mms-malayalam-binary

This model is a fine-tuned version of facebook/mms-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1517
• Accuracy: 0.9773
• F1: 0.9787

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.6653	1.0	23	0.6820	0.7955	0.7805
0.5764	2.0	46	0.4134	0.9205	0.9213
0.3843	3.0	69	0.1825	0.9659	0.9677
0.2285	4.0	92	0.1695	0.9659	0.9677
0.1649	5.0	115	0.1108	0.9773	0.9787
0.1394	6.0	138	0.1390	0.9659	0.9677
0.1141	7.0	161	0.1329	0.9659	0.9677
0.0798	8.0	184	0.1043	0.9773	0.9787
0.0743	9.0	207	0.2201	0.9432	0.9451
0.0846	10.0	230	0.2657	0.9205	0.9213
0.0701	11.0	253	0.1470	0.9659	0.9677
0.082	12.0	276	0.1956	0.9545	0.9565
0.0145	13.0	299	0.1265	0.9773	0.9787
0.0191	14.0	322	0.1345	0.9773	0.9787
0.0355	15.0	345	0.1394	0.9773	0.9787
0.038	16.0	368	0.1432	0.9773	0.9787
0.0437	17.0	391	0.1467	0.9773	0.9787
0.0374	18.0	414	0.1494	0.9773	0.9787
0.0125	19.0	437	0.1511	0.9773	0.9787
0.0575	20.0	460	0.1517	0.9773	0.9787

Framework versions

• Transformers 4.53.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.21.2