bridenmj/wound-llama-adapter

Wound LLaMA-Adapter

This repository contains the wound-specific adapter delta used to adapt the 7B Large Vision Model (LVM) for longitudinal porcine wound image extrapolation.

The artifact is not a standalone model checkpoint. It contains only the 84 project-trained adapter tensors. Users must obtain the matching LVM base checkpoint and VQ image tokenizer separately.

Released artifact

Field	Value
File	wound_llama_adapter_delta_v1.pth
Format	wound_forecasting_adapter_delta_v1
Size	33,786,121 bytes
Adapter tensors	84
Base model	Emma02/LVM_ckpts
VQ tokenizer	Emma02/vqvae_ckpts

Required external artifacts

• Base LVM checkpoint: https://huggingface.co/Emma02/LVM_ckpts
• VQ tokenizer: https://huggingface.co/Emma02/vqvae_ckpts
• Project code: https://github.com/bridenmj/wound-forecasting

The upstream base-model and VQ-tokenizer weights are not mirrored in this repository.

Loading

Construct the matching LVM adapter architecture from the upstream base checkpoint before applying this delta.

import torch
from huggingface_hub import hf_hub_download

from wound_forecasting.llama_adapter import (
    load_adapter_delta,
)

adapter_path = hf_hub_download(
    repo_id="bridenmj/wound-llama-adapter",
    filename="wound_llama_adapter_delta_v1.pth",
)

adapter_package = torch.load(
    adapter_path,
    map_location="cpu",
    weights_only=False,
)

load_adapter_delta(
    model,
    adapter_package,
    expected_base_model="Emma02/LVM_ckpts",
)

model.eval()

The exact construction parameters are recorded in adapter_config.json.

Training data

The model was trained on processed longitudinal images derived from the public porcine wound-healing dataset:

• Dryad source: https://doi.org/10.5061/dryad.0rxwdbsbr
• Processed release: bridenmj/porcine-wound-forecasting-processed

Provenance and licensing

This repository distributes only the project-trained adapter delta. It does not redistribute the upstream LVM base model or VQ tokenizer.

Project license

The original project-trained weights released in this repository are licensed under the Creative Commons Attribution–NonCommercial 4.0 International license (CC BY-NC 4.0).

Third-party implementations, base models, tokenizers, and other external dependencies are not relicensed by this repository and remain subject to their respective licenses and terms. Any upstream license or notice files included in this repository continue to apply to the corresponding upstream materials.