wav2vec2-base-mn-pretrain-42h-finetuned-speech-commands

This model is a fine-tuned version of bayartsogt/wav2vec2-base-mn-pretrain-42h on the Mongolian Speech Commands dataset. It achieves the following results on the evaluation set:

• Loss: 0.1007
• Accuracy: 0.9762
• F1: 0.9758

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: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
2.2273	1.0	17	2.2714	0.1190	0.0253
1.7478	2.0	34	1.2036	0.8452	0.8242
0.775	3.0	51	0.4755	0.9524	0.9526
0.4738	4.0	68	0.2056	0.9881	0.9878
0.3146	5.0	85	0.1485	0.9762	0.9765
0.2677	6.0	102	0.1277	0.9762	0.9758
0.2636	7.0	119	0.0919	0.9881	0.9880
0.2122	8.0	136	0.0903	0.9762	0.9758
0.1817	9.0	153	0.0782	0.9881	0.9880
0.198	10.0	170	0.0982	0.9762	0.9758
0.1436	11.0	187	0.1053	0.9762	0.9758
0.1111	12.0	204	0.1004	0.9762	0.9758
0.1607	13.0	221	0.1176	0.9762	0.9758
0.1209	14.0	238	0.1097	0.9762	0.9758
0.0974	15.0	255	0.1136	0.9762	0.9758
0.1351	16.0	272	0.0986	0.9762	0.9758
0.1008	17.0	289	0.1010	0.9762	0.9758
0.097	18.0	306	0.0781	0.9762	0.9758
0.0806	19.0	323	0.1106	0.9762	0.9758
0.0744	20.0	340	0.1007	0.9762	0.9758

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.0
• Datasets 2.14.4
• Tokenizers 0.13.3