{"id":30220775,"url":"https://github.com/mkirchmeyer/dino","last_synced_at":"2025-08-14T09:07:52.067Z","repository":{"id":61186508,"uuid":"546128380","full_name":"mkirchmeyer/DINo","owner":"mkirchmeyer","description":"Time- and space-continuous neural PDE forecaster based on INRs and ODEs","archived":false,"fork":false,"pushed_at":"2024-01-13T17:33:06.000Z","size":61,"stargazers_count":36,"open_issues_count":0,"forks_count":7,"subscribers_count":4,"default_branch":"main","last_synced_at":"2024-01-14T02:58:17.607Z","etag":null,"topics":["dynamical-systems","implicit-neural-representation","pde"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2209.14855","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/mkirchmeyer.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.md","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2022-10-05T15:15:51.000Z","updated_at":"2024-01-04T09:24:17.000Z","dependencies_parsed_at":"2024-01-13T19:00:41.515Z","dependency_job_id":"e875412f-c7a6-4e14-9ed9-e03e6e0763b1","html_url":"https://github.com/mkirchmeyer/DINo","commit_stats":null,"previous_names":[],"tags_count":0,"template":null,"template_full_name":null,"purl":"pkg:github/mkirchmeyer/DINo","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkirchmeyer%2FDINo","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkirchmeyer%2FDINo/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkirchmeyer%2FDINo/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkirchmeyer%2FDINo/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/mkirchmeyer","download_url":"https://codeload.github.com/mkirchmeyer/DINo/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mkirchmeyer%2FDINo/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":270393033,"owners_count":24575818,"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":["dynamical-systems","implicit-neural-representation","pde"],"created_at":"2025-08-14T09:07:49.212Z","updated_at":"2025-08-14T09:07:52.033Z","avatar_url":"https://github.com/mkirchmeyer.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Continuous PDE Dynamics Forecasting with Implicit Neural Representations\n\nOfficial PyTorch implementation of DINo (Dynamics-aware Implicit Neural Representation) | [Accepted at ICLR 2023 (Notable-Top-25%, Spotlight)](https://openreview.net/forum?id=B73niNjbPs) | [Arxiv](https://arxiv.org/abs/2209.14855) \n\n[Yuan Yin](https://yuan-yin.github.io/) (equal contribution), [Matthieu Kirchmeyer](https://mkirchmeyer.github.io/) (equal contribution), [Jean-Yves Franceschi](https://jyfranceschi.fr) (equal contribution), [Alain Rakotomamonjy](http://asi.insa-rouen.fr/enseignants/~arakoto/), [Patrick Gallinari](http://www-connex.lip6.fr/~gallinar/gallinari/pmwiki.php)\n\nhttps://user-images.githubusercontent.com/15007187/194546732-dafb7627-4be3-4e8e-910a-9b08ad375158.mp4\n\nhttps://user-images.githubusercontent.com/15007187/194546967-557280ff-bde0-4bb4-9650-c9ec1f7bf986.mp4\n\nThe `requirements.txt` file lists Python package dependencies.\n\n## Data \n\n* For `navier_stokes`, `wave`, data is generated as part of our script c.f. `data_pdes.py`.\n* For `shallow_water`, data can be found [here](https://doi.org/10.6084/m9.figshare.21298179).\nIt should be stored in `./results/shallow_water` or in a custom location given as argument to `-f`.\n\n## Pretrained models\n\nWe provide the checkpoints [here](https://doi.org/10.6084/m9.figshare.21298251) \n* `NS_100` for `navier_stokes` (100% subsampling rate)\n* `Wave_100` for `wave` (100% subsampling rate)\n* `SW` for `shallow_water` (c.f. our paper)\n\nThey should be stored in `./results/\u003cDATASET\u003e` and can be used for:\n* warm start via `train.py`; the path of the model should be given as argument via `-c`. \n* inference via `test.py`.\n\nWe do not control the behavior of the checkpoints on other datasets than those generated by our code.\n\n## Training\n\n```\npython3 train.py -d \u003cDATASET\u003e -g 0 -r \u003cRATE\u003e\n```\n\n* `-c`: checkpoint location for warm-start with a pretrained model (default: no warmstart)\n* `-d`: input dataset (`navier_stokes`, `wave`, `shallow_water`)\n* `-f`: home path (defaults to `\"./results\"`)\n* `-g`: gpu id (defaults to `0`)\n* `-r`: subsampling rate (defaults to `1.0`)\n* `-w`: the length of conditioning window for non-Markovian dynamics (defaults to Markovian - `0`)\n\nBy running the train script, it will generate an unique id of each run, called the run_id.\n\nLogs are available in `./results/\u003cRUN_ID\u003e/log` and display the result over In/Out-s + In/Out-t for both train and test trajectories (cf the Figure below where red represents observed data and our paper for more details). \nThese are the numbers reported in Table 2.\n\n![task](https://user-images.githubusercontent.com/15007187/215505653-843c1b0e-f7e1-41ce-819b-a16aec1d09d5.png)\n\n## Inference on new conditions\n\nThe run_id can also be used to run the inference script on test trajectories for the following settings:\n* Evaluation on a new grid for `navier_stokes`, `wave` (Table 3.a).\n* Super-resolution for `shallow_water` (Figure 5).\n\n```\npython3 test.py -d \u003cDATASET\u003e -p \u003cRUN_ID\u003e -g 0 -r \u003cRATE\u003e -s \u003cSEED\u003e\n```\n\n* `-d`: input dataset (`navier_stokes`, `wave`, `shallow_water_hr`)\n* `-f`: home path (defaults to `\"./results\"`)\n* `-g`: gpu id (defaults to `0`)\n* `-p`: run id\n* `-r`: subsampling rate (defaults to `1.0`)\n* `-s`: subsampling seed (defaults to `1` the train subsampling seed; for Table 3.a. we chose `-s 2`)\n\nhttps://user-images.githubusercontent.com/15007187/194547306-e0f151cf-a4fc-43be-a907-35816124020d.mp4\n\n## Citation\n\n```\n@inproceedings{Yin2023,\ntitle={Continuous PDE Dynamics Forecasting with Implicit Neural Representations},\nauthor={Yin, Yuan and Kirchmeyer, Matthieu and Franceschi, Jean-Yves and Rakotomamonjy, Alain and Gallinari, Patrick},\nbooktitle={International Conference on Learning Representations},\nyear={2023},\nurl={https://openreview.net/forum?id=B73niNjbPs}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkirchmeyer%2Fdino","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmkirchmeyer%2Fdino","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmkirchmeyer%2Fdino/lists"}