structured-data-models 0.2.1__tar.gz → 0.3.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.agents/skills/docstring/SKILL.md +3 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.agents/skills/processor-development/SKILL.md +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.coderabbit.yaml +18 -4
- structured_data_models-0.3.0/.github/ISSUE_TEMPLATE/bug_report.yml +59 -0
- structured_data_models-0.3.0/.github/PULL_REQUEST_TEMPLATE/default.md +35 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/doc.yml +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/test-gpu.yml +22 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/test.yml +21 -10
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.gitignore +5 -0
- structured_data_models-0.3.0/.readthedocs.yaml +17 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/AGENTS.md +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/CLAUDE.md +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/CONTRIBUTING.md +8 -21
- {structured_data_models-0.2.1/structured_data_models.egg-info → structured_data_models-0.3.0}/PKG-INFO +34 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/README.md +25 -4
- structured_data_models-0.3.0/THIRD_PARTY_LICENSES.md +87 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/README.md +4 -0
- structured_data_models-0.3.0/benchmark/relational/README.md +3 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/relational/relarena/README.md +3 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/README.md +19 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/beyondarena/main.py +23 -9
- structured_data_models-0.3.0/benchmark/tabular/finetune.py +237 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/model.py +179 -91
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/scoringbench/README.md +13 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/scoringbench/main.py +54 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/scoringbench/models.py +40 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/tabarena/main.py +23 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/talent/README.md +17 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/talent/main.py +43 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/talent/models.py +69 -16
- structured_data_models-0.3.0/benchmark/timeseries/README.md +3 -0
- structured_data_models-0.3.0/benchmark/timeseries/__init__.py +2 -0
- structured_data_models-0.3.0/benchmark/timeseries/fev_bench/README.md +47 -0
- structured_data_models-0.3.0/benchmark/timeseries/fev_bench/__init__.py +4 -0
- structured_data_models-0.3.0/benchmark/timeseries/fev_bench/evaluate.py +42 -0
- structured_data_models-0.3.0/benchmark/timeseries/fev_bench/main.py +70 -0
- structured_data_models-0.3.0/benchmark/timeseries/fev_bench/predictor.py +216 -0
- structured_data_models-0.3.0/benchmark/timeseries/gift_eval/README.md +46 -0
- structured_data_models-0.3.0/benchmark/timeseries/gift_eval/__init__.py +7 -0
- structured_data_models-0.3.0/benchmark/timeseries/gift_eval/datasets.py +91 -0
- structured_data_models-0.3.0/benchmark/timeseries/gift_eval/evaluate.py +46 -0
- structured_data_models-0.3.0/benchmark/timeseries/gift_eval/main.py +155 -0
- structured_data_models-0.3.0/benchmark/timeseries/model.py +44 -0
- structured_data_models-0.3.0/benchmark/timeseries/predictor.py +128 -0
- structured_data_models-0.3.0/benchmark/timeseries/time_bench/README.md +46 -0
- structured_data_models-0.3.0/benchmark/timeseries/time_bench/__init__.py +2 -0
- structured_data_models-0.3.0/benchmark/timeseries/time_bench/evaluate.py +68 -0
- structured_data_models-0.3.0/benchmark/timeseries/time_bench/main.py +101 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_templates/autosummary/class.rst +4 -0
- structured_data_models-0.3.0/docs/source/_templates/autosummary/model_class.rst +78 -0
- structured_data_models-0.3.0/docs/source/api/internal/stype_config.rst +19 -0
- structured_data_models-0.3.0/docs/source/api/models.rst +63 -0
- structured_data_models-0.3.0/docs/source/api/prior.rst +94 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/processing.rst +2 -0
- structured_data_models-0.3.0/docs/source/api/sdm.rst +18 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/conf.py +80 -1
- structured_data_models-0.3.0/docs/source/finetuning.md +172 -0
- structured_data_models-0.3.0/docs/source/forecasting.md +202 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/icl.md +41 -16
- structured_data_models-0.3.0/docs/source/images/chronos2_dark.svg +3203 -0
- structured_data_models-0.3.0/docs/source/images/chronos2_light.svg +3202 -0
- structured_data_models-0.3.0/docs/source/images/finetuning_dark.svg +5 -0
- structured_data_models-0.3.0/docs/source/images/finetuning_light.svg +5 -0
- structured_data_models-0.3.0/docs/source/images/forecasting_dark.svg +4 -0
- structured_data_models-0.3.0/docs/source/images/forecasting_light.svg +4 -0
- structured_data_models-0.3.0/docs/source/images/kumo_anomaly_light.png +0 -0
- structured_data_models-0.3.0/docs/source/images/kumo_forecast_light.png +0 -0
- structured_data_models-0.3.0/docs/source/images/nanotabicl_dark.svg +4 -0
- structured_data_models-0.3.0/docs/source/images/nanotabicl_light.svg +4 -0
- structured_data_models-0.3.0/docs/source/images/timesfm3_dark.png +0 -0
- structured_data_models-0.3.0/docs/source/images/timesfm3_light.png +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/index.md +5 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/install.md +7 -1
- structured_data_models-0.3.0/docs/source/nanotabicl.md +287 -0
- structured_data_models-0.3.0/docs/source/nim.md +56 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/overview.md +13 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/processing.md +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/tensor.md +23 -8
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/README.md +2 -1
- structured_data_models-0.3.0/examples/relational/relbench/README.md +53 -0
- structured_data_models-0.3.0/examples/relational/relbench/lag_target.py +47 -0
- structured_data_models-0.2.1/examples/relational/rel_bench.py → structured_data_models-0.3.0/examples/relational/relbench/main.py +45 -69
- structured_data_models-0.3.0/examples/relational/relbench/text.py +56 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/relational/salt/README.md +7 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/tabular/finetune.py +2 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/tabular/quickstart.py +1 -1
- structured_data_models-0.3.0/examples/timeseries/anomaly.py +27 -0
- structured_data_models-0.3.0/examples/timeseries/forecast.py +91 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/pyproject.toml +26 -20
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/__init__.py +14 -2
- structured_data_models-0.3.0/sdm/_kernels/flash_attention.py +75 -0
- structured_data_models-0.3.0/sdm/_memory.py +73 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_warnings.py +7 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/cache.py +38 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/ensemble.py +180 -98
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/evaluation/classification.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/explain/gradient.py +45 -2
- structured_data_models-0.3.0/sdm/models/__init__.py +40 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/_huggingface.py +14 -12
- structured_data_models-0.3.0/sdm/models/base/__init__.py +16 -0
- structured_data_models-0.3.0/sdm/models/base/callback.py +71 -0
- structured_data_models-0.2.1/sdm/models/base.py → structured_data_models-0.3.0/sdm/models/base/icl.py +119 -131
- structured_data_models-0.3.0/sdm/models/base/model.py +198 -0
- structured_data_models-0.3.0/sdm/models/chronos2/LICENSE +179 -0
- structured_data_models-0.3.0/sdm/models/chronos2/__init__.py +10 -0
- structured_data_models-0.3.0/sdm/models/chronos2/block.py +88 -0
- structured_data_models-0.3.0/sdm/models/chronos2/ckpt.py +52 -0
- structured_data_models-0.3.0/sdm/models/chronos2/icl.py +104 -0
- structured_data_models-0.3.0/sdm/models/chronos2/model.py +431 -0
- structured_data_models-0.3.0/sdm/models/chronos2/recipe.py +18 -0
- structured_data_models-0.3.0/sdm/models/chronos2/util.py +57 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/__init__.py +11 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/__init__.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/model.py +41 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/block.py +1 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/cell_embedding.py +10 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/icl.py +49 -11
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/model.py +71 -13
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/recipe.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/row_embedding.py +21 -3
