hubert-large-ls960-ft-V2-5

This model is a fine-tuned version of facebook/hubert-large-ls960-ft on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5760
• Wer: 0.1085
• Per: 0.0892

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• 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	Wer	Per
19.707	1.0	82	3.5254	1.0	1.0
3.4906	2.0	164	3.2483	1.0	1.0
3.233	3.0	246	3.1368	1.0	1.0
3.0468	4.0	328	2.9600	1.0	1.0
2.6751	5.0	410	2.3348	1.0	1.0
2.0881	6.0	492	1.7351	0.8568	0.8726
1.4875	7.0	574	1.2264	0.6059	0.6134
1.0922	8.0	656	0.9666	0.4068	0.3972
0.8148	9.0	738	0.7746	0.3249	0.3138
0.6332	10.0	820	0.6755	0.2477	0.2313
0.4797	11.0	902	0.6262	0.1612	0.1410
0.3807	12.0	984	0.5765	0.1384	0.1172
0.3195	13.0	1066	0.5666	0.1191	0.0992
0.2526	14.0	1148	0.5759	0.1165	0.0970
0.2417	15.0	1230	0.5460	0.1138	0.0946
0.2072	16.0	1312	0.5551	0.1095	0.0912
0.1881	17.0	1394	0.5745	0.1102	0.0917
0.1888	18.0	1476	0.5731	0.1094	0.0907
0.202	19.0	1558	0.5774	0.1081	0.0893
0.1813	20.0	1640	0.5760	0.1085	0.0892

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0