ssc-bas-mms-model-mix-adapt-max-longcv2

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: 0.2587
• Cer: 0.1279
• Wer: 0.4567

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: 8
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
3.0661	0.8016	200	2.9063	0.9241	0.9982
2.6227	1.6012	400	2.5538	0.8764	0.9879
2.37	2.4008	600	2.3405	0.8121	0.9507
2.0787	3.2004	800	1.8165	0.7173	0.9465
1.4179	4.0	1000	1.1348	0.4501	0.8524
1.0596	4.8016	1200	0.7486	0.2924	0.7123
0.8973	5.6012	1400	0.7237	0.2807	0.6842
0.8	6.4008	1600	0.6260	0.2360	0.6437
0.7781	7.2004	1800	0.5515	0.2156	0.6349
0.7073	8.0	2000	0.4825	0.1989	0.5913
0.7165	8.8016	2200	0.4225	0.1764	0.5626
0.6925	9.6012	2400	0.3848	0.1649	0.5309
0.6639	10.4008	2600	0.4002	0.1678	0.5420
0.6281	11.2004	2800	0.3676	0.1580	0.5215
0.5809	12.0	3000	0.3508	0.1552	0.5094
0.5696	12.8016	3200	0.3709	0.1579	0.5218
0.5627	13.6012	3400	0.3586	0.1548	0.5121
0.52	14.4008	3600	0.3307	0.1499	0.5079
0.5261	15.2004	3800	0.3141	0.1408	0.4882
0.4907	16.0	4000	0.3317	0.1535	0.4903
0.4883	16.8016	4200	0.2957	0.1403	0.4815
0.4508	17.6012	4400	0.3082	0.1395	0.4888
0.4399	18.4008	4600	0.3386	0.1435	0.4858
0.4519	19.2004	4800	0.2933	0.1392	0.4843
0.4395	20.0	5000	0.3248	0.1452	0.4906
0.4344	20.8016	5200	0.2868	0.1363	0.4761
0.3868	21.6012	5400	0.2769	0.1319	0.4646
0.385	22.4008	5600	0.2918	0.1350	0.4710
0.377	23.2004	5800	0.2655	0.1311	0.4634
0.3728	24.0	6000	0.2750	0.1327	0.4664
0.3377	24.8016	6200	0.2808	0.1311	0.4634
0.3606	25.6012	6400	0.2683	0.1307	0.4589
0.3368	26.4008	6600	0.2666	0.1309	0.4595
0.3563	27.2004	6800	0.2600	0.1303	0.4649
0.3224	28.0	7000	0.2585	0.1277	0.4604
0.3132	28.8016	7200	0.2596	0.1270	0.4555
0.3335	29.6012	7400	0.2587	0.1279	0.4567

Framework versions

• Transformers 4.52.1
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.21.4