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/__init__.py +18 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/NOTICE +5 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/__init__.py +11 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/block.py +161 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/ckpt.py +24 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/denoiser.py +107 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/embedding.py +168 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/model.py +357 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/noise_schedule.py +103 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/recipe.py +16 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/sampling.py +149 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/schedule.py +85 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/anomaly/threshold.py +226 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/__init__.py +11 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/block.py +83 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/ckpt.py +59 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/cross.py +48 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/darr.py +41 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/icl.py +87 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/model.py +390 -0
- structured_data_models-0.3.0/sdm/models/kumo/timeseries/forecast/recipe.py +22 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/block.py +9 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/cell_embedding.py +40 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/model.py +31 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/recipe.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/model.py +27 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/recipe.py +1 -1
- structured_data_models-0.3.0/sdm/models/timesfm3/LICENSE +209 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/__init__.py +11 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/block.py +107 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/ckpt.py +82 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/icl.py +110 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/model.py +387 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/recipe.py +35 -0
- structured_data_models-0.3.0/sdm/models/timesfm3/util.py +92 -0
- structured_data_models-0.3.0/sdm/models/timestral/LICENSE +204 -0
- structured_data_models-0.3.0/sdm/models/timestral/__init__.py +6 -0
- structured_data_models-0.3.0/sdm/models/timestral/block.py +127 -0
- structured_data_models-0.3.0/sdm/models/timestral/icl.py +93 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/__init__.py +6 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/attention.py +214 -52
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/glu.py +4 -4
- structured_data_models-0.3.0/sdm/nn/mlp.py +57 -0
- structured_data_models-0.3.0/sdm/nn/revin.py +173 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/rope.py +11 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/scaling.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/set_transformer.py +2 -2
- structured_data_models-0.3.0/sdm/nn/sinusoidal.py +39 -0
- structured_data_models-0.3.0/sdm/prior/__init__.py +159 -0
- structured_data_models-0.3.0/sdm/prior/_cdist.py +72 -0
- structured_data_models-0.3.0/sdm/prior/activations/__init__.py +53 -0
- structured_data_models-0.3.0/sdm/prior/activations/base.py +25 -0
- structured_data_models-0.3.0/sdm/prior/activations/callable.py +26 -0
- structured_data_models-0.3.0/sdm/prior/activations/fixed.py +59 -0
- structured_data_models-0.3.0/sdm/prior/activations/mixture.py +35 -0
- structured_data_models-0.3.0/sdm/prior/activations/parametric.py +86 -0
- structured_data_models-0.3.0/sdm/prior/activations/standardized.py +60 -0
- structured_data_models-0.3.0/sdm/prior/converters/__init__.py +35 -0
- structured_data_models-0.3.0/sdm/prior/converters/base.py +72 -0
- structured_data_models-0.3.0/sdm/prior/converters/distributional.py +293 -0
- structured_data_models-0.3.0/sdm/prior/converters/identity.py +19 -0
- structured_data_models-0.3.0/sdm/prior/converters/kumaraswamy.py +57 -0
- structured_data_models-0.3.0/sdm/prior/converters/mixture.py +64 -0
- structured_data_models-0.3.0/sdm/prior/converters/nearest_neighbor.py +138 -0
- structured_data_models-0.3.0/sdm/prior/converters/softmax.py +202 -0
- structured_data_models-0.3.0/sdm/prior/correlated.py +197 -0
- structured_data_models-0.3.0/sdm/prior/dag.py +243 -0
- structured_data_models-0.3.0/sdm/prior/filter.py +131 -0
- structured_data_models-0.3.0/sdm/prior/functions/__init__.py +39 -0
- structured_data_models-0.3.0/sdm/prior/functions/base.py +57 -0
- structured_data_models-0.3.0/sdm/prior/functions/concat.py +26 -0
- structured_data_models-0.3.0/sdm/prior/functions/discretization.py +55 -0
- structured_data_models-0.3.0/sdm/prior/functions/em.py +100 -0
- structured_data_models-0.3.0/sdm/prior/functions/gp.py +119 -0
- structured_data_models-0.3.0/sdm/prior/functions/linear.py +38 -0
- structured_data_models-0.3.0/sdm/prior/functions/mixture.py +72 -0
- structured_data_models-0.3.0/sdm/prior/functions/nn.py +76 -0
- structured_data_models-0.3.0/sdm/prior/functions/product.py +32 -0
- structured_data_models-0.3.0/sdm/prior/functions/quadratic.py +59 -0
- structured_data_models-0.3.0/sdm/prior/functions/reduce.py +46 -0
- structured_data_models-0.3.0/sdm/prior/functions/tree.py +85 -0
- structured_data_models-0.3.0/sdm/prior/hp.py +255 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/__init__.py +30 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/_normalize.py +20 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/activation.py +57 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/base.py +31 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/gaussian.py +33 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/kernel.py +128 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/mixture.py +42 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/power_law.py +36 -0
- structured_data_models-0.3.0/sdm/prior/matrix_samplers/singular.py +51 -0
- structured_data_models-0.3.0/sdm/prior/mixture.py +91 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/__init__.py +20 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/ball.py +31 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/base.py +27 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/correlated.py +44 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/mixture.py +43 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/normal.py +40 -0
- structured_data_models-0.3.0/sdm/prior/root_samplers/uniform.py +39 -0
- structured_data_models-0.3.0/sdm/prior/tabular/__init__.py +13 -0
- structured_data_models-0.3.0/sdm/prior/tabular/schema.py +150 -0
- structured_data_models-0.3.0/sdm/prior/tabular/scm.py +290 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/__init__.py +55 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/_scale.py +58 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/_table.py +100 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/base.py +34 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/choice.py +50 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/coarsen.py +146 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/correlate.py +175 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/identity.py +20 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/__init__.py +29 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/_count.py +27 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/latent.py +156 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/near_empty.py +145 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/random.py +45 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/missing/structured.py +372 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/recode.py +169 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/reject_nonfinite.py +27 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/relabel.py +185 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/scale_features.py +51 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/scale_target.py +28 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/shuffle_classes.py +30 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/shuffle_columns.py +40 -0
- structured_data_models-0.3.0/sdm/prior/tabular/transform/warp.py +94 -0
- structured_data_models-0.3.0/sdm/prior/util.py +83 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/__init__.py +24 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/base.py +29 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/__init__.py +2 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/_categorical.py +24 -12
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/align.py +141 -60
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/count.py +17 -7
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/impute.py +10 -6
- structured_data_models-0.3.0/sdm/processing/categorical/one_hot.py +62 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/categorical/shuffle.py +14 -7
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/callable.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/choice.py +1 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/drop.py +3 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/ensemble.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/identity.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/select.py +34 -8
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/sequential.py +6 -8
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/shuffle.py +5 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/stype.py +11 -10
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/table.py +2 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/task.py +3 -7
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/to_numerical.py +5 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/datetime/calendar.py +4 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/ensemble.py +16 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/execution.py +159 -67
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/__init__.py +10 -3
- structured_data_models-0.3.0/sdm/processing/numerical/asinh.py +18 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/cast.py +1 -1
- structured_data_models-0.3.0/sdm/processing/numerical/clip.py +91 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/constant.py +13 -5
- structured_data_models-0.3.0/sdm/processing/numerical/detrend.py +133 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/flip_sign.py +5 -2
- structured_data_models-0.3.0/sdm/processing/numerical/impute.py +102 -0
- structured_data_models-0.3.0/sdm/processing/numerical/interpolate.py +140 -0
- structured_data_models-0.3.0/sdm/processing/numerical/min_max_scale.py +52 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/pca.py +21 -12
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/power.py +34 -15
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/quantile.py +19 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/quantile_clip.py +21 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/random_projection.py +7 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/rank_gaussian.py +11 -7
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/robust_scale.py +10 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/sigma_clip.py +26 -13
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/standardize.py +36 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/output/__init__.py +2 -0
- structured_data_models-0.3.0/sdm/processing/output/clip.py +40 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/output/reduce.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/output/softmax.py +1 -1
- structured_data_models-0.3.0/sdm/processing/output/sort.py +39 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/recipe.py +49 -7
- structured_data_models-0.3.0/sdm/processing/series.py +17 -0
- structured_data_models-0.3.0/sdm/processing/target.py +13 -0
- structured_data_models-0.3.0/sdm/processing/text/llm.py +120 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/text/sentence_transformer.py +17 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/text/tfidf.py +5 -3
- structured_data_models-0.3.0/sdm/processing/text/tile.py +56 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/backend/_cugraph.py +8 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/data.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/sampler.py +6 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/task.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/stype.py +138 -61
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/categorical.py +114 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/columnar.py +185 -51
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/mixin.py +11 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/nullable.py +16 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/string.py +24 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/table.py +186 -64
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/var_len.py +134 -55
- {structured_data_models-0.2.1 → structured_data_models-0.3.0/structured_data_models.egg-info}/PKG-INFO +34 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/structured_data_models.egg-info/SOURCES.txt +249 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/structured_data_models.egg-info/requires.txt +3 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/structured_data_models.egg-info/scm_file_list.json +248 -8
- structured_data_models-0.3.0/structured_data_models.egg-info/scm_version.json +8 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/_kernels/test_segment_multi_reduce.py +1 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/conftest.py +19 -0
- structured_data_models-0.3.0/test/docs/test_nanotabicl.py +20 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/explain/test_base.py +1 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/explain/test_gradient.py +2 -0
- structured_data_models-0.2.1/test/models/test_base.py → structured_data_models-0.3.0/test/models/base/test_icl.py +56 -111
- structured_data_models-0.3.0/test/models/chronos2/test_model.py +54 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_batching.py +6 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_block.py +12 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_cell_embedding.py +22 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_model.py +106 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_recipe.py +4 -9
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_block.py +156 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_denoiser.py +41 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_embedding.py +78 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_model.py +25 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_noise_schedule.py +136 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_sampling.py +66 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_schedule.py +96 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_side_embedding.py +108 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/anomaly/test_threshold.py +24 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/forecast/test_cross.py +31 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/forecast/test_icl.py +37 -0
- structured_data_models-0.3.0/test/models/kumo/timeseries/forecast/test_model.py +77 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/tabfm/test_cell_embedding.py +26 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/tabiclv2/test_model.py +1 -1
- structured_data_models-0.3.0/test/models/timesfm3/test_icl.py +30 -0
- structured_data_models-0.3.0/test/models/timesfm3/test_model.py +41 -0
- structured_data_models-0.3.0/test/models/timesfm3/test_util.py +65 -0
- structured_data_models-0.3.0/test/models/timestral/test_icl.py +29 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_attention.py +38 -39
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_compile.py +25 -18
- structured_data_models-0.3.0/test/nn/test_mlp.py +24 -0
- structured_data_models-0.3.0/test/nn/test_revin.py +86 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_rms_norm.py +8 -0
- structured_data_models-0.3.0/test/nn/test_sinusoidal.py +26 -0
- structured_data_models-0.3.0/test/prior/tabular/test_schema.py +97 -0
- structured_data_models-0.3.0/test/prior/tabular/test_scm.py +227 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_choice.py +33 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_coarsen.py +64 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_correlate.py +39 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_latent.py +30 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_near_empty.py +36 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_random.py +49 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_recode.py +55 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_reject_nonfinite.py +26 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_relabel.py +51 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_scale_features.py +33 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_scale_target.py +23 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_shuffle_classes.py +22 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_shuffle_columns.py +33 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_structured.py +142 -0
- structured_data_models-0.3.0/test/prior/tabular/transform/test_warp.py +48 -0
- structured_data_models-0.3.0/test/prior/test_activation.py +126 -0
- structured_data_models-0.3.0/test/prior/test_converter.py +300 -0
- structured_data_models-0.3.0/test/prior/test_correlated.py +81 -0
- structured_data_models-0.3.0/test/prior/test_dag.py +114 -0
- structured_data_models-0.3.0/test/prior/test_filter.py +48 -0
- structured_data_models-0.3.0/test/prior/test_functions.py +285 -0
- structured_data_models-0.3.0/test/prior/test_hp.py +63 -0
- structured_data_models-0.3.0/test/prior/test_matrix_sampler.py +161 -0
- structured_data_models-0.3.0/test/prior/test_mixture.py +32 -0
- structured_data_models-0.3.0/test/prior/test_root.py +134 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/categorical/test_align.py +36 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/categorical/test_count.py +3 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/categorical/test_impute.py +4 -4
