MMS_10langs_simultane

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: 0.3003
• Wer: 0.3716
• Bleu: 0.4691
• Rouge1: 0.6871
• Rouge2: 0.5597
• Rougel: 0.6859

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 10
• total_train_batch_size: 40
• 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	Bleu	Rouge1	Rouge2	Rougel
2.2952	1.0	450	0.3890	0.4483	0.3755	0.6177	0.4654	0.6160
0.631	2.0	900	0.3822	0.4494	0.3774	0.6166	0.4639	0.6149
0.6019	3.0	1350	0.3704	0.4405	0.3805	0.6263	0.4729	0.6248
0.5863	4.0	1800	0.3545	0.4255	0.4021	0.6419	0.4948	0.6399
0.5718	5.0	2250	0.3480	0.4241	0.4044	0.6414	0.4948	0.6397
0.561	6.0	2700	0.3464	0.4195	0.4081	0.6465	0.5013	0.6448
0.5568	7.0	3150	0.3487	0.4214	0.4055	0.6433	0.4976	0.6414
0.547	8.0	3600	0.3442	0.4180	0.4101	0.6443	0.5009	0.6427
0.5462	9.0	4050	0.3395	0.4210	0.4101	0.6478	0.5042	0.6462
0.5354	10.0	4500	0.3377	0.4136	0.4141	0.6510	0.5083	0.6493
0.5358	11.0	4950	0.3405	0.4141	0.4107	0.6495	0.5059	0.6481
0.5295	12.0	5400	0.3399	0.4138	0.4167	0.6501	0.5070	0.6485
0.5261	13.0	5850	0.3351	0.4084	0.4222	0.6557	0.5144	0.6540
0.5238	14.0	6300	0.3345	0.4110	0.4199	0.6513	0.5084	0.6494
0.52	15.0	6750	0.3346	0.4104	0.4198	0.6511	0.5083	0.6498
0.5126	16.0	7200	0.3332	0.4127	0.4181	0.6515	0.5087	0.6498
0.5128	17.0	7650	0.3331	0.4043	0.4263	0.6597	0.5199	0.6578
0.5069	18.0	8100	0.3284	0.4024	0.4290	0.6613	0.5231	0.6597
0.5074	19.0	8550	0.3351	0.4090	0.4207	0.6565	0.5168	0.6547
0.4996	20.0	9000	0.3342	0.4035	0.4308	0.6565	0.5177	0.6544
0.4991	21.0	9450	0.3281	0.4030	0.4275	0.6678	0.5312	0.6664
0.4931	22.0	9900	0.3268	0.4075	0.4276	0.6526	0.5140	0.6508
0.4959	23.0	10350	0.3290	0.4043	0.4282	0.6594	0.5211	0.6580
0.4937	24.0	10800	0.3304	0.4115	0.4215	0.6516	0.5108	0.6495
0.4889	25.0	11250	0.3226	0.3999	0.4333	0.6605	0.5221	0.6587
0.4871	26.0	11700	0.3221	0.3974	0.4357	0.6618	0.5251	0.6604
0.4828	27.0	12150	0.3338	0.4048	0.4310	0.6510	0.5130	0.6499
0.4843	28.0	12600	0.3206	0.3974	0.4343	0.6661	0.5283	0.6642
0.4782	29.0	13050	0.3215	0.3994	0.4357	0.6603	0.5234	0.6587
0.4738	30.0	13500	0.3218	0.3964	0.4375	0.6637	0.5254	0.6618
0.4735	31.0	13950	0.3231	0.4000	0.4317	0.6632	0.5260	0.6610
0.4707	32.0	14400	0.3183	0.3917	0.4421	0.6705	0.5355	0.6688
0.4692	33.0	14850	0.3198	0.3985	0.4351	0.6663	0.5312	0.6651
0.4672	34.0	15300	0.3137	0.3932	0.4395	0.6717	0.5394	0.6699
0.4668	35.0	15750	0.3135	0.3947	0.4391	0.6676	0.5321	0.6657
0.4645	36.0	16200	0.3169	0.3958	0.4397	0.6672	0.5324	0.6652
0.4645	37.0	16650	0.3147	0.3923	0.4402	0.6694	0.5338	0.6678
0.4617	38.0	17100	0.3160	0.3924	0.4448	0.6668	0.5301	0.6647
0.4592	39.0	17550	0.3132	0.3883	0.4477	0.6736	0.5399	0.6718
0.456	40.0	18000	0.3108	0.3888	0.4474	0.6729	0.5391	0.6710
0.4562	41.0	18450	0.3138	0.3921	0.4435	0.6680	0.5340	0.6662
0.4507	42.0	18900	0.3137	0.3918	0.4426	0.6723	0.5385	0.6707
0.4521	43.0	19350	0.3147	0.3899	0.4479	0.6687	0.5335	0.6671
0.4492	44.0	19800	0.3121	0.3892	0.4473	0.6693	0.5353	0.6679
0.4481	45.0	20250	0.3109	0.3903	0.4474	0.6696	0.5353	0.6682
0.4458	46.0	20700	0.3146	0.3861	0.4505	0.6733	0.5397	0.6720
0.4469	47.0	21150	0.3107	0.3877	0.4495	0.6731	0.5407	0.6717
0.446	48.0	21600	0.3100	0.3877	0.4500	0.6742	0.5426	0.6728
0.4453	49.0	22050	0.3099	0.3885	0.4506	0.6732	0.5410	0.6715
0.4412	50.0	22500	0.3136	0.3860	0.4485	0.6779	0.5459	0.6763
0.4396	51.0	22950	0.3181	0.3879	0.4488	0.6701	0.5377	0.6688
0.4371	52.0	23400	0.3102	0.3860	0.4499	0.6772	0.5446	0.6757
0.4376	53.0	23850	0.3098	0.3884	0.4489	0.6727	0.5391	0.6704
0.4356	54.0	24300	0.3096	0.3837	0.4552	0.6731	0.5426	0.6716
0.4324	55.0	24750	0.3115	0.3832	0.4548	0.6801	0.5497	0.6787
0.4331	56.0	25200	0.3089	0.3869	0.4527	0.6756	0.5458	0.6740
0.4301	57.0	25650	0.3084	0.3848	0.4541	0.6778	0.5467	0.6763
0.4307	58.0	26100	0.3128	0.3823	0.4553	0.6759	0.5460	0.6741
0.43	59.0	26550	0.3070	0.3813	0.4559	0.6799	0.5502	0.6782
0.4244	60.0	27000	0.3076	0.3833	0.4539	0.6781	0.5458	0.6767
0.4236	61.0	27450	0.3109	0.3846	0.4554	0.6748	0.5449	0.6735
0.4257	62.0	27900	0.3085	0.3814	0.4544	0.6808	0.5496	0.6793
0.4226	63.0	28350	0.3068	0.3837	0.4537	0.6776	0.5454	0.6760
0.4239	64.0	28800	0.3052	0.3821	0.4561	0.6798	0.5491	0.6783
0.4206	65.0	29250	0.3095	0.3820	0.4548	0.6762	0.5457	0.6749
0.4212	66.0	29700	0.3055	0.3822	0.4541	0.6771	0.5457	0.6756
0.4191	67.0	30150	0.3063	0.3787	0.4605	0.6809	0.5520	0.6797
0.4137	68.0	30600	0.3056	0.3792	0.4577	0.6818	0.5536	0.6804
0.4156	69.0	31050	0.3023	0.3783	0.4602	0.6808	0.5507	0.6793
0.413	70.0	31500	0.3034	0.3785	0.4597	0.6821	0.5530	0.6803
0.4112	71.0	31950	0.3022	0.3805	0.4577	0.6804	0.5509	0.6790
0.4116	72.0	32400	0.3031	0.3793	0.4586	0.6794	0.5496	0.6782
0.4101	73.0	32850	0.3021	0.3766	0.4632	0.6819	0.5540	0.6804
0.4073	74.0	33300	0.3039	0.3788	0.4608	0.6816	0.5526	0.6805
0.4071	75.0	33750	0.3076	0.3776	0.4622	0.6823	0.5529	0.6809
0.4063	76.0	34200	0.3034	0.3776	0.4624	0.6794	0.5496	0.6783
0.407	77.0	34650	0.3058	0.3755	0.4637	0.6816	0.5524	0.6800
0.4039	78.0	35100	0.3048	0.3760	0.4620	0.6813	0.5510	0.6800
0.4052	79.0	35550	0.3063	0.3777	0.4620	0.6822	0.5526	0.6811
0.4066	80.0	36000	0.3029	0.3782	0.4612	0.6804	0.5489	0.6792
0.4036	81.0	36450	0.3041	0.3781	0.4603	0.6829	0.5520	0.6815
0.3987	82.0	36900	0.3048	0.3760	0.4625	0.6838	0.5549	0.6824
0.4007	83.0	37350	0.3008	0.3736	0.4659	0.6863	0.5573	0.6849
0.4016	84.0	37800	0.3011	0.3739	0.4653	0.6865	0.5586	0.6849
0.3981	85.0	38250	0.3007	0.3731	0.4666	0.6864	0.5588	0.6845
0.3986	86.0	38700	0.3005	0.3719	0.4670	0.6860	0.5583	0.6846
0.3955	87.0	39150	0.3002	0.3737	0.4656	0.6857	0.5576	0.6844
0.3942	88.0	39600	0.2999	0.3729	0.4672	0.6860	0.5596	0.6848
0.3951	89.0	40050	0.3021	0.3756	0.4646	0.6826	0.5536	0.6811
0.3963	90.0	40500	0.3000	0.3713	0.4683	0.6880	0.5624	0.6867
0.3941	91.0	40950	0.2998	0.3716	0.4677	0.6884	0.5622	0.6872
0.3913	92.0	41400	0.3018	0.3722	0.4687	0.6871	0.5600	0.6859
0.3938	93.0	41850	0.3009	0.3726	0.4687	0.6857	0.5583	0.6845
0.3912	94.0	42300	0.3003	0.3717	0.4679	0.6879	0.5617	0.6868
0.3921	95.0	42750	0.2997	0.3712	0.4693	0.6876	0.5621	0.6863
0.392	96.0	43200	0.3012	0.3710	0.4700	0.6884	0.5617	0.6870
0.3885	97.0	43650	0.3015	0.3713	0.4694	0.6875	0.5602	0.6861
0.3893	98.0	44100	0.3006	0.3718	0.4692	0.6868	0.5598	0.6856
0.3872	99.0	44550	0.3002	0.3712	0.4692	0.6874	0.5602	0.6859
0.3903	100.0	45000	0.3003	0.3716	0.4691	0.6871	0.5597	0.6859

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0