mms-tamil-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: 0.5679
• Accuracy: 0.8235
• F1: 0.7568

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
0.6797	1.0	13	0.6920	0.5686	0.0
0.6731	2.0	26	0.6815	0.6667	0.4516
0.6339	3.0	39	0.5852	0.7451	0.6977
0.5413	4.0	52	0.5228	0.7843	0.7442
0.4976	5.0	65	0.6092	0.7451	0.6667
0.5311	6.0	78	0.4797	0.7843	0.7660
0.4852	7.0	91	0.5751	0.7255	0.6316
0.4213	8.0	104	0.5290	0.7647	0.7000
0.3725	9.0	117	0.6625	0.7059	0.5455
0.3615	10.0	130	0.5822	0.7647	0.6842
0.3544	11.0	143	0.6254	0.7451	0.6061
0.3403	12.0	156	0.7538	0.7255	0.5625
0.3213	13.0	169	0.4957	0.7647	0.7000
0.2983	14.0	182	0.6817	0.7647	0.6471
0.3201	15.0	195	0.7149	0.7451	0.6061
0.2984	16.0	208	0.4992	0.8039	0.7368
0.2923	17.0	221	0.5695	0.8235	0.7568
0.2573	18.0	234	0.5558	0.8039	0.7368
0.2383	19.0	247	0.5647	0.8235	0.7568
0.2281	20.0	260	0.5679	0.8235	0.7568

Framework versions

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