quanliang/LLaMA32-11B-FullFT-FractureMorph-AugmentedDataset

fracture-morph-sft

This model is a fine-tuned version of meta-llama/Llama-3.2-11B-Vision-Instruct on the fracture_morph_gen dataset. It achieves the following results on the evaluation set:

• Loss: 2.3144

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: 1.5e-05
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 4
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
1.6931	0.3901	200	1.6400
1.5323	0.7801	400	1.5292
1.1249	1.1697	600	1.4814
1.141	1.5597	800	1.4272
1.0932	1.9498	1000	1.3718
0.5647	2.3393	1200	1.5874
0.5141	2.7294	1400	1.5496
0.1684	3.1190	1600	1.9035
0.1714	3.5090	1800	1.9125
0.1667	3.8991	2000	1.9118
0.0433	4.2886	2200	2.3167
0.0424	4.6787	2400	2.3123

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1