mms-1b-all-lin-Fleurs_AMMI_AFRIVOICE_LRSC-1hrs-v1

This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7464
• Wer: 0.4690
• Cer: 0.1857

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: 0.001
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
20.2969	1.0	20	15.4344	6.0492	1.9074
8.6598	2.0	40	4.6574	1.0	0.9999
4.0919	3.0	60	3.6494	1.0177	0.8475
3.5479	4.0	80	3.1117	0.9919	0.7177
2.9834	5.0	100	2.9935	0.9976	0.6736
2.9875	6.0	120	2.6889	0.9079	0.6285
2.4481	7.0	140	2.5176	0.8753	0.5858
2.1694	8.0	160	2.3393	0.8359	0.5096
1.9925	9.0	180	2.1131	0.8073	0.4976
1.7516	10.0	200	2.1095	0.7797	0.4831
1.5563	11.0	220	1.9102	0.7538	0.4248
1.3814	12.0	240	1.8076	0.7351	0.4033
1.2278	13.0	260	1.7135	0.7027	0.3646
1.0993	14.0	280	1.6182	0.6884	0.3347
0.992	15.0	300	1.5669	0.6584	0.3120
0.8873	16.0	320	1.5504	0.6491	0.2943
1.0836	17.0	340	1.5024	0.6229	0.2854
0.9552	18.0	360	1.4542	0.6003	0.2668
0.6556	19.0	380	1.4784	0.5956	0.2609
0.6097	20.0	400	1.4696	0.5847	0.2499
0.5452	21.0	420	1.4980	0.5749	0.2443
0.5073	22.0	440	1.4862	0.5674	0.2359
0.5055	23.0	460	1.4599	0.5463	0.2297
0.4414	24.0	480	1.4955	0.5574	0.2279
0.4092	25.0	500	1.4877	0.5516	0.2280
0.3752	26.0	520	1.4439	0.5471	0.2241
0.3616	27.0	540	1.4631	0.5346	0.2196
0.3228	28.0	560	1.4857	0.5285	0.2142
0.3132	29.0	580	1.4594	0.5257	0.2156
0.2877	30.0	600	1.5222	0.5288	0.2123
0.2993	31.0	620	1.5179	0.5294	0.2132
0.2699	32.0	640	1.5192	0.5233	0.2130
0.2635	33.0	660	1.5113	0.5144	0.2090
0.2331	34.0	680	1.5547	0.5217	0.2098
0.2374	35.0	700	1.5334	0.5034	0.2026
0.2317	36.0	720	1.5067	0.5143	0.2057
0.2121	37.0	740	1.5589	0.5067	0.2051
0.2023	38.0	760	1.5842	0.5053	0.2048
0.2002	39.0	780	1.6043	0.5167	0.2033
0.1981	40.0	800	1.5686	0.5096	0.2049
0.1902	41.0	820	1.6279	0.5056	0.2026
0.1743	42.0	840	1.5759	0.5089	0.2022
0.1892	43.0	860	1.6137	0.5065	0.1999
0.1641	44.0	880	1.6219	0.4974	0.1981
0.4117	45.0	900	1.5925	0.4973	0.1964
0.1497	46.0	920	1.6325	0.4967	0.1973
0.1592	47.0	940	1.5893	0.4992	0.1976
0.1399	48.0	960	1.6137	0.4985	0.1980
0.145	49.0	980	1.5947	0.4862	0.1967
0.129	50.0	1000	1.6438	0.4972	0.1977
0.1198	51.0	1020	1.6853	0.5019	0.1970
0.1266	52.0	1040	1.6279	0.4993	0.1967
0.1176	53.0	1060	1.6381	0.4965	0.1973
0.1157	54.0	1080	1.6612	0.4891	0.1942
0.1126	55.0	1100	1.6502	0.4867	0.1946
0.1078	56.0	1120	1.7070	0.4874	0.1939
0.1074	57.0	1140	1.6772	0.4939	0.1961
0.1145	58.0	1160	1.7170	0.4863	0.1927
0.0998	59.0	1180	1.7114	0.4850	0.1923
0.1024	60.0	1200	1.7010	0.4838	0.1932
0.1089	61.0	1220	1.6970	0.4762	0.1897
0.0886	62.0	1240	1.7086	0.4762	0.1898
0.0971	63.0	1260	1.6991	0.4776	0.1901
0.0932	64.0	1280	1.6583	0.4761	0.1889
0.0847	65.0	1300	1.7442	0.4809	0.1887
0.0913	66.0	1320	1.7359	0.4772	0.1910
0.0892	67.0	1340	1.6782	0.4797	0.1902
0.0818	68.0	1360	1.7167	0.4844	0.1920
0.0857	69.0	1380	1.7230	0.4797	0.1897
0.0878	70.0	1400	1.6981	0.4860	0.1928
0.0739	71.0	1420	1.7313	0.4735	0.1862
0.0757	72.0	1440	1.7053	0.4767	0.1883
0.0716	73.0	1460	1.7432	0.4792	0.1884
0.075	74.0	1480	1.7191	0.4770	0.1878
0.0666	75.0	1500	1.7157	0.4760	0.1895
0.0701	76.0	1520	1.7501	0.4779	0.1873
0.0743	77.0	1540	1.7318	0.4750	0.1873
0.0676	78.0	1560	1.7164	0.4736	0.1867
0.062	79.0	1580	1.7338	0.4658	0.1851
0.0637	80.0	1600	1.7325	0.4730	0.1876
0.0697	81.0	1620	1.7165	0.4693	0.1861
0.0584	82.0	1640	1.7529	0.4700	0.1863
0.0597	83.0	1660	1.7759	0.4687	0.1850
0.0658	84.0	1680	1.7439	0.4683	0.1860
0.0555	85.0	1700	1.7395	0.4721	0.1870
0.0617	86.0	1720	1.7525	0.4685	0.1855
0.0641	87.0	1740	1.7330	0.4705	0.1849
0.0542	88.0	1760	1.7356	0.4704	0.1862
0.0582	89.0	1780	1.7464	0.4690	0.1857

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0