wav2vec2-xlsr-korean-speech-emotion-recognition

This model is a fine-tuned version of jungjongho/wav2vec2-large-xlsr-korean-demo-colab_epoch15 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.6651
Accuracy: 0.7667

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
gradient_accumulation_steps: 2
total_train_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 2
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.7098	0.02	20	1.6849	0.1986
1.6093	0.05	40	1.6102	0.2237
1.5673	0.07	60	1.5126	0.3696
1.554	0.1	80	1.4260	0.3624
1.3676	0.12	100	1.2504	0.4761
1.3763	0.14	120	1.3997	0.4091
1.2727	0.17	140	1.3728	0.4211
1.4632	0.19	160	1.2427	0.4653
1.2167	0.22	180	1.1379	0.5371
1.1789	0.24	200	1.1165	0.5514
1.1639	0.26	220	1.0029	0.5813
1.1664	0.29	240	1.1682	0.4844
1.1223	0.31	260	1.0946	0.5526
1.2748	0.33	280	1.0023	0.5837
1.1422	0.36	300	1.0873	0.5215
0.9999	0.38	320	1.0378	0.5742
1.0444	0.41	340	1.0049	0.5921
1.0099	0.43	360	0.9367	0.6160
0.9251	0.45	380	1.0681	0.5694
0.8257	0.48	400	1.0516	0.5813
1.3777	0.5	420	1.0250	0.6280
0.9415	0.53	440	0.9033	0.6316
0.9927	0.55	460	0.9360	0.6184
1.0245	0.57	480	0.9133	0.6232
1.145	0.6	500	0.8876	0.6459
0.916	0.62	520	0.9421	0.6256
0.9833	0.65	540	0.8567	0.6400
0.7927	0.67	560	0.9217	0.6579
1.0793	0.69	580	0.8519	0.6663
0.9642	0.72	600	0.8192	0.6758
0.9433	0.74	620	0.8412	0.6627
0.5469	0.77	640	0.8177	0.6794
0.9734	0.79	660	0.8130	0.6794
0.7071	0.81	680	0.9579	0.6579
0.9347	0.84	700	0.7870	0.6878
0.8024	0.86	720	0.9395	0.6411
0.9428	0.88	740	0.8817	0.6663
0.8024	0.91	760	0.7953	0.6974
0.7028	0.93	780	0.8279	0.6890
0.8357	0.96	800	0.8291	0.6950
0.8304	0.98	820	0.7778	0.6914
0.8157	1.0	840	0.7726	0.7081
0.5431	1.03	860	0.7628	0.7117
0.6816	1.05	880	0.7928	0.7057
0.7498	1.08	900	0.7732	0.7069
0.6785	1.1	920	0.8298	0.7033
0.5951	1.12	940	0.8020	0.7225
0.4782	1.15	960	0.7541	0.7297
0.6825	1.17	980	0.9647	0.6687
0.5793	1.2	1000	0.9230	0.7057
0.4872	1.22	1020	0.8361	0.7141
0.5422	1.24	1040	0.8629	0.7225
0.5671	1.27	1060	0.8376	0.7261
0.6285	1.29	1080	0.7478	0.7524
0.6647	1.32	1100	0.8001	0.7261
0.7846	1.34	1120	0.7344	0.7428
0.6362	1.36	1140	0.6980	0.75
0.4498	1.39	1160	0.7186	0.7536
0.5002	1.41	1180	0.7003	0.7656
0.5366	1.44	1200	0.7670	0.7404
0.6324	1.46	1220	0.7095	0.7536
0.6089	1.48	1240	0.7791	0.7428
0.5368	1.51	1260	0.6889	0.7679
0.6087	1.53	1280	0.7078	0.7620
0.5514	1.55	1300	0.7207	0.7644
0.6472	1.58	1320	0.6864	0.7691
0.5265	1.6	1340	0.7005	0.7679
0.5339	1.63	1360	0.6983	0.7667
0.5886	1.65	1380	0.7075	0.7739
0.4888	1.67	1400	0.6967	0.7715
0.5152	1.7	1420	0.6864	0.7679
0.3835	1.72	1440	0.7277	0.7608
0.5798	1.75	1460	0.7023	0.7644
0.4918	1.77	1480	0.7007	0.7667
0.6233	1.79	1500	0.7429	0.7596
0.5679	1.82	1520	0.6910	0.7488
0.6393	1.84	1540	0.7151	0.7512
0.6167	1.87	1560	0.6984	0.7512
0.4379	1.89	1580	0.6685	0.7656
0.4348	1.91	1600	0.6662	0.7703
0.5685	1.94	1620	0.6648	0.7691
0.4126	1.96	1640	0.6656	0.7667
0.4846	1.99	1660	0.6651	0.7667

Framework versions

Transformers 4.22.0.dev0
Pytorch 1.12.1+cu113
Datasets 2.4.1.dev0
Tokenizers 0.12.1

jungjongho/wav2vec2-xlsr-korean-speech-emotion-recognition

wav2vec2-xlsr-korean-speech-emotion-recognition

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions