hubert-large-korean-finetuned-korspeech-ser2

This model is a fine-tuned version of team-lucid/hubert-large-korean on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.4456
Macro F1: 0.6275
Accuracy: 0.6278
Weighted f1: 0.6275
Micro f1: 0.6278
Weighted recall: 0.6278
Micro recall: 0.6278
Macro recall: 0.6278
Weighted precision: 0.6296
Micro precision: 0.6278
Macro precision: 0.6296

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: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
lr_scheduler_warmup_steps: 100
num_epochs: 12

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Accuracy	Weighted f1	Micro f1	Weighted recall	Micro recall	Macro recall	Weighted precision	Micro precision	Macro precision
1.3419	0.28	100	1.2665	0.4034	0.4276	0.4034	0.4276	0.4276	0.4276	0.4276	0.4520	0.4276	0.4520
1.2338	0.57	200	1.1790	0.4459	0.4638	0.4459	0.4638	0.4638	0.4638	0.4638	0.4618	0.4638	0.4618
1.1842	0.85	300	1.1540	0.4560	0.4972	0.4560	0.4972	0.4972	0.4972	0.4972	0.4966	0.4972	0.4966
1.1304	1.13	400	1.1268	0.5056	0.5249	0.5056	0.5249	0.5249	0.5249	0.5249	0.5269	0.5249	0.5269
1.0815	1.42	500	1.0745	0.5390	0.5447	0.5390	0.5447	0.5447	0.5447	0.5447	0.5396	0.5447	0.5396
1.0822	1.7	600	1.0715	0.5071	0.5270	0.5071	0.5270	0.5270	0.5270	0.5270	0.5260	0.5270	0.5260
1.0484	1.99	700	1.0213	0.5555	0.5632	0.5555	0.5632	0.5632	0.5632	0.5632	0.5573	0.5632	0.5573
0.9784	2.27	800	1.0601	0.5640	0.5739	0.5640	0.5739	0.5739	0.5739	0.5739	0.5715	0.5739	0.5715
0.9627	2.55	900	1.0287	0.5606	0.5746	0.5606	0.5746	0.5746	0.5746	0.5746	0.5714	0.5746	0.5714
0.9614	2.84	1000	0.9945	0.5705	0.5753	0.5705	0.5753	0.5753	0.5753	0.5753	0.5782	0.5753	0.5782
0.9379	3.12	1100	1.0166	0.5852	0.5881	0.5852	0.5881	0.5881	0.5881	0.5881	0.5899	0.5881	0.5899
0.8982	3.4	1200	1.0289	0.5685	0.5724	0.5685	0.5724	0.5724	0.5724	0.5724	0.5905	0.5724	0.5905
0.8651	3.69	1300	1.0100	0.5967	0.6001	0.5967	0.6001	0.6001	0.6001	0.6001	0.6005	0.6001	0.6005
0.9017	3.97	1400	1.0405	0.5702	0.5739	0.5702	0.5739	0.5739	0.5739	0.5739	0.5884	0.5739	0.5884
0.8152	4.26	1500	0.9874	0.6016	0.6030	0.6016	0.6030	0.6030	0.6030	0.6030	0.6090	0.6030	0.6090
0.8149	4.54	1600	0.9994	0.6001	0.6044	0.6001	0.6044	0.6044	0.6044	0.6044	0.6092	0.6044	0.6092
0.7978	4.82	1700	1.0319	0.5945	0.6080	0.5945	0.6080	0.6080	0.6080	0.6080	0.6093	0.6080	0.6093
0.7674	5.11	1800	1.0800	0.5884	0.5909	0.5884	0.5909	0.5909	0.5909	0.5909	0.6128	0.5909	0.6128
0.7126	5.39	1900	1.0071	0.6177	0.6200	0.6177	0.6200	0.6200	0.6200	0.6200	0.6229	0.6200	0.6229