- structured_data_models-0.3.0/test/processing/categorical/test_one_hot.py +83 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/categorical/test_shuffle.py +12 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_choice.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_drop.py +10 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_select.py +2 -4
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_sequential.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_shuffle.py +4 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_stype.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_table.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_to_numerical.py +6 -7
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/datetime/test_calendar.py +5 -5
- structured_data_models-0.3.0/test/processing/numerical/test_asinh.py +17 -0
- structured_data_models-0.3.0/test/processing/numerical/test_clip.py +109 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_constant.py +4 -4
- structured_data_models-0.3.0/test/processing/numerical/test_detrend.py +78 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_flip_sign.py +20 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_impute.py +35 -1
- structured_data_models-0.3.0/test/processing/numerical/test_interpolate.py +75 -0
- structured_data_models-0.3.0/test/processing/numerical/test_min_max_scale.py +27 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_pca.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_power.py +1 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_sigma_clip.py +37 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_standardize.py +48 -0
- structured_data_models-0.3.0/test/processing/output/test_clip.py +33 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/output/test_sort.py +18 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/test_contract.py +18 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/test_ensemble.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/test_execution.py +1 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/test_fitted_guard.py +2 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/test_recipe.py +13 -3
- structured_data_models-0.3.0/test/processing/text/test_llm.py +36 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/text/test_sentence_transformer.py +64 -2
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/text/test_tfidf.py +2 -2
- structured_data_models-0.3.0/test/processing/text/test_tile.py +52 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/relational/backend/test_cugraph.py +10 -5
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/relational/test_sampler.py +9 -9
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_categorical.py +72 -1
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_columnar.py +39 -6
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_string.py +24 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_table.py +57 -3
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_var_len.py +67 -1
- structured_data_models-0.3.0/test/test_cache.py +120 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/test_ensemble.py +58 -44
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/test_stype.py +1 -36
- structured_data_models-0.3.0/third_party/dpm-solver/LICENSE +21 -0
- structured_data_models-0.3.0/train/kumo/tabular/README.md +91 -0
- structured_data_models-0.3.0/train/kumo/tabular/loader.py +299 -0
- structured_data_models-0.3.0/train/kumo/tabular/optim.py +186 -0
- structured_data_models-0.3.0/train/kumo/tabular/prior.py +414 -0
- structured_data_models-0.3.0/train/kumo/tabular/train.py +557 -0
- structured_data_models-0.3.0/uv.lock +2721 -0
- structured_data_models-0.2.1/THIRD_PARTY_LICENSES.md +0 -40
- structured_data_models-0.2.1/docs/source/_templates/autosummary/model_class.rst +0 -42
- structured_data_models-0.2.1/docs/source/api/models.rst +0 -54
- structured_data_models-0.2.1/docs/source/api/sdm.rst +0 -10
- structured_data_models-0.2.1/sdm/_memory.py +0 -34
- structured_data_models-0.2.1/sdm/models/__init__.py +0 -20
- structured_data_models-0.2.1/sdm/models/callback.py +0 -106
- structured_data_models-0.2.1/sdm/processing/numerical/clip.py +0 -48
- structured_data_models-0.2.1/sdm/processing/numerical/clip_soft.py +0 -54
- structured_data_models-0.2.1/sdm/processing/numerical/impute.py +0 -52
- structured_data_models-0.2.1/sdm/processing/output/sort.py +0 -16
- structured_data_models-0.2.1/structured_data_models.egg-info/scm_version.json +0 -8
- structured_data_models-0.2.1/test/processing/numerical/test_clip.py +0 -37
- structured_data_models-0.2.1/test/processing/numerical/test_clip_soft.py +0 -82
- structured_data_models-0.2.1/test/test_cache.py +0 -47
- structured_data_models-0.2.1/uv.lock +0 -2432
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.agents/skills/port-external-models/SKILL.md +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.agents/skills/recipe-development/SKILL.md +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/CODEOWNERS +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/copy-pr-bot.yaml +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/lint.yml +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/publish.yml +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/.pre-commit-config.yaml +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/CODE_OF_CONDUCT.md +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/LICENSE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/SECURITY.md +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/relational/relarena/kumo_relational.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/beyondarena/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/beyondarena/evaluate.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/scoringbench/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/scoringbench/evaluate.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/tabarena/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/tabarena/evaluate.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/talent/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/benchmark/tabular/talent/evaluate.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/Makefile +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_static/custom.css +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_templates/autosummary/function.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_templates/autosummary/processor_class.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_templates/autosummary/tensor_class.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/_templates/partials/site-foot.html +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/evaluation.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/explain.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/cache.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/device_mixin.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/ensemble_table.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/kv_cache_entry.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/related_tables_schema.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/relational_sampler_output.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/internal/table_schema.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/nn.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/relational.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/api/tensor.rst +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/ensemble_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/ensemble_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/graph_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/graph_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/kumo_relational_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/kumo_relational_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/kumo_tabular.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/logo_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/logo_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/nvidia.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/pipeline_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/pipeline_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/tabfm_dark.png +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/tabfm_light.png +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/tabicl_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/tabicl_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/table_dark.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/docs/source/images/table_light.svg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/relational/salt/main.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/examples/tabular/text.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_constants.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_inference.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/rmsnorm_cast.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/segment_multi_reduce.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/triton/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/triton/rmsnorm_cast.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/_kernels/triton/segment_multi_reduce.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/evaluation/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/explain/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/explain/base.py +0 -0
