{"success":true,"data":{"author":"Maurizio Pierini, Javier Duarte, Nhan Tran, Marat Freytsis (hls4ml / Fast Machine Learning Lab)","bucketAccessType":"public","bucketName":"fastmachinelearning/hls4ml_lhc_jets_150p","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-16T02:38:29.443Z","description":"Simulated high transverse momentum (about 1 TeV) jets from LHC proton-proton collisions, labelled as gluon, light quark, W boson, Z boson or top quark, prepared by Pierini, Duarte, Tran and Freytsis for the hls4ml studies of low-latency neural network inference on FPGAs in the Level-1 trigger. Each jet keeps up to 150 constituents with 16 features (px, py, pz, energy, relative and rotated pT, eta and phi, delta-R, cos theta) and 53 jet-level substructure observables: N-subjettiness tau ratios, energy correlation functions C2, D2, M2 and N2, groomed masses (trimmed, mMDT, pruned, soft drop) and multiplicity. This repository, published by the fastmachinelearning organisation, reshapes the original HDF5 files to one Parquet row per jet with an integer label column, dropping only the jet images: 496,000 training and 124,000 validation jets cut from the Zenodo train archive and 260,000 test jets from its val archive, roughly balanced across the five classes. Files are data/{train,validation,test}-*.parquet (about 6 GB), loadable with Hugging Face datasets; a Hydra-driven loader that selects leading constituents, normalises features and returns torch tensors is included. The original record (10.5281/zenodo.3602260) is CC-BY-4.0; cite Duarte et al., JINST 13 P07027 (2018).","displayTitle":"hls4ml LHC Jet Dataset (150 Particles): Five-Class Jet Tagging for FPGA Trigger Studies","doi":"10.5281/zenodo.3602260","id":"105e9f12-021a-531a-85fd-f0b14ef7df4b","license":"CC-BY-4.0","licenseEvidenceUrl":"https://zenodo.org/records/3602260","objectCountEstimate":18,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1088/1748-0221/13/07/P07027","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"fastmachinelearning/hls4ml_lhc_jets_150p\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"fastmachinelearning/hls4ml_lhc_jets_150p\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=fastmachinelearning, license:cc-by-4.0, size_categories:100K<n<1M, format:parquet, downloads=1252"],"format":"mixed","generatedAt":"2026-11-16T02:38:38.405Z","layout":{"example":"data/test-00000-of-00002.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"504152311500159e84830115fcd681012c15ce8a2015001506150600008fc241c06756000003feaa010103fe560103fe620103fe720103fe340103fe740103fe5a0103fe540103fe3e0103fe82020103fe1e0d1d406c0103fe440103fe380103fe520103fea2014200ac010500420d22009a010d012330780103fe3c0103fe760103fe600d3320360103fe5c0103fe820125011d14660103fe4e010969011900a00119103a0103fe4805820805fefb019f106e0103fe220d43001e0d0c204c0103fe700103fe960132082a0103050410200103fe180d3e10640103fea4011d018101d500b8010d0158014f202c0103fe680103fe8a011501ca01a3011d242e0103fec8010105feef015e00be011f10320103fe8e0109211f009e010901310199018501310139007e0d9901252138019401b50531106a0103fe400dc208460103052d01352179014121500090015a20500103fe1c0103fe580dc601250824010305182129011c01e605b8011d00b40136217000a60109017501af218e2147251e004a2d820115119e01442100015c2121217d055401c7014f01350121007a31af0192012601f821dc0570015301040143012101bf011000c8018d081001032586101a0103fe8801114157012200dc010d2182014c4103211b008601152139256b412a01ac2142412901b0082601032572018601dd18140103fe3001030518210b","encoding":"hex","key":"data/test-00000-of-00002.parquet","truncatedAt":512}],"sampleObjects":["configs/config.yaml","configs/data/basis.yaml","configs/paths/default.yaml","data/test-00000-of-00002.parquet","data/test-00001-of-00002.parquet"],"sidecars":[{"bytes":"Creative Commons Attribution 4.0 International (CC BY 4.0)\n\nThis dataset is a mirror of\n\n    Pierini, Maurizio; Duarte, Javier; Tran, Nhan; Freytsis, Marat\n    HLS4ML LHC Jet dataset (150 particles)\n    Zenodo, 2020\n    https://doi.org/10.5281/zenodo.3602260\n\nwhich is published under the Creative Commons Attribution 4.0 International\nlicence. The mirror is published under the same licence.\n\nYou are free to share and adapt the material for any purpose, including\ncommercially, provided you give appropriate credit to the original authors,\nlink to the licence, and indicate whether changes were made. The rows here are\nthe rows of the original HDF5 files, reshaped; no values were recomputed.