0.7229	5.67	2000	1.0267	0.6141	0.6165	0.6141	0.6165	0.6165	0.6165	0.6165	0.6141	0.6165	0.6141
0.7272	5.96	2100	1.0179	0.6119	0.6143	0.6119	0.6143	0.6143	0.6143	0.6143	0.6147	0.6143	0.6147
0.6519	6.24	2200	1.0576	0.6246	0.6257	0.6246	0.6257	0.6257	0.6257	0.6257	0.6322	0.6257	0.6322
0.6287	6.52	2300	1.0537	0.6275	0.6307	0.6275	0.6307	0.6307	0.6307	0.6307	0.6382	0.6307	0.6382
0.6103	6.81	2400	1.0323	0.6305	0.6328	0.6305	0.6328	0.6328	0.6328	0.6328	0.6329	0.6328	0.6329
0.5639	7.09	2500	1.1021	0.6306	0.6335	0.6306	0.6335	0.6335	0.6335	0.6335	0.6336	0.6335	0.6336
0.5706	7.38	2600	1.1086	0.6328	0.6342	0.6328	0.6342	0.6342	0.6342	0.6342	0.6349	0.6342	0.6349
0.529	7.66	2700	1.1428	0.6194	0.6186	0.6194	0.6186	0.6186	0.6186	0.6186	0.6260	0.6186	0.6260
0.5336	7.94	2800	1.1523	0.6128	0.6136	0.6128	0.6136	0.6136	0.6136	0.6136	0.6131	0.6136	0.6131
0.4776	8.23	2900	1.3509	0.5922	0.5959	0.5922	0.5959	0.5959	0.5959	0.5959	0.6070	0.5959	0.6070
0.4603	8.51	3000	1.2143	0.6036	0.6023	0.6036	0.6023	0.6023	0.6023	0.6023	0.6058	0.6023	0.6058
0.4734	8.79	3100	1.2464	0.6056	0.6051	0.6056	0.6051	0.6051	0.6051	0.6051	0.6063	0.6051	0.6063
0.4358	9.08	3200	1.3027	0.6110	0.6108	0.6110	0.6108	0.6108	0.6108	0.6108	0.6178	0.6108	0.6178
0.3808	9.36	3300	1.3469	0.6265	0.6328	0.6265	0.6328	0.6328	0.6328	0.6328	0.6304	0.6328	0.6304
0.4184	9.65	3400	1.3317	0.6168	0.6165	0.6168	0.6165	0.6165	0.6165	0.6165	0.6220	0.6165	0.6220
0.3748	9.93	3500	1.3316	0.6232	0.625	0.6232	0.625	0.625	0.625	0.625	0.6344	0.625	0.6344
0.3785	10.21	3600	1.3792	0.6144	0.6158	0.6144	0.6158	0.6158	0.6158	0.6158	0.6172	0.6158	0.6172
0.3339	10.5	3700	1.4025	0.6263	0.6264	0.6263	0.6264	0.6264	0.6264	0.6264	0.6296	0.6264	0.6296
0.367	10.78	3800	1.3871	0.6108	0.6115	0.6108	0.6115	0.6115	0.6115	0.6115	0.6135	0.6115	0.6135
0.3307	11.06	3900	1.3996	0.6170	0.6179	0.6170	0.6179	0.6179	0.6179	0.6179	0.6181	0.6179	0.6181
0.3188	11.35	4000	1.4383	0.6251	0.6271	0.6251	0.6271	0.6271	0.6271	0.6271	0.6252	0.6271	0.6252
0.3129	11.63	4100	1.4338	0.6209	0.6214	0.6209	0.6214	0.6214	0.6214	0.6214	0.6217	0.6214	0.6217
0.3112	11.91	4200	1.4456	0.6275	0.6278	0.6275	0.6278	0.6278	0.6278	0.6278	0.6296	0.6278	0.6296

Framework versions

Transformers 4.30.2
Pytorch 2.0.0
Datasets 2.1.0
Tokenizers 0.13.3

byc3230/hubert-large-korean-finetuned-korspeech-ser2

hubert-large-korean-finetuned-korspeech-ser2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

为此模型做种