Instructions to use ahmedtaha100/spinefairbench-generator with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use ahmedtaha100/spinefairbench-generator with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("stable-diffusion-v1-5", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("ahmedtaha100/spinefairbench-generator") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
SpineFairBench Generator Checkpoints
This repository provides SpineFairBench generator checkpoint artifacts for inspection and reproducibility support.
Do not use any generic Hugging Face auto-generated "Use this model" snippet for SpineFairBench regeneration. This LoRA is released for reviewer inspection and optional local img2img inference using the project command in the paired GitHub repository.
Recommended checkpoint
Use spinefairbench_sd15_lora.safetensors for inference inspection. This is a
LoRA-only export derived from the preserved epoch-2 checkpoint identified below.
Optional archival checkpoint
latest.pt is the preserved full PyTorch checkpoint whose bytes match the retained Stage-1 epoch-2 file. It is provided
for archival/reproducibility inspection. It contains training state such as
optimizer and discriminator components and is not the recommended inference
artifact.
Not included
Raw source radiographs, source masks, runnable training data, credentials, private run roots, and provider-client orchestration are not included.
Download
Recommended LoRA only:
hf download ahmedtaha100/spinefairbench-generator \
spinefairbench_sd15_lora.safetensors \
generator_config.yaml \
SHA256SUMS.recommended.txt \
--repo-type model \
--local-dir generator_assets
cd generator_assets
shasum -a 256 -c SHA256SUMS.recommended.txt
All generator assets, including the optional archival checkpoint:
hf download ahmedtaha100/spinefairbench-generator \
--repo-type model \
--local-dir generator_assets
Verify checksums after downloading all generator assets:
cd generator_assets
shasum -a 256 -c SHA256SUMS.txt
Release files
| File | Purpose |
|---|---|
spinefairbench_sd15_lora.safetensors |
Recommended LoRA-only checkpoint for inference inspection. |
latest.pt |
Optional archival PyTorch training checkpoint; not recommended for standard inference. |
generator_config.yaml |
Public generator settings used for inspection. |
SHA256SUMS.recommended.txt |
SHA256 checksums for the recommended LoRA-only inspection download. |
SHA256SUMS.txt |
SHA256 checksums for all released generator files, including latest.pt. |
Provenance clarification, 2026-09-10
The archival checkpoint SHA-256 is bc3f71631f07f3d0ca645808b0622f28da2c863b21622cab0ed65e8e44f329ff; the LoRA SHA-256 is 21bbfa4ee50ff389c3a49edc9fa39f4e1f7c691fad85c34cbcce595f62b8d56c. Original checkpoint/configuration/export bytes are unchanged. The LoRA header names the source checkpoint hash. The selected epoch-2 weights precede the later Stage-2 updates in the recovered training log. April generation calls record a checkpoint pathname and settings but lack a hash binding to the bytes loaded at run time.
The public illustrative generator adapter differs from the historical mask/drift implementation and does not reproduce the released images. See the verified provenance report for reconstructed corpus/split membership, recovered masks and preprocessed source images, original logs, access instructions, and unresolved historical records. No training, generation, or model experiment was run for this clarification.
- Downloads last month
- 17