ssc-ady-mms-model-mix-adapt-max-longcv

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: 12.1108
• Cer: 0.8989
• Wer: 1.0

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: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
3.7443	0.2717	200	3.2557	0.9926	1.0
0.7396	0.5435	400	0.5329	0.1720	0.8128
0.5625	0.8152	600	0.4470	0.1455	0.7283
0.5028	1.0870	800	0.4204	0.1425	0.7189
0.4808	1.3587	1000	0.4070	0.1394	0.7088
0.4613	1.6304	1200	0.3806	0.1298	0.6834
0.4574	1.9022	1400	0.3863	0.1286	0.6820
0.4247	2.1739	1600	0.3623	0.1257	0.6674
0.4233	2.4457	1800	0.3794	0.1330	0.6916
0.4101	2.7174	2000	0.3497	0.1213	0.6532
0.4057	2.9891	2200	0.3569	0.1278	0.6727
0.4033	3.2609	2400	0.3481	0.1232	0.6552
0.4023	3.5326	2600	0.3437	0.1206	0.6482
0.3862	3.8043	2800	0.3484	0.1204	0.6566
0.4157	4.0761	3000	0.3586	0.1264	0.6568
0.4017	4.3478	3200	0.3744	0.1282	0.6628
0.4148	4.6196	3400	0.3591	0.1244	0.6530
0.4446	4.8913	3600	0.3905	0.1289	0.6561
0.4784	5.1630	3800	0.3949	0.1315	0.6882
0.4966	5.4348	4000	0.4179	0.1328	0.6741
0.5704	5.7065	4200	0.4569	0.1335	0.6825
0.6411	5.9783	4400	0.6167	0.1527	0.7170
0.8632	6.25	4600	0.9967	0.2227	0.8538
2.0204	6.5217	4800	2.4109	0.8268	1.0
2.7834	6.7935	5000	2.6203	0.9210	1.0
2.7752	7.0652	5200	2.7062	0.9524	1.0
2.8283	7.3370	5400	2.6976	0.9280	1.0
2.8589	7.6087	5600	2.7020	0.9235	1.0
3.0579	7.8804	5800	2.9812	0.8943	0.9998
3.5998	8.1522	6000	3.3562	0.8770	0.9993
3.98	8.4239	6200	3.8742	0.7393	0.9950
4.6409	8.6957	6400	4.5709	0.7550	0.9945
5.2842	8.9674	6600	5.5003	0.6588	0.9923
6.9814	9.2391	6800	7.0443	0.8092	1.0
8.6704	9.5109	7000	9.0629	0.7697	1.0
10.1129	9.7826	7200	10.1659	0.7456	1.0
10.964	10.0543	7400	10.9963	0.7948	1.0
11.7005	10.3261	7600	11.9229	0.7517	1.0
12.0909	10.5978	7800	12.3057	0.9080	1.0
12.385	10.8696	8000	12.3060	0.9079	1.0
12.1537	11.1413	8200	12.1111	0.8990	1.0
12.0604	11.4130	8400	12.1109	0.8993	1.0
11.9687	11.6848	8600	12.1105	0.8991	1.0
12.0688	11.9565	8800	12.1108	0.8990	1.0
11.9894	12.2283	9000	12.1103	0.8990	1.0
12.1037	12.5	9200	12.1110	0.8990	1.0
11.9453	12.7717	9400	12.1102	0.8986	1.0
11.959	13.0435	9600	12.1109	0.8990	1.0
12.1515	13.3152	9800	12.1104	0.8991	1.0
11.8908	13.5870	10000	12.1108	0.8992	1.0
12.0349	13.8587	10200	12.1108	0.8995	1.0
12.0769	14.1304	10400	12.1110	0.8993	1.0
11.8601	14.4022	10600	12.1110	0.8990	1.0
12.2427	14.6739	10800	12.1108	0.8989	1.0
11.9567	14.9457	11000	12.1108	0.8989	1.0

Framework versions

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