- {structured_data_models-0.2.1/sdm/models → structured_data_models-0.3.0/sdm/models/base}/ecoc.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/NOTICE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/graph.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/invariant_gnn.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/recipe.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/relational/task.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/kumo/tabular/batching.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/LICENSE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/ckpt.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/icl.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabfm/row_embedding.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/LICENSE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/block.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/ckpt.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/icl.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/models/tabiclv2/row_embedding.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/_buffer.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/linear.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/rms_norm.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/nn/softplus.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/common/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/datetime/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/numerical/_stats.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/processing/text/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/backend/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/backend/_pyg_lib.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/relational/join.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/task.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/io/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/io/arrow.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/tensor/io/cudf.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/testing/__init__.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/testing/decorators.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/sdm/testing/memory.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/setup.cfg +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/structured_data_models.egg-info/dependency_links.txt +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/structured_data_models.egg-info/top_level.txt +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/_kernels/test_rmsnorm_cast.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/evaluation/test_classification.py +0 -0
- {structured_data_models-0.2.1/test/models → structured_data_models-0.3.0/test/models/base}/test_ecoc.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/relational/test_model.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_icl.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/kumo/tabular/test_row_embedding.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/tabfm/test_model.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/models/tabiclv2/test_icl.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_buffer.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_glu.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_linear.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_rope.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_scaling.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/nn/test_set_transformer.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_identity.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/common/test_task.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_cast.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_quantile.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_quantile_clip.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_random_projection.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_rank_gaussian.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/numerical/test_robust_scale.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/output/test_reduce.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/processing/output/test_softmax.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/relational/backend/test_sampler_parity.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/relational/test_data.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/relational/test_join.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_io.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_nullable.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/test/tensor/test_record_stream.py +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/third_party/contributor-covenant/LICENSE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/third_party/pytorch/LICENSE +0 -0
- {structured_data_models-0.2.1 → structured_data_models-0.3.0}/third_party/tabpfn-extensions/LICENSE +0 -0
{structured_data_models-0.2.1 → structured_data_models-0.3.0}/.agents/skills/docstring/SKILL.md
RENAMED
|
@@ -48,9 +48,6 @@ class MyClass:
|
|
|
48
48
|
Returns:
|
|
49
49
|
Tensor with shape ``[..., S, C]``.
|
|
50
50
|
"""
|
|
51
|
-
|
|
52
|
-
def citation(self) -> str:
|
|
53
|
-
"""Return the key from :func:`sdm.refs.resolve_default_citation_key`.""" # noqa: E501
|
|
54
51
|
```
|
|
55
52
|
|
|
56
53
|
## Sphinx Cross-Referencing Reference
|
|
@@ -75,12 +72,12 @@ in `docs/source/conf.py`).
|
|
|
75
72
|
```python
|
|
76
73
|
r"""
|
|
77
74
|
- :mod:`sdm.nn`: Module reference.
|
|
78
|
-
- :class:`~sdm.nn.
|
|
75
|
+
- :class:`~sdm.nn.Attention`: Class reference. External targets use the same
|
|
79
76
|
role: :class:`pandas.DataFrame`, :class:`torch.nn.Module`, :class:`dict`,
|
|
80
77
|
:class:`TypeError`, :class:`collections.abc.Mapping`.
|
|
81
78
|
- :class:`~pyarrow.Array`: Leading ``~`` renders only ``Array``. Use it in
|
|
82
79
|
summary lines; keep the full path in ``Args:`` entries.
|
|
83
|
-
- :meth:`~sdm.nn.
|
|
80
|
+
- :meth:`~sdm.nn.Attention.forward`: Method reference.
|
|
84
81
|
- :func:`torch.nn.functional.scaled_dot_product_attention`: Function
|
|
85
82
|
reference. Use :func: for free functions and :meth: only for methods.
|
|
86
83
|
- :attr:`attribute`: Attribute reference.
|
|
@@ -99,5 +96,5 @@ r"""
|
|
|
99
96
|
|
|
100
97
|
## Verification
|
|
101
98
|
|
|
102
|
-
Run `
|
|
99
|
+
Run `uvx ruff check`. This is a structural backstop only: it flags missing public class/method/function docstrings, capitalization, end punctuation, `r"""` for backslashes, and `Args:`/signature mismatches.
|
|
103
100
|
It does **not** require an `Args:` section to exist, or verify citations, shape notation, or behavior notes — confirm those by hand against the principles above.
|
|
@@ -11,7 +11,7 @@ description: Create or modify reusable SDM processors and focused tests. Use whe
|
|
|
11
11
|
|
|
12
12
|
02. **Keep processors general**: Before adding behavior to a processor, check whether it describes the processing operation itself or a specific model/recipe use case. Keep only the former in sdm.processing.
|
|
13
13
|
|
|
14
|
-
03. **Learn from existing processors**: Before implementing, inspect the target processor and a few relevant existing processors and tests. Reuse established SDM API & coding patterns rather than introducing a new approach. When modifying an existing processor, understand its current behavior and tests before changing it. For standard preprocessing operations, check the corresponding scikit-learn API and behavior before designing a new interface. Consult cuML when a comparable GPU implementation exists and its implementation is useful for SDM.
|
|
14
|
+
03. **Learn from existing processors**: Before implementing, inspect the target processor and a few relevant existing processors and tests. Reuse established SDM API & coding patterns rather than introducing a new approach. When modifying an existing processor, understand its current behavior and tests before changing it. For standard preprocessing operations, check the corresponding scikit-learn API and behavior before designing a new interface. Consult cuML when a comparable GPU implementation exists and its implementation is useful for SDM. Use model-specific libraries such as TabICL or TabPFN to establish processor needs and numerical semantics, while deriving API design and implementation style from SDM conventions and established general-purpose libraries such as scikit-learn and cuML.
|
|
15
15
|
|
|
16
16
|
04. **Choose the simplest processor type**: Use a regular `Processor` when its state and behavior are defined per table or batch; `EnsembleProcessorAdapter` handles its use with ensembles. Implement `EnsembleProcessor` directly only when state or behavior is scoped to logical ensemble members, depends on multiple members, or changes the ensemble structure.
