{"success":true,"data":{"author":"Zane Blood","bucketAccessType":"public","bucketName":"zaneblood1/cmb_lensing_data","bucketPrefix":"","category":"Astronomy & Space","cloudProvider":"huggingface","createdAt":"2026-11-21T20:46:01.546Z","description":"Data products behind CMBLensing.py, a JAX port of the Julia package CMBLensing.jl that simulates gravitationally lensed cosmic microwave background (CMB) temperature and polarization maps on a flat sky, reconstructs the unlensed field and lensing potential (map_joint), and jointly samples LCDM cosmological parameters from a lensed map (sample_joint). The repository holds a precomputed 5D grid of CAMB power spectra over omch2, ombh2, n_s, theta_MC and log A (camb_grid_spline.npz, 7.9 GB) that the sampler interpolates as a cubic spline, plus the results reported for the package: LCDM parameter chains (joint_inference_data), empirical delensed field covariances for a Fisher forecast, per-mode lensing reconstruction noise of map_joint and of the sampler posterior mean, and timing and reconstruction comparisons between the Python and Julia map_joint. Unless noted, experiments use 128 x 128 maps at 2.5 arcmin pixels with white noise of 2.5, 5 or 10 uK-arcmin. About 23 GB of NumPy .npz and text files; read with numpy.load and numpy.loadtxt. Released under the MIT license by the author, the same as the code.","displayTitle":"CMBLensing.py Data: 5D CAMB LCDM Spectra Grid, Joint Lensing Chains and Delensing Covariances for Flat-Sky CMB Lensing","huggingFaceUrl":"https://huggingface.co/datasets/zaneblood1/cmb_lensing_data","id":"41822a9f-e908-5a4c-a244-39b1375a9718","license":"MIT","licenseEvidenceUrl":"https://huggingface.co/datasets/zaneblood1/cmb_lensing_data","objectCountEstimate":14123,"probeStatus":"ok","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"zaneblood1/cmb_lensing_data\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"zaneblood1/cmb_lensing_data\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=zaneblood1, license:mit, size_categories:10K<n<100K, downloads=114"],"format":"mixed","generatedAt":"2026-11-21T20:48:55.538Z","layout":{"example":"camb_grid_spline.npz","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"camb_grid_spline.npz","truncatedAt":512}],"sampleObjects":["camb_grid_spline.npz","camb_logA_grid.npz","camb_ns_grid.npz","camb_ombh2_grid.npz","camb_omch2_grid.npz"],"sidecars":[{"bytes":"---\nlicense: mit\npretty_name: CMBLensing.py data\nviewer: false\nsize_categories:\n- 10K<n<100K\ntags:\n- physics\n- cosmology\n- astrophysics\n- cmb\n- gravitational-lensing\n- mcmc\n- camb\n---\n\n# CMBLensing.py data\n\nThis dataset holds every large data product behind [CMBLensing.py](https://github.com/zaneblood1/CMBLensing.py), a [JAX](https://github.com/jax-ml/jax) port of the Julia package [CMBLensing.jl](https://github.com/marius311/CMBLensing.jl). CMBLensing.py simulates gravitationally lensed CMB temperature and polarization maps on a flat sky, recovers the unlensed field and the lensing potential from a lensed data map (`map_joint`), and jointly samples the LCDM parameters from that map (`sample_joint`).\n\nDocumentation for the package: <https://zaneblooddev.com/CMBLensing.py/>\n\n## Why this dataset exists\n\nThe code lives on GitHub, but the data it needs and the data it produced do not fit there:\n\n- The sampler cannot run without a precomputed **5D grid of CAMB spectra** (~7.9 GB). Calling CAMB at every proposed parameter combination would dominate the run time, so the spectra are computed once on a grid and interpolated.\n- The **results** reported for the package (LCDM chains, lensing reconstruction noise measurements, delensed covariances for the Fisher forecast, and the Python versus Julia `map_joint` comparison) took many cluster hours to generate. Storing the raw outputs here lets anyone regenerate the figures, check the analysis, or run a new analysis without repeating those jobs.\n\nEverything is stored in the layout the cluster jobs wrote it in, so the scripts in the repository read it without conversion.