finetuning-1-12-2022

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

• Loss: 0.0357
• Wer: 0.0836

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: 0.0003
• 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: 500
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	45	2.6669	1.0
No log	2.0	90	1.4650	0.9652
No log	3.0	135	0.8614	0.7491
No log	4.0	180	0.8031	0.7735
No log	5.0	225	0.7993	0.7909
No log	6.0	270	0.5708	0.6411
No log	7.0	315	0.5728	0.7178
No log	8.0	360	0.5439	0.6341
1.2911	9.0	405	0.5072	0.7213
1.2911	10.0	450	0.3578	0.5331
1.2911	11.0	495	0.4871	0.6411
1.2911	12.0	540	0.3034	0.4634
1.2911	13.0	585	0.4684	0.6028
1.2911	14.0	630	0.2638	0.4216
1.2911	15.0	675	0.2657	0.4948
1.2911	16.0	720	0.2593	0.3972
1.2911	17.0	765	0.2770	0.4634
0.3079	18.0	810	0.2936	0.4530
0.3079	19.0	855	0.4168	0.5436
0.3079	20.0	900	0.2642	0.3693
0.3079	21.0	945	0.1827	0.3519
0.3079	22.0	990	0.1807	0.2962
0.3079	23.0	1035	0.2134	0.3484
0.3079	24.0	1080	0.1317	0.2474
0.3079	25.0	1125	0.0950	0.2021
0.3079	26.0	1170	0.0985	0.1707
0.1678	27.0	1215	0.1444	0.2753
0.1678	28.0	1260	0.0816	0.1289
0.1678	29.0	1305	0.1103	0.1916
0.1678	30.0	1350	0.0878	0.1777
0.1678	31.0	1395	0.1436	0.1568
0.1678	32.0	1440	0.1097	0.1882
0.1678	33.0	1485	0.0995	0.1777
0.1678	34.0	1530	0.0917	0.1882
0.1678	35.0	1575	0.0691	0.1254
0.0743	36.0	1620	0.0394	0.0941
0.0743	37.0	1665	0.0592	0.1185
0.0743	38.0	1710	0.0680	0.1220
0.0743	39.0	1755	0.0748	0.0941
0.0743	40.0	1800	0.0651	0.1010
0.0743	41.0	1845	0.0688	0.1045
0.0743	42.0	1890	0.0489	0.0871
0.0743	43.0	1935	0.0524	0.0976
0.0743	44.0	1980	0.0415	0.1080
0.0234	45.0	2025	0.0489	0.0767
0.0234	46.0	2070	0.0337	0.0732
0.0234	47.0	2115	0.0456	0.0662
0.0234	48.0	2160	0.0326	0.0871
0.0234	49.0	2205	0.0319	0.0976
0.0234	50.0	2250	0.0357	0.0836

Framework versions

• Transformers 4.23.0
• Pytorch 1.13.0+cpu
• Datasets 2.5.2
• Tokenizers 0.13.1