llava-1.5_RSI

This model is a fine-tuned version of llava-hf/llava-1.5-7b-hf on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.9393

Model description

INCORRECT PROMPT FORMATING WAS USED: [INST] \nDescribe the aerial image:\n[\INST]

INSTEAD OF: USER: \nDescribe this aerial image in detail.\nASSISTANT:

Intended uses & limitations

More information needed

Training and evaluation data

RSI-CAP DATASET

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2.5e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5
• training_steps: 200

Training results

Training Loss	Epoch	Step	Validation Loss
1.2226	0.1546	25	1.2343
1.0431	0.3091	50	1.0435
0.9912	0.4637	75	0.9912
1.0529	0.6182	100	0.9578
0.9986	0.7728	125	0.9454
0.9295	0.9274	150	0.9419
0.8554	1.0819	175	0.9391
0.993	1.2365	200	0.9393

Framework versions

• PEFT 0.12.0
• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1