{"id":22668598,"url":"https://github.com/lanl/blackout-diffusion","last_synced_at":"2025-04-12T11:08:18.195Z","repository":{"id":166521836,"uuid":"642021357","full_name":"lanl/Blackout-Diffusion","owner":"lanl","description":null,"archived":false,"fork":false,"pushed_at":"2024-06-09T19:32:34.000Z","size":599,"stargazers_count":27,"open_issues_count":1,"forks_count":4,"subscribers_count":6,"default_branch":"main","last_synced_at":"2024-06-27T00:19:12.128Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/lanl.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null}},"created_at":"2023-05-17T16:41:33.000Z","updated_at":"2024-06-10T20:53:40.000Z","dependencies_parsed_at":null,"dependency_job_id":"560169c2-97fa-4c07-b95a-d1644478e016","html_url":"https://github.com/lanl/Blackout-Diffusion","commit_stats":{"total_commits":14,"total_committers":2,"mean_commits":7.0,"dds":0.4285714285714286,"last_synced_commit":"0320e35997f8ed810ea87dc55ffd55a50949c329"},"previous_names":["lanl/blackout-diffusion"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lanl%2FBlackout-Diffusion","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lanl%2FBlackout-Diffusion/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lanl%2FBlackout-Diffusion/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/lanl%2FBlackout-Diffusion/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/lanl","download_url":"https://codeload.github.com/lanl/Blackout-Diffusion/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":228911888,"owners_count":17990774,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-12-09T15:15:54.379Z","updated_at":"2024-12-09T15:15:54.926Z","avatar_url":"https://github.com/lanl.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Blackout-Diffusion\n\nThis repository host the source codes for the project \"Blackout Diffusion: Generative Diffusion Models in Discrete-State Spaces\", authored by Javier E. Santos, Zachary R. Fox (former CCS-3/LANL, currently at ORNL), Nicholas Lubbers (CCS-3, LANL), and Yen Ting Lin (CCS-3, LANL). The paper has been accepted by the 40th International Conference on Machine Learning. \n\nThe source codes were reviewed and approved with a LANL C-number C23047 by the Richard P. Feynman Center for Innovation (FCI) at the Los Alamos National Laboratory.\n\n## NCSN++ code base\nThe NCSN++ network, which was used to learn the reverse-time transition rates, was cloned from https://github.com/yang-song/score_sde_pytorch/tree/1618ddea340f3e4a2ed7852a0694a809775cf8d0\n\nWe used Python 3.7 and PyTorch 1.12.1+cu116.\nNCSN++ used CUDA complied layers `fused` or `upfirdn2d`.\nIf error message occurs during compiling `fused` or `upfirdn2d`, it may be useful to remove the respective folders in `~/.cache/torch_extensions/py37_cu116/'. \n\n#### Differential changes to the SDE code base\n- The file `util.py` and `datasets.py` was updated for our purpose \n- The file `loadDataPipeline.py` was added (for some reason, the original integrated data pipeline could not be used on our internal computational cluster.)\n\n## Blackout Diffusion codes\n\nTraining and generation (inference) are separated into two Jupyter Notebooks. The first time when a training code is executed, the forward solution of the $\\beta$-decay process at observation times is solved and stored in `forwardSolution.npz`. Later on, running the same training codes would only read the stored solution.\n\n### CIFAR-10 dataset\n- `blackout-cifar10-train-by-finiteTime.ipynb`:  Training by the finite-time formulation of the loss function\n- `blackout-binarizedMNIST-train-by-instantaneous.ipynb`:  Training by the instantaneous formulation of the loss function\n- `blackout-binarizedMNIST-generate-by-binomailBridge.ipynb`:  Generation using the binomial bridge formula\n- `blackout-cifar10-generate-by-tauLeaping.ipynb`:  Generation using the $\\tau$-leaping formulation\n\n### Binarized MNIST dataset\n- `blackout-binarizedMNIST-train-by-instantaneous.ipynb`:  Training by the instantaneous formulation of the loss function\n- `blackout-binarizedMNIST-generate-by-binomailBridge.ipynb`:  Generation by the binomial bridge formula\n\n### CelebA dataset\n- `blackout-celebA64-train-by-instantaneous.ipynb`:  Training by the instantaneous formulation of the loss function\n- `blackout-celebA64-generate-by-binomialBridge.ipynb`:  Generation by the binomial bridge formula\n\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flanl%2Fblackout-diffusion","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flanl%2Fblackout-diffusion","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flanl%2Fblackout-diffusion/lists"}