{"id":15899712,"url":"https://github.com/kisonho/torchmanager-diffusion","last_synced_at":"2025-08-14T10:33:51.997Z","repository":{"id":219766688,"uuid":"642875775","full_name":"kisonho/torchmanager-diffusion","owner":"kisonho","description":"The torchmanager implementation for diffusion models.","archived":false,"fork":false,"pushed_at":"2025-07-25T18:37:46.000Z","size":3721,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-07-26T00:59:02.509Z","etag":null,"topics":["ddpm","diffusion-models","pytorch","torchmanager"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/kisonho.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2023-05-19T14:41:51.000Z","updated_at":"2025-07-25T18:09:37.000Z","dependencies_parsed_at":"2024-04-08T22:26:00.005Z","dependency_job_id":"9d646a08-ddd9-4d80-baee-eba2284f8d48","html_url":"https://github.com/kisonho/torchmanager-diffusion","commit_stats":null,"previous_names":["kisonho/diffusion"],"tags_count":43,"template":false,"template_full_name":null,"purl":"pkg:github/kisonho/torchmanager-diffusion","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kisonho%2Ftorchmanager-diffusion","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kisonho%2Ftorchmanager-diffusion/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kisonho%2Ftorchmanager-diffusion/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kisonho%2Ftorchmanager-diffusion/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/kisonho","download_url":"https://codeload.github.com/kisonho/torchmanager-diffusion/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kisonho%2Ftorchmanager-diffusion/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":270405933,"owners_count":24578154,"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","status":"online","status_checked_at":"2025-08-14T02:00:10.309Z","response_time":75,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"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":["ddpm","diffusion-models","pytorch","torchmanager"],"created_at":"2024-10-06T10:22:39.368Z","updated_at":"2025-08-14T10:33:51.678Z","avatar_url":"https://github.com/kisonho.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Torchmanager Diffusion Models Plug-in\nThe torchmanager implementation for diffusion models.\n\n## Pre-requisites\n* Python \u003e= 3.9\n* [SciPy](https://www.scipy.org) \u003e= 1.11.4\n* [PyTorch](https://pytorch.org) \u003e= 2.0.1\n* [LPIPS](https://github.com/richzhang/PerceptualSimilarity)\n* [torchmanager](https://github.com/kisonho/torchmanager) \u003e= 1.2\n* [einops](https://github.com/arogozhnikov/einops) \u003e= 0.6.1\n\n## Installation\n* PyPi: `pip install torchmanager-diffusion`\n\n## DDPM Manager Usage\n### Train DDPM\nDirect compile `DDPMManager` with a model, a beta space, and a number of time steps. Then, use `fit` method to train the model.\n\n```python\nimport diffusion\nfrom diffusion import DDPMManager\nfrom torchmanager import callbacks, data, losses\n\n# initialize dataset\ndataset: data.Dataset = ...\n\n# initialize model, beta_space, and time_steps\nmodel: torch.nn.Module = ...\nbeta_space: diffusion.scheduling.BetaSpace = ...\ntime_steps: int = ...\n\n# initialize optimizer and loss function\noptimizer: torch.optim.Optimizer = ...\nloss_fn: losses.Loss = ...\n\n# compile the ddpm manager\nmanager = DDPMManager(model, beta_space, time_steps, optimizer=optimizer, loss_fn=loss_fn)\n\n# initialize callbacks\ncallback_list: list[callbacks.Callback] = ...\n\n# train the model\ntrained_model = manager.fit(dataset, epochs=..., callbacks=callback_list)\n```\n\n### Evaluate DDPM\nAdd necessary metrics and use `test` method with `sampling_images` as `True` to evaluate the trained model.\n\n```python\nimport torch\nfrom diffusion import DDPMManager\nfrom torchmanager import data, metrics\nfrom torchvision import models\n\n# load manager from checkpoints\nmanager = DDPMManager.from_checkpoint(...)\nassert isinstance(manager, DDPMManager), \"manager is not a DDPMManager.\"\n\n# initialize dataset\ntesting_dataset: data.Dataset = ...\n\n# add neccessary metrics\ninception = models.inception_v3(pretrained=True)\ninception.fc = torch.nn.Identity()  # type: ignore\ninception.eval()\nfid = metrics.FID(inception)\nmanager.metrics.update({\"FID\": fid})\n\n# evaluate the model\nsummary = manager.test(testing_dataset, sampling_images=True)\n```\n\n## Customize Diffusion Algorithm\nInherit `DiffusionManager` and implement abstract methods `forward_diffusion` and `sampling_step` to customize the diffusion algorithm.\n\n```python\nfrom diffusion import DiffusionManager\n\nclass CustomizedManager(DiffusionManager):\n    def forward_diffusion(self, data: Any, condition: Optional[torch.Tensor] = None, t: Optional[torch.Tensor] = None) -\u003e tuple[Any, torch.Tensor]:\n        ...\n\n    def sampling_step(self, data: DiffusionData, i: int, /, *, return_noise: bool = False) -\u003e Union[torch.Tensor, tuple[torch.Tensor, torch.Tensor]]:\n        ...\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkisonho%2Ftorchmanager-diffusion","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fkisonho%2Ftorchmanager-diffusion","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkisonho%2Ftorchmanager-diffusion/lists"}