200	0.2181	12.7155
0.0598	8.0132	2400	0.2192	12.7076
0.054	9.0048	2600	0.2348	13.0048
0.0452	9.0249	2800	0.2241	13.0940
0.0433	10.0165	3000	0.2406	12.6362
0.0433	11.0082	3200	0.2283	12.5332
0.0377	11.0282	3400	0.2293	12.2201
0.0317	12.0198	3600	0.2323	12.6144
0.0297	13.0114	3800	0.2309	12.2974
0.0267	14.0031	4000	0.2342	11.9011
0.0267	14.0231	4200	0.2286	12.1171
0.0243	15.0147	4400	0.2364	12.0854
0.0218	16.0064	4600	0.2405	12.1805
0.021	16.0264	4800	0.2422	12.0338
0.0173	17.0180	5000	0.2416	11.9387

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.0.1
• Datasets 3.0.0
• Tokenizers 0.19.1