{"version":1,"type":"story","url":"https://digestai.news/story/nvidia-structured-data-models-enable-zerotraining-tabular-predictions","json":"https://digestai.news/story/nvidia-structured-data-models-enable-zerotraining-tabular-predictions.json","markdown":"https://digestai.news/story/nvidia-structured-data-models-enable-zerotraining-tabular-predictions.md","slug":"nvidia-structured-data-models-enable-zerotraining-tabular-predictions","headline":"NVIDIA Structured Data Models enable zero‑training tabular predictions","summary":"The tutorial shows how NVIDIA’s Structured Data Models (SDM) library wraps the pretrained TabICLv2, KumoTabular and KumoRelational foundation models behind a single PyTorch interface. After installing the library from a specific Git commit and fixing a pyg‑lib wheel URL issue, users create typed TableTensors that preserve data types and handle datetime microsecond precision.\n\nZero‑training classification is demonstrated on the breast‑cancer dataset with three hundred labelled rows and two hundred and sixty‑nine rows to predict. TabICLv2 and KumoTabular achieve 0.970 accuracy and AUCs of 0.996 and 0.995, matching logistic regression (0.970 accuracy, 0.993 AUC) and out‑performing gradient boosting (0.967 accuracy, 0.992 AUC). Regression on the diabetes dataset returns 999 quantiles per query; the median prediction has an RMSE of 53.6, slightly better than RidgeCV (54.3) and far better than gradient boosting (64.7). Adding calendar and TFIDF preprocessing to TabICLv2 raises AUC from 0.525 to 0.789 on a support‑ticket escalation task. KumoRelational applied to a synthetic shop database (800 customers, 60 products, many orders) reaches 0.823 AUC, comparable to hand‑engineered features (0.828) but drops to 0.940 when time columns are omitted, exposing leakage.","keyPoints":["TabICLv2 and KumoTabular reach 0.970 accuracy and up to 0.996 AUC with no training on the breast‑cancer dataset (300 labelled rows).","Adding calendar and TFIDF features lifts TabICLv2 AUC from 0.525 to 0.789 on a support‑ticket escalation task.","KumoRelational attains 0.823 AUC on raw shop tables, matching hand‑built 0.828 baseline; without time columns it jumps to 0.940."],"whyItMatters":"Pretrained tabular models that require no training lower data‑labeling effort and compute cost, enabling fast deployment for many business analytics tasks.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["NVIDIA"],"models":["TabICLv2","KumoTabular","KumoRelational","TabFM","TimesFM 3"],"people":[]},"firstPublishedAt":"2026-10-18T05:56:05Z","updatedAt":"2026-10-18T05:56:05Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"MarkTechPost","title":"How to Make Zero-Training Tabular Predictions with NVIDIA Structured Data Models","url":"https://marktechpost.com/2026/10/17/how-to-make-zero-training-tabular-predictions-with-nvidia-structured-data-models","publishedAt":"2026-10-18T05:56:05Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"NVIDIA Structured Data Models enable zero‑training tabular predictions\", 18 October 2026, https://digestai.news/story/nvidia-structured-data-models-enable-zerotraining-tabular-predictions","publisher":"Digest AI","title":"NVIDIA Structured Data Models enable zero‑training tabular predictions","datePublished":"2026-10-18T05:56:05Z","url":"https://digestai.news/story/nvidia-structured-data-models-enable-zerotraining-tabular-predictions"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}