Wav2Vec2_Fine_tuned_on_CremaD_Speech_Emotion_Recognition

This model is a fine-tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-english.

The dataset used to fine-tune the original pre-trained model is the CremaD dataset. This dataset provides 7442 samples of recordings from actors performing on 6 different emotions in English, which are:

emotions = ['angry', 'disgust', 'fearful', 'happy', 'neutral', 'sad']

It achieves the following results on the evaluation set:

• Loss: 0.6258
• Accuracy: 0.7890

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 2.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.7923	0.01	10	1.8102	0.2554
1.7712	0.03	20	1.7128	0.2560
1.6854	0.04	30	1.7213	0.2823
1.6129	0.05	40	1.5384	0.3851
1.5121	0.07	50	1.5442	0.3810
1.532	0.08	60	1.4817	0.4234
1.3681	0.09	70	1.6103	0.3474
1.6408	0.11	80	1.5118	0.3495
1.4527	0.12	90	1.3684	0.4671
1.3219	0.13	100	1.3871	0.4698
1.5121	0.15	110	1.4060	0.4328
1.4013	0.16	120	1.5057	0.4180
1.3605	0.17	130	1.3576	0.4348
1.3813	0.19	140	1.3194	0.4933
1.2232	0.2	150	1.2804	0.5114
1.3133	0.22	160	1.2345	0.5356
1.2686	0.23	170	1.2445	0.5161
1.2539	0.24	180	1.1071	0.5766
1.1747	0.26	190	1.2424	0.5060
1.1644	0.27	200	1.3082	0.4892
1.2624	0.28	210	1.3811	0.5155
1.2036	0.3	220	1.2410	0.5349
1.2191	0.31	230	1.0329	0.5988
1.1212	0.32	240	1.1005	0.5806
1.1243	0.34	250	1.2593	0.5262
1.1951	0.35	260	1.0575	0.5981
1.0971	0.36	270	1.1753	0.5565
1.0209	0.38	280	1.0568	0.5840
1.1628	0.39	290	1.1174	0.5793
1.1894	0.4	300	1.0343	0.6183
1.0605	0.42	310	1.1357	0.5578
1.0701	0.43	320	1.0726	0.6042
0.9606	0.44	330	1.2933	0.5222
0.9128	0.46	340	1.1310	0.5827
1.1218	0.47	350	1.1245	0.6102
0.9566	0.48	360	1.0386	0.6116
1.1211	0.5	370	0.9842	0.6324
1.2184	0.51	380	0.9250	0.6593
1.1452	0.52	390	0.9282	0.6573
0.7752	0.54	400	1.0523	0.6102
1.0063	0.55	410	0.9372	0.6364
1.1807	0.56	420	1.0236	0.6176
1.0624	0.58	430	0.9036	0.6606
1.1832	0.59	440	0.9229	0.6458
1.0186	0.6	450	0.8801	0.6707
0.8184	0.62	460	0.9526	0.6398
0.8863	0.63	470	0.8996	0.6761
0.9068	0.65	480	0.8378	0.7030
0.8077	0.66	490	0.9574	0.6694
0.9824	0.67	500	1.0673	0.6499
0.8002	0.69	510	0.8819	0.6922
0.9411	0.7	520	0.8553	0.6815
1.0061	0.71	530	0.9180	0.6673
0.7496	0.73	540	0.9676	0.6505
0.8208	0.74	550	0.9990	0.6519
0.9846	0.75	560	0.8613	0.6962
0.9968	0.77	570	0.8798	0.6949
0.9485	0.78	580	0.9894	0.6223
0.9165	0.79	590	0.9384	0.6465
0.9393	0.81	600	0.7944	0.7137
0.9086	0.82	610	0.8543	0.6767
0.9175	0.83	620	0.8039	0.6996
0.8692	0.85	630	0.8488	0.6949
0.759	0.86	640	0.8890	0.6895
1.0115	0.87	650	1.0963	0.6210
0.766	0.89	660	0.9505	0.6277