|
|
17
17
|
|
|
@@ -19,6 +19,7 @@ reviews:
|
|
|
19
19
|
in_progress_fortune: false
|
|
20
20
|
suggested_labels: false
|
|
21
21
|
suggested_reviewers: false
|
|
22
|
+
related_prs: false
|
|
22
23
|
|
|
23
24
|
path_filters:
|
|
24
25
|
- '!uv.lock'
|
|
@@ -46,11 +47,21 @@ reviews:
|
|
|
46
47
|
dependencies optional). In `__init__.py`, imports and `__all__`
|
|
47
48
|
are intentionally ordered by dependency, not alphabetically — do
|
|
48
49
|
not suggest alphabetical sorting there.
|
|
50
|
+
- path: 'sdm/_kernels/**'
|
|
51
|
+
instructions: >-
|
|
52
|
+
For changes introducing or expanding a custom GPU kernel, check
|
|
53
|
+
whether the performance evidence includes a practical
|
|
54
|
+
compiler-generated or existing-library alternative. If absent,
|
|
55
|
+
request a comparison or a concrete explanation of incompatibility.
|
|
56
|
+
Check that comparisons preserve workload and numerical requirements
|
|
57
|
+
and distinguish startup from prediction latency. Treat this as a
|
|
58
|
+
justification question; do not infer performance from source code
|
|
59
|
+
alone.
|
|
49
60
|
- path: 'test/**/*.py'
|
|
50
61
|
instructions: >-
|
|
51
62
|
Pytest-based tests. Check that behavior changes in sdm/ have
|
|
52
|
-
corresponding test coverage,
|
|
53
|
-
|
|
63
|
+
corresponding test coverage, seeds or generators are used only
|
|
64
|
+
when required by the tested behavior, and parametrization covers relevant
|
|
54
65
|
dtype/device variants. Plain pytest functions are preferred over
|
|
55
66
|
unittest-style classes.
|
|
56
67
|
- path: 'examples/**'
|
|
@@ -65,6 +76,8 @@ reviews:
|
|
|
65
76
|
|
|
66
77
|
slop_detection:
|
|
67
78
|
enabled: true
|
|
79
|
+
include_all_authors: true
|
|
80
|
+
label: slop
|
|
68
81
|
|
|
69
82
|
finishing_touches:
|
|
70
83
|
docstrings:
|
|
@@ -125,5 +138,6 @@ code_generation:
|
|
|
125
138
|
instructions: >-
|
|
126
139
|
Place tests under test/ mirroring the sdm/ package layout. Use
|
|
127
140
|
plain pytest functions with `pytest.mark.parametrize` for
|
|
128
|
-
dtype/device variants
|
|
129
|
-
|
|
141
|
+
dtype/device variants and `sdm.testing` utilities where applicable.
|
|
142
|
+
Do not set seeds unless the tested behavior requires them; use
|
|
143
|
+
explicit generators when controlled randomness is necessary.
|
|
@@ -0,0 +1,59 @@
|
|
|
1
|
+
name: 🐛 Bug Report
|
|
2
|
+
description: Create a report to help us reproduce and fix the bug
|
|
3
|
+
labels: [bug]
|
|
4
|
+
body:
|
|
5
|
+
- type: markdown
|
|
6
|
+
attributes:
|
|
7
|
+
value: |
|
|
8
|
+
Thanks for taking the time to report this problem! Issues that do not follow the checklist below may be closed.
|
|
9
|
+
- type: checkboxes
|
|
10
|
+
attributes:
|
|
11
|
+
label: Checklist
|
|
12
|
+
description: Please confirm the following before submitting.
|
|
13
|
+
options:
|
|
14
|
+
- label: I have reviewed the [contributing guidelines](https://github.com/Nvidia/structured-data-models/blob/main/CONTRIBUTING.md) and [AI policy](https://github.com/Nvidia/structured-data-models/blob/main/AGENTS.md#ai-policy).
|
|
15
|
+
required: true
|
|
16
|
+
- label: I have searched the [existing and past issues](https://github.com/Nvidia/structured-data-models/issues?q=is%3Aissue+sort%3Acreated-desc+) and this bug has not been reported yet.
|
|
17
|
+
required: true
|
|
18
|
+
- label: I have checked whether the bug also reproduces on the latest `main`.
|
|
19
|
+
- type: textarea
|
|
20
|
+
attributes:
|
|
21
|
+
label: Describe the bug
|
|
22
|
+
description: A clear and concise description of what the bug is, and what you expected to happen instead.
|
|
23
|
+
validations:
|
|
24
|
+
required: true
|
|
25
|
+
- type: textarea
|
|
26
|
+
attributes:
|
|
27
|
+
label: Reproducible code example
|
|
28
|
+
description: |
|
|
29
|
+
A minimal, self-contained script that we can copy and run to reproduce the bug. Please trim down any irrelevant code.
|
|
30
|
+
Use synthetic or publicly available data, and include any required checkpoint name or revision.
|
|
31
|
+
render: python
|
|
32
|
+
placeholder: |
|
|
33
|
+
# All necessary imports at the beginning
|
|
34
|
+
import torch
|
|
35
|
+
import sdm
|
|
36
|
+
|
|
37
|
+
# A succinct reproducing example trimmed down to the essential parts:
|
|
38
|
+
...
|
|
39
|
+
validations:
|
|
40
|
+
required: true
|
|
41
|
+
- type: textarea
|
|
42
|
+
attributes:
|
|
43
|
+
label: Error message or observed output
|
|
44
|
+
description: The **full** traceback of the exception, or the output that shows the incorrect behavior.
|
|
45
|
+
render: text
|
|
46
|
+
validations:
|
|
47
|
+
required: true
|
|
48
|
+
- type: textarea
|
|
49
|
+
attributes:
|
|
50
|
+
label: Versions
|
|
51
|
+
description: |
|
|
52
|
+
Your structured-data-models version or Git commit, how you installed it (e.g. pip, editable source checkout, or container image and tag), and the output of `python -m torch.utils.collect_env`.
|
|
53
|
+
render: text
|
|
54
|
+
validations:
|
|
55
|
+
required: true
|
|
56
|
+
- type: markdown
|
|
57
|
+
attributes:
|
|
58
|
+
value: |
|
|
59
|
+
Thanks for contributing 🎉!
|
|
@@ -0,0 +1,35 @@
|
|
|
1
|
+
<!--
|
|
2
|
+
Keep these headings and fill each section.
|
|
3
|
+
Do not list changes file by file, symbol by symbol, or commit by commit in the three main sections.
|
|
4
|
+
Do not add an attribution footer, badge, or emoji signature.
