uaspeech-mms1ball-Nov30

This model is a fine-tuned version of facebook/mms-1b-all on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1373
• Wer: 0.5222

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: 3
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
45.2252	0.0063	100	3.7562	1.0
2.9851	0.0127	200	2.0328	0.8281
2.0067	0.0190	300	1.8614	0.6848
2.1249	0.0254	400	2.0929	0.7247
1.8786	0.0317	500	1.6966	0.6522
1.8295	0.0381	600	1.6550	0.6579
1.8186	0.0444	700	1.6961	0.6457
1.839	0.0508	800	1.7019	0.6642
1.8854	0.0571	900	1.6492	0.6455
2.0186	0.0635	1000	1.5746	0.6682
1.8061	0.0698	1100	1.5590	0.6345
1.6994	0.0761	1200	1.6544	0.6509
1.6444	0.0825	1300	1.6297	0.6115
1.7308	0.0888	1400	1.7002	0.6337
1.8179	0.0952	1500	1.5325	0.6179
1.7174	0.1015	1600	1.5668	0.6343
1.7043	0.1079	1700	1.5762	0.6415
1.5174	0.1142	1800	1.4478	0.5991
1.6108	0.1206	1900	1.4394	0.6054
1.6227	0.1269	2000	1.5185	0.5964
1.773	0.1333	2100	1.6450	0.6124
1.6231	0.1396	2200	1.3599	0.6016
1.6005	0.1459	2300	1.3782	0.5943
1.524	0.1523	2400	1.6099	0.5953
1.5881	0.1586	2500	1.4681	0.6160
1.5577	0.1650	2600	1.3934	0.6052
1.5556	0.1713	2700	1.3960	0.6099
1.609	0.1777	2800	1.4837	0.6008
1.3661	0.1840	2900	1.2992	0.5800
1.4908	0.1904	3000	1.3942	0.5844
1.6022	0.1967	3100	1.4685	0.5907
1.7082	0.2031	3200	1.4685	0.6214
1.5526	0.2094	3300	1.4331	0.6038
1.4424	0.2157	3400	1.3674	0.5791
1.3544	0.2221	3500	1.3960	0.5882
1.4868	0.2284	3600	1.3507	0.5882
1.5313	0.2348	3700	1.5923	0.5764
1.5763	0.2411	3800	1.3493	0.5681
1.3742	0.2475	3900	1.3979	0.5827
1.5003	0.2538	4000	1.4011	0.5886
1.5894	0.2602	4100	1.3652	0.5519
1.5496	0.2665	4200	1.2970	0.5860
1.6228	0.2729	4300	1.4056	0.5753
1.4173	0.2792	4400	1.3485	0.5515
1.5236	0.2856	4500	1.2558	0.5473
1.3976	0.2919	4600	1.3535	0.5505
1.4656	0.2982	4700	1.3634	0.5865
1.4089	0.3046	4800	1.3172	0.5458
1.2844	0.3109	4900	1.3553	0.5606
1.3984	0.3173	5000	1.2458	0.5679
1.4784	0.3236	5100	1.3808	0.5484
1.4339	0.3300	5200	1.3194	0.5578
1.6029	0.3363	5300	1.3393	0.5631
1.4778	0.3427	5400	1.3053	0.5639
1.5093	0.3490	5500	1.2234	0.5445
1.5348	0.3554	5600	1.2514	0.5583
1.5195	0.3617	5700	1.2954	0.5561
1.5035	0.3680	5800	1.1854	0.5435
1.532	0.3744	5900	1.2768	0.5486
1.5833	0.3807	6000	1.3575	0.5547
1.4789	0.3871	6100	1.2984	0.5418
1.3557	0.3934	6200	1.2763	0.5326
1.2738	0.3998	6300	1.3761	0.5351
1.454	0.4061	6400	1.3057	0.5325
1.403	0.4125	6500	1.3707	0.5662
1.5306	0.4188	6600	1.3330	0.5583
1.4576	0.4252	6700	1.2403	0.5349
1.4214	0.4315	6800	1.3175	0.5363
1.5853	0.4378	6900	1.2071	0.5281
1.522	0.4442	7000	1.2593	0.5387
1.2743	0.4505	7100	1.3154	0.5549
1.3778	0.4569	7200	1.2544	0.5365
1.2875	0.4632	7300	1.2312	0.5281
