fine-tuned-vctkdataset

This model is a fine-tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1289
• Wer: 0.1950

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: 32
• 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	125	2.8536	1.0
No log	2.0	250	1.6881	1.0168
No log	3.0	375	0.3762	0.4656
1.8553	4.0	500	0.1728	0.2567
1.8553	5.0	625	0.1234	0.2073
1.8553	6.0	750	0.1231	0.2007
1.8553	7.0	875	0.1225	0.2013
0.1847	8.0	1000	0.1211	0.1959
0.1847	9.0	1125	0.1217	0.2007
0.1847	10.0	1250	0.1221	0.1980
0.1847	11.0	1375	0.1258	0.1983
0.1016	12.0	1500	0.1307	0.1959
0.1016	13.0	1625	0.1179	0.1942
0.1016	14.0	1750	0.1248	0.1983
0.1016	15.0	1875	0.1281	0.1954
0.0756	16.0	2000	0.1199	0.1967
0.0756	17.0	2125	0.1284	0.1967
0.0756	18.0	2250	0.1338	0.1982
0.0756	19.0	2375	0.1270	0.1987
0.0699	20.0	2500	0.1298	0.1964
0.0699	21.0	2625	0.1319	0.1950
0.0699	22.0	2750	0.1254	0.1925
0.0699	23.0	2875	0.1285	0.1949
0.0491	24.0	3000	0.1303	0.1948
0.0491	25.0	3125	0.1293	0.1964
0.0491	26.0	3250	0.1294	0.1969
0.0491	27.0	3375	0.1279	0.1952
0.043	28.0	3500	0.1318	0.1942
0.043	29.0	3625	0.1287	0.1950
0.043	30.0	3750	0.1289	0.1950

Framework versions

• Transformers 4.29.0.dev0
• Pytorch 2.0.0+cu118
• Datasets 1.14.0
• Tokenizers 0.13.3