DurainKim/gym_color_large_old

ft_paligemma2_gym_color_large

This model is a fine-tuned version of google/paligemma2-3b-pt-224 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1356

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: 6
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 24
• optimizer: Use adamw_hf 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: 2
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3299	0.2020	100	0.0672
0.071	0.4040	200	0.0674
0.0615	0.6061	300	0.0792
0.0547	0.8081	400	0.0696
0.0524	1.0101	500	0.0683
0.035	1.2121	600	0.1216
0.029	1.4141	700	0.1179
0.0261	1.6162	800	0.1077
0.0198	1.8182	900	0.1219
0.0154	2.0202	1000	0.1440
0.0123	2.2222	1100	0.1356

Framework versions

• PEFT 0.14.0
• Transformers 4.47.0
• Pytorch 2.1.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0