All_balanced-lang_tag-whisper-lg-3-Nov30

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

• Loss: 0.2030
• Wer: 18.0679

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: 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: 100
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0883	0.3210	100	0.5905	33.2422
0.4857	0.6421	200	0.4462	26.5892
0.3709	0.9631	300	0.3049	27.3639
0.1935	1.2841	400	0.2699	22.2602
0.1615	1.6051	500	0.2412	21.6906
0.1504	1.9262	600	0.2297	23.1032
0.0921	2.2472	700	0.2316	20.8931
0.0736	2.5682	800	0.2132	19.8679
0.0782	2.8892	900	0.2108	22.6475
0.0555	3.2103	1000	0.2226	19.4577
0.0489	3.5313	1100	0.2099	20.5742
0.0418	3.8523	1200	0.2068	19.9134
0.0364	4.1734	1300	0.2309	22.5564
0.0296	4.4944	1400	0.2175	22.5564
0.0285	4.8154	1500	0.2040	19.3210
0.0213	5.1364	1600	0.2037	18.6147
0.0156	5.4575	1700	0.2159	18.6375
0.0172	5.7785	1800	0.2068	19.0704
0.0183	6.0995	1900	0.2134	18.2046
0.0184	6.4205	2000	0.2085	18.1362
0.0142	6.7416	2100	0.1998	17.4755
0.0163	7.0626	2200	0.2059	18.1590
0.009	7.3836	2300	0.1967	18.3185
0.012	7.7047	2400	0.1976	17.5894
0.0119	8.0257	2500	0.1894	19.5944
0.0085	8.3467	2600	0.1961	18.4780
0.0059	8.6677	2700	0.2018	17.3844
0.0068	8.9888	2800	0.1821	17.5439
0.0056	9.3098	2900	0.1996	18.0451
0.0053	9.6308	3000	0.2143	17.8856
0.0077	9.9518	3100	0.1810	16.4502
0.0069	10.2729	3200	0.1873	17.3160
0.0076	10.5939	3300	0.1897	18.6375
0.0095	10.9149	3400	0.2144	18.6147
0.0051	11.2360	3500	0.2006	17.2477
0.0085	11.5570	3600	0.2106	17.0198
0.013	11.8780	3700	0.2030	18.0679

Framework versions

• Transformers 4.43.4
• Pytorch 2.4.1
• Datasets 3.0.0
• Tokenizers 0.19.1