AgroNVILA (SFT)
Supervised fine-tuned checkpoint of AgroNVILA, built on NVILA-Lite-8B.
Training
- Corpus: AgroOmni, 268,119 question-answer pairs over 97,926 images
- Views: Ground 192,800 / Satellite 56,732 / UAV 18,587 QA pairs
- Stage: supervised fine-tuning only (no reinforcement learning)
- Code: https://github.com/Jrookie2005/AgroOmni
Results
AgroMind benchmark, 13 tasks: 59.52 overall micro-average. After GRPO alignment the same recipe reaches 59.86; that checkpoint is AgroOmni/AgroNVILA.
Usage
from llava.model.builder import load_pretrained_model
tokenizer, model, image_processor, _ = load_pretrained_model(
"AgroOmni/AgroNVILA-SFT", None, "llava_llama"
)
License
Apache-2.0, inherited from the NVILA-Lite-8B base model.
- Downloads last month
- 25
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for AgroOmni/AgroNVILA-SFT
Base model
Efficient-Large-Model/NVILA-Lite-8B