llama3.2-3b-medium

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.6810

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.0835	0.5198	100	1.9059
1.8694	1.0396	200	1.8444
1.7751	1.5595	300	1.8023
1.7621	2.0793	400	1.7662
1.6951	2.5991	500	1.7426
1.6917	3.1189	600	1.7270
1.6354	3.6387	700	1.7141
1.6231	4.1585	800	1.7076
1.5944	4.6784	900	1.6972
1.5959	5.1982	1000	1.6925
1.5429	5.7180	1100	1.6878
1.5584	6.2378	1200	1.6880
1.5347	6.7576	1300	1.6839
1.5536	7.2775	1400	1.6818
1.5061	7.7973	1500	1.6823
1.5282	8.3171	1600	1.6814
1.5187	8.8369	1700	1.6809
1.5232	9.3567	1800	1.6809
1.5131	9.8765	1900	1.6810

Framework versions

• PEFT 0.12.0
• Transformers 4.45.0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.20.1