from transformers import LlavaForConditionalGeneration, LlavaProcessor
import torch

# Load fine-tuned model
model = LlavaForConditionalGeneration.from_pretrained("aparaselli/llava-7b-ft-unlin-10tok", trust_remote_code=True)

# Load original processor (recommended approach)
processor = LlavaProcessor.from_pretrained("llava-hf/llava-1.5-7b-hf", trust_remote_code=True)

# Your inference code here...

Important Notes

Base Model: llava-hf/llava-1.5-7b-hf
Model Type: Vision-Language Model
Architecture: LlavaForConditionalGeneration
Processor: Use the original processor from llava-hf/llava-1.5-7b-hf

This repository contains only the fine-tuned model weights
Always use the processor from llava-hf/llava-1.5-7b-hf for best compatibility
The model was fine-tuned on top of the base model but uses the original tokenization and image processing

Training Details

Add details about your training process, dataset, and hyperparameters here.

Evaluation

Add evaluation results here.

Citation

If you use this model, please cite the original LLaVA paper and mention your fine-tuning work.