|
|
5
|
+
-->
|
|
6
|
+
|
|
7
|
+
# Summary
|
|
8
|
+
|
|
9
|
+
<!-- Summarize the change for a human in at most two prose sentences. A short code block may follow. -->
|
|
10
|
+
|
|
11
|
+
# Description of the change
|
|
12
|
+
|
|
13
|
+
<!--
|
|
14
|
+
Explain what changed and why.
|
|
15
|
+
Include only details needed to review it.
|
|
16
|
+
Do not repeat the summary.
|
|
17
|
+
For bug fixes, link an existing GitHub issue or provide a minimal Python script, full error messages, and environment details sufficient to reproduce the bug.
|
|
18
|
+
-->
|
|
19
|
+
|
|
20
|
+
# Validation
|
|
21
|
+
|
|
22
|
+
<!--
|
|
23
|
+
Report manual checks, benchmarks, or tests CI does not run; write n/a if none.
|
|
24
|
+
Omit results already reported by CI.
|
|
25
|
+
-->
|
|
26
|
+
|
|
27
|
+
<details><summary>Additional details</summary>
|
|
28
|
+
|
|
29
|
+
<!--
|
|
30
|
+
Put any remaining useful detail here.
|
|
31
|
+
This section is usually not intended for humans to read.
|
|
32
|
+
Delete this block if empty.
|
|
33
|
+
-->
|
|
34
|
+
|
|
35
|
+
</details>
|
|
@@ -34,7 +34,7 @@ jobs:
|
|
|
34
34
|
- name: Build documentation
|
|
35
35
|
working-directory: ./docs
|
|
36
36
|
run: |
|
|
37
|
-
uv run --no-project sphinx-build -M html "source" "build" -W --keep-going
|
|
37
|
+
uv run --no-project sphinx-build -M html "source" "build" -n -W --keep-going
|
|
38
38
|
|
|
39
39
|
- name: Upload documentation
|
|
40
40
|
uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
|
{structured_data_models-0.2.1 → structured_data_models-0.3.0}/.github/workflows/test-gpu.yml
RENAMED
|
@@ -133,7 +133,13 @@ jobs:
|
|
|
133
133
|
uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
|
|
134
134
|
with:
|
|
135
135
|
path: /home/runner/.cache/data_foundry
|
|
136
|
-
key: beyondarena-
|
|
136
|
+
key: beyondarena-parkinsons-biomedical-voice-measurements
|
|
137
|
+
|
|
138
|
+
- name: Cache TabArena dataset
|
|
139
|
+
uses: actions/cache@27d5ce7f107fe9357f9df03efb73ab90386fccae # v5.0.5
|
|
140
|
+
with:
|
|
141
|
+
path: /home/runner/.cache/openml
|
|
142
|
+
key: tabarena-blood-transfusion-service-center
|
|
137
143
|
|
|
138
144
|
- name: Install uv
|
|
139
145
|
uses: astral-sh/setup-uv@fac544c07dec837d0ccb6301d7b5580bf5edae39 # v8.2.0
|
|
@@ -199,7 +205,21 @@ jobs:
|
|
|
199
205
|
env:
|
|
200
206
|
HF_TOKEN: ${{ secrets.HF_TOKEN }}
|
|
201
207
|
run: |
|
|
202
|
-
python examples/relational/
|
|
208
|
+
python examples/relational/relbench/main.py --dataset rel-f1 --task driver-position
|
|
209
|
+
|
|
210
|
+
# Keep slow OpenML downloads outside the benchmark's timeout limit.
|
|
211
|
+
- name: Download TabArena dataset
|
|
212
|
+
timeout-minutes: 2
|
|
213
|
+
run: |
|
|
214
|
+
python - <<'PY'
|
|
215
|
+
from tabarena.contexts import TabArenaContext
|
|
216
|
+
|
|
217
|
+
context = TabArenaContext()
|
|
218
|
+
tasks = context.task_metadata_collection.subset_tasks(
|
|
219
|
+
dataset_names=["blood-transfusion-service-center"],
|
|
220
|
+
)
|
|
221
|
+
tasks.materialize()
|
|
222
|
+
PY
|
|
203
223
|
|
|
204
224
|
- name: Test TabArena benchmark
|
|
205
225
|
timeout-minutes: 5
|
|
@@ -17,6 +17,10 @@ permissions:
|
|
|
17
17
|
jobs:
|
|
18
18
|
pytest:
|
|
19
19
|
runs-on: ubuntu-latest
|
|
20
|
+
strategy:
|
|
21
|
+
fail-fast: false
|
|
22
|
+
matrix:
|
|
23
|
+
python-version: ['3.11', '3.15']
|
|
20
24
|
|
|
21
25
|
steps:
|
|
22
26
|
- uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6.0.3
|
|
@@ -24,7 +28,7 @@ jobs:
|
|
|
24
28
|
- name: Install uv
|
|
25
29
|
uses: astral-sh/setup-uv@fac544c07dec837d0ccb6301d7b5580bf5edae39 # v8.2.0
|
|
26
30
|
with:
|
|
27
|
-
python-version:
|
|
31
|
+
python-version: ${{ matrix.python-version }}
|
|
28
32
|
activate-environment: true
|
|
29
33
|
|
|
30
34
|
- name: Install package
|
|
@@ -33,18 +37,25 @@ jobs:
|
|
|
33
37
|
uv pip install \
|
|
34
38
|
--torch-backend=cpu \
|
|
35
39
|
--find-links=https://data.pyg.org/whl/torch-2.14.0+cpu.html \
|
|
36
|
-
--group=test \
|
|
40
|
+
--group=test-cpu \
|
|
37
41
|
-e .