1.3484	0.4696	7400	1.2599	0.5340
1.5618	0.4759	7500	1.3038	0.5568
1.4391	0.4823	7600	1.2523	0.5355
1.4676	0.4886	7700	1.2205	0.5311
1.371	0.4950	7800	1.1996	0.5275
1.361	0.5013	7900	1.1635	0.5186
1.463	0.5076	8000	1.2232	0.5344
1.348	0.5140	8100	1.1407	0.5174
1.4943	0.5203	8200	1.3098	0.5361
1.3703	0.5267	8300	1.1393	0.5226
1.4188	0.5330	8400	1.1914	0.4995
1.3048	0.5394	8500	1.1555	0.5170
1.4468	0.5457	8600	1.1515	0.5149
1.2279	0.5521	8700	1.2047	0.5168
1.2367	0.5584	8800	1.1868	0.5208
1.3536	0.5648	8900	1.1398	0.5189
1.3496	0.5711	9000	1.2705	0.5176
1.4915	0.5774	9100	1.2007	0.5079
1.474	0.5838	9200	1.2192	0.5378
1.3093	0.5901	9300	1.1065	0.5125
1.1994	0.5965	9400	1.2120	0.5201
1.502	0.6028	9500	1.1464	0.5290
1.5327	0.6092	9600	1.2010	0.5384
1.4533	0.6155	9700	1.1614	0.5146
1.4554	0.6219	9800	1.1353	0.5155
1.3605	0.6282	9900	1.1788	0.5058
1.3606	0.6346	10000	1.1877	0.5178
1.3295	0.6409	10100	1.2601	0.5452
1.4172	0.6472	10200	1.1926	0.5083
1.4859	0.6536	10300	1.1426	0.5077
1.5264	0.6599	10400	1.1495	0.5121
1.3455	0.6663	10500	1.1535	0.5140
1.2375	0.6726	10600	1.1857	0.5087
1.2808	0.6790	10700	1.2156	0.5005
1.3358	0.6853	10800	1.1543	0.5075
1.361	0.6917	10900	1.2553	0.5290
1.3859	0.6980	11000	1.1458	0.5245
1.5061	0.7044	11100	1.1547	0.5399
1.2777	0.7107	11200	1.1677	0.5180
1.2645	0.7171	11300	1.2715	0.5191
1.3514	0.7234	11400	1.1811	0.5109
1.2733	0.7297	11500	1.1079	0.5134
1.2961	0.7361	11600	1.3171	0.5188
1.3744	0.7424	11700	1.1271	0.5106
1.2626	0.7488	11800	1.0821	0.5001
1.317	0.7551	11900	1.1156	0.5037
1.1732	0.7615	12000	1.1828	0.5287
1.3693	0.7678	12100	1.1063	0.5172
1.3227	0.7742	12200	1.1450	0.4948
1.41	0.7805	12300	1.0917	0.5079
1.2909	0.7869	12400	1.1321	0.5066
1.1509	0.7932	12500	1.1678	0.4921
1.3629	0.7995	12600	1.1482	0.4923
1.2959	0.8059	12700	1.1488	0.5064
1.3537	0.8122	12800	1.1356	0.5014
1.1468	0.8186	12900	1.0872	0.4970
1.3582	0.8249	13000	1.1221	0.4913
1.3872	0.8313	13100	1.2243	0.4925
1.4433	0.8376	13200	1.1536	0.5205
1.4121	0.8440	13300	1.1322	0.5050
1.0979	0.8503	13400	1.1252	0.4927
1.3989	0.8567	13500	1.0632	0.5035
1.1527	0.8630	13600	1.1041	0.4942
1.167	0.8693	13700	1.0895	0.4913
1.3945	0.8757	13800	1.1256	0.5050
1.2615	0.8820	13900	1.1768	0.5226
1.3221	0.8884	14000	1.2068	0.5050
1.4254	0.8947	14100	1.1757	0.5028
1.1403	0.9011	14200	1.1589	0.5161
1.304	0.9074	14300	1.1399	0.5140
1.3728	0.9138	14400	1.0997	0.5132
1.4233	0.9201	14500	1.0702	0.4948
1.1177	0.9265	14600	1.0930	0.4936
1.186	0.9328	14700	1.1956	0.4919
1.3834	0.9391	14800	1.1244	0.5007
1.3236	0.9455	14900	1.1373	0.5222

Framework versions

• Transformers 4.43.4
• Pytorch 2.4.1
• Datasets 3.0.0
• Tokenizers 0.19.1