1.2062	0.9	670	0.8218	0.6962
0.8678	0.91	680	0.7918	0.7056
0.9055	0.93	690	0.7626	0.7204
0.7303	0.94	700	0.8733	0.6714
0.9239	0.95	710	0.8488	0.6962
0.8024	0.97	720	0.7996	0.7083
0.7927	0.98	730	0.8690	0.6821
0.8371	0.99	740	0.9029	0.6727
0.8419	1.01	750	0.7640	0.7211
0.5163	1.02	760	0.8040	0.7292
0.4603	1.03	770	0.7946	0.7211
0.7675	1.05	780	0.9796	0.6774
0.9771	1.06	790	0.7548	0.7433
0.6141	1.08	800	0.7334	0.7386
0.71	1.09	810	0.7037	0.7547
0.6074	1.1	820	0.8142	0.7137
1.0638	1.12	830	0.8786	0.7036
0.7303	1.13	840	0.7548	0.7292
0.5361	1.14	850	0.7000	0.7513
0.6014	1.16	860	0.8950	0.6902
0.5635	1.17	870	0.7070	0.75
0.5585	1.18	880	0.7612	0.7473
0.8462	1.2	890	1.0107	0.6761
0.6256	1.21	900	0.7899	0.7272
0.7361	1.22	910	0.7397	0.7312
0.5147	1.24	920	0.8835	0.7003
0.5843	1.25	930	0.8751	0.7016
0.5077	1.26	940	0.7542	0.7278
0.6421	1.28	950	0.8593	0.7090
0.7138	1.29	960	0.7012	0.7601
0.5414	1.3	970	0.7669	0.7372
0.662	1.32	980	0.7620	0.7272
0.6002	1.33	990	0.6881	0.7628
0.8094	1.34	1000	0.7783	0.7433
0.6081	1.36	1010	0.7272	0.75
0.5943	1.37	1020	0.7667	0.7440
0.6295	1.38	1030	0.7453	0.7399
0.6415	1.4	1040	0.7053	0.7560
0.4686	1.41	1050	0.8764	0.7171
0.5586	1.42	1060	0.7406	0.75
0.4292	1.44	1070	0.7160	0.7708
0.6343	1.45	1080	0.8051	0.7298
0.6209	1.47	1090	0.9153	0.7198
0.834	1.48	1100	0.7113	0.7614
0.5106	1.49	1110	0.7978	0.7352
0.6587	1.51	1120	0.7805	0.7440
0.5694	1.52	1130	0.7192	0.7587
0.6949	1.53	1140	0.7119	0.7614
0.4578	1.55	1150	0.7249	0.7594
0.6219	1.56	1160	0.7289	0.7554
0.6857	1.57	1170	0.6933	0.7587
0.631	1.59	1180	0.6719	0.7749
0.6944	1.6	1190	0.7028	0.7587
0.5063	1.61	1200	0.6815	0.7587
0.6884	1.63	1210	0.7068	0.7534
0.797	1.64	1220	0.7583	0.7426
0.5841	1.65	1230	0.7034	0.7446
0.7062	1.67	1240	0.7050	0.7513
0.7438	1.68	1250	0.6894	0.7560
0.6627	1.69	1260	0.6438	0.7769
0.4233	1.71	1270	0.6523	0.7695
0.5555	1.72	1280	0.6859	0.7634
0.7625	1.73	1290	0.7076	0.7513
0.6136	1.75	1300	0.6515	0.7769
0.5207	1.76	1310	0.6463	0.7708
0.5175	1.77	1320	0.6442	0.7762
0.6413	1.79	1330	0.6515	0.7742
0.7482	1.8	1340	0.6608	0.7735
0.5284	1.81	1350	0.6717	0.7681
0.7004	1.83	1360	0.6800	0.7628
0.7958	1.84	1370	0.6577	0.7769
0.3887	1.85	1380	0.6428	0.7829
0.4225	1.87	1390	0.6465	0.7809
0.7193	1.88	1400	0.6590	0.7776
0.5101	1.9	1410	0.6519	0.7789
0.7712	1.91	1420	0.6510	0.7789
0.3919	1.92	1430	0.6566	0.7809
0.4867	1.94	1440	0.6531	0.7755
0.5402	1.95	1450	0.6441	0.7789
0.7002	1.96	1460	0.6344	0.7809
0.5943	1.98	1470	0.6278	0.7870
0.5144	1.99	1480	0.6258	0.7890

Framework versions

• Transformers 4.40.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.1.dev0
• Tokenizers 0.15.2