Views
No views yet
pip install synthefy-norinori.pt (this repo)| Suite | Datasets | Mean R² | Median R² |
|---|---|---|---|
| TabArena | 13 | 0.8148 | 0.8834 |
| TALENT | 72 | 0.7575 | 0.8844 |
| OpenML | 11 | 0.6459 | 0.6212 |
| Overall | 96 | 0.7525 | 0.8745 |
1Look at my code/task/report here and figure out where Nori would best fit — it's
2Synthefy's tabular foundation model, a drop-in scikit-learn estimator that predicts
3a continuous target by in-context learning: no training loop, no hyperparameters,
4and it uses the GPU automatically when one's available (CPU otherwise).
5
61. Install it with this project's package manager
7 (e.g. `uv add synthefy-nori`, or `pip install -U synthefy-nori`).
8
92. Use it wherever a tabular regression / prediction step fits:
10
11 ```python
12 from synthefy_nori import NoriRegressor
13
14 reg = NoriRegressor(model="nori-30m") # downloads these weights from the Hub on first predict
15 reg.fit(X_train, y_train) # stores your rows as context — no training happens
16 y_pred = reg.predict(X_test) # point predictions (predictive-distribution mean)
17
18 # Prediction intervals come free — no conformal/quantile add-ons:
19 lo, mid, hi = reg.predict(X_test, output_type="quantiles", quantiles=[0.1, 0.5, 0.9])
20 ```
21
22X is a numeric feature matrix (encode categoricals as ordinals/one-hot, leave
23missing values as NaN, no scaling needed); y is a finite continuous target. If I
24already have a model, wire Nori up alongside it on the same train/test split and
25metric so I can compare them. If the best place to plug Nori in isn't obvious,
26show me where you'd put it and confirm with me before making changes.
27
28Going deeper: synthefy-nori ships a ready-made nori-regression skill for AI coding
29assistants with vetted recipes — calibrated prediction intervals, honest baseline
30comparison under fixed CV, SHAP/PDP interpretability, and leak-safe one-step
31time-series forecasting. Read and follow it if relevant:
32https://github.com/Synthefy/synthefy-nori/tree/main/.claude/skills/nori-regressionpip install synthefy-nori1from sklearn.datasets import load_diabetes
2from sklearn.model_selection import train_test_split
3from synthefy_nori import NoriRegressor
4
5X, y = load_diabetes(return_X_y=True)
6X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)
7
8model = NoriRegressor(model="nori-30m") # downloads these weights from the Hub on first use
9model.fit(X_train, y_train) # "fit" just stores the labeled rows as context
10pred = model.predict(X_test) # predictions in a single forward pass, no training1from synthefy_nori import predict
2pred = predict(X_train, y_train, X_test, task="regression", model="nori-30m")predict follows the TabPFNRegressor.predict contract: pass output_type="mean" (default),
"median", or "mode" to choose the point estimate drawn from the model's predictive
distribution.NoriRegressor(model_path="path/to/nori.pt").export HF_TOKEN=hf_..., hf auth login, or
NoriRegressor(model="nori-30m", token="hf_...").1@software{synthefy_2026_20710462,
2 author = {Synthefy and
3 Li, Po-han and
4 Narayanan, Aditya and
5 Narasimhan, Sai Shankar and
6 Mallampalli, Raghav and
7 Agrawal, Aahan and
8 Ajan, Bekzat and
9 Shah, Raimi and
10 Agarwal, Shubhankar},
11 title = {Synthefy Nori: Tabular Foundation Model for Regression},
12 month = jun,
13 year = 2026,
14 publisher = {Zenodo},
15 version = {0.6.0},
16 doi = {10.5281/zenodo.20710462},
17 url = {https://doi.org/10.5281/zenodo.20710462},
18}