ssc-ady-mms-model-mix-adapt-max3

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: 0.5765
• Cer: 0.1402
• Wer: 0.6993

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 6
• 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.9047	0.2710	200	0.6095	0.1972	0.8507
0.5954	0.5420	400	0.4688	0.1525	0.7410
0.5498	0.8130	600	0.4338	0.1466	0.7218
0.4841	1.0840	800	0.4005	0.1352	0.6921
0.4703	1.3550	1000	0.3867	0.1314	0.6815
0.4535	1.6260	1200	0.3772	0.1311	0.6870
0.4435	1.8970	1400	0.3557	0.1252	0.6715
0.4235	2.1680	1600	0.3442	0.1216	0.6588
0.4034	2.4390	1800	0.3496	0.1242	0.6621
0.3985	2.7100	2000	0.3347	0.1199	0.6497
0.3996	2.9810	2200	0.3365	0.1190	0.6456
0.3794	3.2520	2400	0.3270	0.1203	0.6463
0.3742	3.5230	2600	0.3236	0.1170	0.6393
0.3818	3.7940	2800	0.3254	0.1152	0.6338
0.3718	4.0650	3000	0.3336	0.1183	0.6314
0.3811	4.3360	3200	0.3318	0.1194	0.6358
0.3798	4.6070	3400	0.3212	0.1164	0.6310
0.3684	4.8780	3600	0.3324	0.1214	0.6554
0.3586	5.1491	3800	0.3205	0.1141	0.6185
0.3617	5.4201	4000	0.3262	0.1163	0.6293
0.395	5.6911	4200	0.3667	0.1171	0.6379
0.4125	5.9621	4400	0.3930	0.1217	0.6552
0.4441	6.2331	4600	0.4093	0.1314	0.6719
0.5343	6.5041	4800	0.4458	0.1372	0.6988
0.6275	6.7751	5000	0.5100	0.1455	0.7081
0.684	7.0461	5200	0.5860	0.1467	0.7232
0.6562	7.3171	5400	0.5362	0.1548	0.7259
0.6384	7.5881	5600	0.5052	0.1430	0.7050
0.621	7.8591	5800	0.5097	0.1475	0.7110
0.6395	8.1301	6000	0.5098	0.1451	0.7048
0.6656	8.4011	6200	0.5299	0.1447	0.7057
0.6787	8.6721	6400	0.5523	0.1523	0.7259
0.7118	8.9431	6600	0.5753	0.1401	0.6995
0.6841	9.2141	6800	0.5741	0.1401	0.6983
0.7023	9.4851	7000	0.5794	0.1388	0.6930
0.687	9.7561	7200	0.5765	0.1402	0.6993

Framework versions

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