jimregan/whisper-small-sv-riksdag

Whisper Small Sv - Riksdag 100h

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

• Loss: 0.4977
• Wer: 1118.4718

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1384	0.11	1000	0.4747	380.8335
0.1186	0.22	2000	0.4513	1032.3900
0.1056	0.33	3000	0.4385	582.0427
0.0824	0.43	4000	0.4465	574.8907
0.0961	0.54	5000	0.4199	1004.9138
0.0939	0.65	6000	0.4478	866.2979
0.0758	0.76	7000	0.4384	907.9496
0.0741	0.87	8000	0.4264	641.1371
0.0692	0.98	9000	0.4206	1142.6550
0.0257	1.08	10000	0.4707	1152.4312
0.0273	1.19	11000	0.4789	1100.2058
0.021	1.3	12000	0.4763	1236.1719
0.0163	1.41	13000	0.5035	924.8006
0.0183	1.52	14000	0.4911	1285.1814
0.024	1.63	15000	0.4861	1140.8284
0.0158	1.73	16000	0.4793	1181.7597
0.0167	1.84	17000	0.4759	1207.3064
0.0231	1.95	18000	0.4801	1139.6964
0.0054	2.06	19000	0.4934	1114.4842
0.006	2.17	20000	0.4977	1118.4718

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1
• Datasets 2.10.1
• Tokenizers 0.13.2