\n\n> This dataset was previously named `zaneblood1/5D_LCDM_CAMB_Grid_Spline` and held only the CAMB grid. The old name redirects here, and the grid files are still at the dataset root, so existing installs keep working.\n\n## Contents\n\n| Path | Size | Files | What it is |\n|---|---|---|---|\n| `camb_grid_spline.npz` | 7.9 GB | 1 | the 5D CAMB grid that `sample_joint` interpolates |\n| `camb_*_grid.npz` | small | 5 | legacy 1D CAMB grids, no longer read by the package |\n| `joint_inference_data/` | 512 MB | 5,061 | LCDM parameter chains from `sample_joint` |\n| `delensed_field_covariance_data/` | 683 MB | 1,506 | empirical delensed covariances, the field block of the Fisher forecast |\n| `map_joint_phi_noise_data/` | 98 MB | 303 | per-mode lensing reconstruction noise of `map_joint` |\n| `sampler_phi_noise_data/` | 787 MB | 622 | per-mode lensing reconstruction noise of the sampler's posterior mean |\n| `map_joint_comparison/` | 13 GB | 6,420 | `map_joint` run times and reconstructions, CMBLensing.py versus CMBLensing.jl |\n\nUnless a folder says otherwise, the experiments use 128 x 128 pixel temperature maps at 2.5 arcminutes per pixel, white noise only (no 1/f term, no beam, no mask), at 2.5, 5 or 10 uK-arcmin, with the ground truth cosmology\n\n| `omch2` | `ombh2` | `ns` | `theta_MC_100` | `logA` |\n|---|---|---|---|---|\n| 0.109381 | 0.022386 | 0.959814 | 1.031732 | 3.218387 |\n\n## Downloading\n\nThe whole dataset is about 23 GB. To take only the part you need, pass a folder pattern:\n\n```bash\npip install -U huggingface_hub\n\n#one folder\nhf download zaneblood1/cmb_lensing_data --repo-type dataset \\\n    --include \"joint_inference_data/*\" --local-dir cmb_lensing_data\n\n#everything\nhf download zaneblood1/cmb_lensing_data --repo-type dataset --local-dir cmb_lensing_data\n```\n\nor from Python:\n\n```python\nfrom huggingface_hub import snapshot_download\n\nsnapshot_download(repo_id = \"zaneblood1/cmb_lensing_data\", repo_type = \"dataset\",\n                  allow_patterns = [\"delensed_field_covariance_data/*\"],\n                  local_dir = \"cmb_lensing_data\")\n```\n\nThe CAMB grid needs no manual step: CMBLensing.py downloads it into `camb_splines/` the first time the sampler asks for it, or ahead of time with `cmb-lensing-download-splines`.\n\nThe files are NumPy `.npz` archives and plain text, so the dataset viewer and `datasets.load_dataset` do not apply. Read them with `numpy.load` and `numpy.loadtxt`.\n\n## The 5D CAMB grid\n\n`camb_grid_spline.npz` holds CAM","encoding":"utf8","key":"README.md","name":"README.md","size":12869,"truncated":true},{"bytes":"# Unsuccessful test run\n\nThis folder holds the chains of a test at a white noise level of 2 uK-arcmin (50 maps, 5 chains per map, 6000 sweeps per chain; see `experimental_details.txt`). The test was unsuccessful: these chains are kept for the record only and should not be used as a result.\n\nThey are deliberately left out of the results visualizer (https://zaneblood1.github.io/CMBLensing.py/results/), and there is no delensed field covariance at 2 uK-arcmin, so no Fisher forecast exists for them.\n\nNote added 2026-11-19.\n","encoding":"utf8","key":"joint_inference_data/omch2_logA_theta_MC_100_joint_inference_08_31_26/README.md","name":"README.md","size":525,"truncated":false}],"sourceObjectKey":"camb_grid_spline.npz","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"zaneblood1/cmb_lensing_data\", filename=\"camb_grid_spline.npz\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"camb_grid_spline.npz","totalObjects":14123},"summary":"Data products behind CMBLensing.py, a JAX port of the Julia package CMBLensing.jl that simulates gravitationally lensed cosmic microwave background (CMB) temperature and…","title":"CMBLensing.py Data: 5D CAMB LCDM Spectra Grid, Joint Lensing Chains and Delensing Covariances for Flat-Sky CMB Lensing","updatedAt":"2026-11-22T01:05:45.918Z","visibility":"public","websiteUrl":"https://github.com/zaneblood1/CMBLensing.py"}}