\n\nLicence: https://creativecommons.org/licenses/by/4.0/\n","encoding":"utf8","key":"LICENSE","name":"LICENSE","size":750,"truncated":false},{"bytes":"---\nlicense: cc-by-4.0\nlanguage:\n- en\ntask_categories:\n- tabular-classification\npretty_name: hls4ml LHC jet dataset (150 particles)\nsize_categories:\n- 100K<n<1M\ntags:\n- physics\n- particle-physics\n- jet-tagging\n- hls4ml\n- lhc\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*.parquet\n  - split: validation\n    path: data/validation-*.parquet\n  - split: test\n    path: data/test-*.parquet\n---\n\n# hls4ml LHC jet dataset (150 particles)\n\n*A mirror of the Zenodo record\n[10.5281/zenodo.3602260](https://doi.org/10.5281/zenodo.3602260), reshaped to one row\nper jet.*\n\nThe data are simulated high transverse-momentum (~1 TeV) jets from proton-proton\ncollisions at the Large Hadron Collider, labelled by what produced them: a light quark,\na gluon, a W boson, a Z boson or a top quark. Each jet is constituted of up to 150\nconstituents together with a set of jet-level observables, and the set was prepared for\nthe hls4ml jet-tagging studies. The mirror holds 880,000 jets.\n\nEvery value is the value of the original HDF5 files. The jet images from the original\ndata files are dropped.\n\n## One row is one jet\n\n| column | type | holds |\n|---|---|---|\n| the 16 `j1_*` columns | `list<float32>` | one entry per constituent of that jet, so the lists of one row all have the same length |\n| the 53 `j_*` columns | `float32` | jet-level observables, one value per jet |\n| `label` | `int8` | 0 gluon, 1 light quark, 2 W boson, 3 Z boson, 4 top quark |\n| `source_file` | `string` | the HDF5 file the jet was read from, without its extension |\n| `source_row` | `int32` | the jet's row within that file |\n\nThe constituent features, in the order the original files list them, are `j1_px`,\n`j1_py`, `j1_pz`, `j1_e`, `j1_erel`, `j1_pt`, `j1_ptrel`, `j1_eta`, `j1_etarel`,\n`j1_etarot`, `j1_phi`, `j1_phirel`, `j1_phirot`, `j1_deltaR`, `j1_costheta` and\n`j1_costhetarel`. The jet-level features are `j_ptfrac`, `j_pt`, `j_eta`, `j_mass`,\n`j_tau1_b1`, `j_tau2_b1`, `j_tau3_b1`, `j_tau1_b2`, `j_tau2_b2`, `j_tau3_b2`,\n`j_tau32_b1`, `j_tau32_b2`, `j_zlogz`, `j_c1_b0`, `j_c1_b1`, `j_c1_b2`, `j_c2_b1`,\n`j_c2_b2`, `j_d2_b1`, `j_d2_b2`, `j_d2_a1_b1`, `j_d2_a1_b2`, `j_m2_b1`, `j_m2_b2`,\n`j_n2_b1`, `j_n2_b2`, `j_tau1_b1_mmdt`, `j_tau2_b1_mmdt`, `j_tau3_b1_mmdt`,\n`j_tau1_b2_mmdt`, `j_tau2_b2_mmdt`, `j_tau3_b2_mmdt`, `j_tau32_b1_mmdt`,\n`j_tau32_b2_mmdt`, `j_c1_b0_mmdt`, `j_c1_b1_mmdt`, `j_c1_b2_mmdt`, `j_c2_b1_mmdt`,\n`j_c2_b2_mmdt`, `j_d2_b1_mmdt`, `j_d2_b2_mmdt`, `j_d2_a1_b1_mmdt`, `j_d2_a1_b2_mmdt`,\n`j_m2_b1_mmdt`, `j_m2_b2_mmdt`, `j_n2_b1_mmdt`, `j_n2_b2_mmdt`, `j_mass_trim`,\n`j_mass_mmdt`, `j_mass_prun`, `j_mass_sdb2`, `j_mass_sdm1` and `j_multiplicity`. The\nlabel is the argmax over the five one-hot columns `j_g`, `j_q`, `j_w`, `j_z`, `j_t` that\nthe original `jets` array ends with. The one-hot columns and the always-zero `j_undef`\nbeside them are not mirrored, but are replaced by `label`.\n\n## Splits\n\n`train` and `validation` come from the Zenodo *train* archive. Its files are\nconcatenated in sorted-filename order and cut with\n\n```python\ntrain_idx, validation_idx = train_test_split(np.arange(n), test_size=0.2, random_state=42)\n```\n\n`test` is the Zenodo *val* archive, in sorted-filename order.\n\n| split | jets | gluon | light quark | W boson | Z boson | top quark |\n|---|---|---|---|---|---|---|\n| `train` | 496,000 | 100,051 | 96,079 | 99,834 | 99,591 | 100,445 |\n| `validation` | 124,000 | 24,797 | 24,132 | 25,103 | 25,063 | 24,905 |\n| `test` | 260,000 | 52,404 | 50,468 | 52,235 | 52,298 | 52,595 |\n\n## Layout\n\n```\ndata/train-NNNNN-of-NNNNN.parquet\ndata/validation-NNNNN-of-NNNNN.parquet\ndata/test-NNNNN-of-NNNNN.parquet\nloader/            the pipeline described below\nconfigs/           the Hydra tree that drives it\nrequirements.txt   what that pipeline needs\n```\n\n## Loading\n\nThe tables need only the `datasets` package:\n\n```python\nfrom datasets import load_dataset\njets = load_dataset(\"fastmachinelearning/hls4ml_lhc_jets_150p\", split=\"train\")\njets[0][\"j1_pt\"]    # the transverse momenta of that jet's constituents\n```\n\nThis data record also contains a data proc","encoding":"utf8","key":"README.md","name":"README.md","size":8601,"truncated":true}],"sourceObjectKey":"data/test-00000-of-00002.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"fastmachinelearning/hls4ml_lhc_jets_150p\")\nprint(ds)","starterSamplePath":"data/test-00000-of-00002.parquet","totalObjects":18},"summary":"Simulated high transverse momentum (about 1 TeV) jets from LHC proton-proton collisions, labelled as gluon, light quark, W boson, Z boson or top quark, prepared by Pierini,…","title":"hls4ml LHC Jet Dataset (150 Particles): Five-Class Jet Tagging for FPGA Trigger Studies","updatedAt":"2026-11-16T17:25:11.073Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/fastmachinelearning/hls4ml_lhc_jets_150p"}}