|
|
38
42
|
|
|
39
43
|
- name: Run tests
|
|
40
44
|
env:
|
|
41
45
|
FULL_TEST: ${{ ((github.event_name == 'push' && github.ref_name == 'main') || contains(github.event.pull_request.labels.*.name, 'ci-full-test')) && '1' || '0' }}
|
|
42
46
|
run: |
|
|
43
|
-
pytest --cov=sdm --cov-config=pyproject.toml --cov-report=term-missing --durations=10
|
|
44
|
-
|
|
45
|
-
|
|
46
|
-
|
|
47
|
-
|
|
48
|
-
|
|
49
|
-
|
|
50
|
-
|
|
47
|
+
pytest --cov=sdm --cov-config=pyproject.toml --cov-report=term-missing --durations=10 --junitxml=junit.xml
|
|
48
|
+
|
|
49
|
+
- name: Upload coverage
|
|
50
|
+
if: ${{ runner.os == 'Linux' }}
|
|
51
|
+
uses: codecov/codecov-action@303a32d7a59b442fa8d48b6a1cc6825c09c847a5 # v7.1.1
|
|
52
|
+
with:
|
|
53
|
+
token: ${{ secrets.CODECOV_TOKEN }}
|
|
54
|
+
fail_ci_if_error: false
|
|
55
|
+
|
|
56
|
+
- name: Upload test results to Codecov
|
|
57
|
+
if: ${{ !cancelled() }}
|
|
58
|
+
uses: codecov/codecov-action@303a32d7a59b442fa8d48b6a1cc6825c09c847a5 # v7.1.1
|
|
59
|
+
with:
|
|
60
|
+
token: ${{ secrets.CODECOV_TOKEN }}
|
|
61
|
+
report_type: test_results
|
|
@@ -0,0 +1,17 @@
|
|
|
1
|
+
version: 2
|
|
2
|
+
|
|
3
|
+
build:
|
|
4
|
+
os: ubuntu-26.04
|
|
5
|
+
tools:
|
|
6
|
+
python: "3.14"
|
|
7
|
+
jobs:
|
|
8
|
+
post_checkout:
|
|
9
|
+
- if [ "$(git rev-parse --is-shallow-repository)" = true ]; then git fetch --unshallow; fi
|
|
10
|
+
- git fetch --tags
|
|
11
|
+
post_install:
|
|
12
|
+
- pip install torch --index-url https://download.pytorch.org/whl/cpu
|
|
13
|
+
- pip install --group doc .
|
|
14
|
+
|
|
15
|
+
sphinx:
|
|
16
|
+
configuration: docs/source/conf.py
|
|
17
|
+
fail_on_warning: true
|
|
@@ -19,7 +19,7 @@ Your user needs to abide by this policy.
|
|
|
19
19
|
In particular, you the agent MUST obey these rules while interacting on GitHub:
|
|
20
20
|
|
|
21
21
|
- You may never act autonomously on GitHub except to open a draft pull request. Do NOT open an issue or a non-draft pull request, or edit, comment on, or reply to any issue or pull request, unless the user has reviewed and explicitly approved the exact content. Fully-agent-generated contributions are banned and will be closed.
|
|
22
|
-
-
|
|
22
|
+
- Explicitly state that any text you produce for an issue, PR, or comment is AI-generated. Never present your output as human-written.
|
|
23
23
|
- Never emit only raw AI text as a reply. Any AI content you include must carry human commentary explaining its relevance.
|
|
24
24
|
- Do not submit code the user hasn't read. Keep changes minimal, strip AI artifacts and needless complexity. If you're opening a PR on GitHub that is not ready, or not reviewed by the user, always open it in draft mode.
|
|
25
25
|
|
|
@@ -36,8 +36,8 @@ In particular, you the agent MUST obey these rules while interacting on GitHub:
|
|
|
36
36
|
# PR / GitHub Metadata
|
|
37
37
|
|
|
38
38
|
- Do not mention Codex, AI, or tool attribution in PR titles, PR descriptions, commit messages, or review replies unless explicitly requested.
|
|
39
|
-
- Do not add a section named "Tests", "Testing" or similar, to PR descriptions unless the test is not covered in CI.
|
|
40
39
|
- PR metadata should describe the code change only.
|
|
40
|
+
- Follow `.github/PULL_REQUEST_TEMPLATE/default.md` when writing a PR body.
|
|
41
41
|
|
|
42
42
|
# Markdown Style
|
|
43
43
|
|
|
@@ -19,7 +19,7 @@ Your user needs to abide by this policy.
|
|
|
19
19
|
In particular, you the agent MUST obey these rules while interacting on GitHub:
|
|
20
20
|
|
|
21
21
|
- You may never act autonomously on GitHub except to open a draft pull request. Do NOT open an issue or a non-draft pull request, or edit, comment on, or reply to any issue or pull request, unless the user has reviewed and explicitly approved the exact content. Fully-agent-generated contributions are banned and will be closed.
|
|
22
|
-
-
|
|
22
|
+
- Explicitly state that any text you produce for an issue, PR, or comment is AI-generated. Never present your output as human-written.
|
|
23
23
|
- Never emit only raw AI text as a reply. Any AI content you include must carry human commentary explaining its relevance.
|
|
24
24
|
- Do not submit code the user hasn't read. Keep changes minimal, strip AI artifacts and needless complexity. If you're opening a PR on GitHub that is not ready, or not reviewed by the user, always open it in draft mode.
|
|
25
25
|
|
|
@@ -36,8 +36,8 @@ In particular, you the agent MUST obey these rules while interacting on GitHub:
|
|
|
36
36
|
# PR / GitHub Metadata
|
|
37
37
|
|
|
38
38
|
- Do not mention Codex, AI, or tool attribution in PR titles, PR descriptions, commit messages, or review replies unless explicitly requested.
|
|
39
|
-
- Do not add a section named "Tests", "Testing" or similar, to PR descriptions unless the test is not covered in CI.
|
|
40
39
|
- PR metadata should describe the code change only.
|
|
40
|
+
- Follow `.github/PULL_REQUEST_TEMPLATE/default.md` when writing a PR body.
|
|
41
41
|
|
|
42
42
|
# Markdown Style
|
|
43
43
|
|
|
@@ -6,11 +6,8 @@ into three categories:
|
|
|
6
6
|
1. You want to report a bug, feature request, or documentation issue
|
|
7
7
|
- File an [issue](https://github.com/NVIDIA/structured-data-models/issues/new/choose)
|
|
8
8
|
describing what you encountered or what you want to see changed.
|
|
9
|
-
- Please run and paste the output of
|
|
10
|
-
|
|
11
|
-
- The RAPIDS team will evaluate the issues and triage them, scheduling
|
|
12
|
-
them for a release. If you believe the issue needs priority attention
|
|
13
|
-
comment on the issue to notify the team.
|
|
9
|
+
- Please run and paste the output of `python -m torch.utils.collect_env` while reporting a bug to gather and report relevant environment details.
|
|
10
|
+
- The project maintainers will evaluate the issues and triage them, scheduling them for a release. If you believe the issue needs priority attention, comment on the issue to notify the team.
|
|
14
11
|
2. You want to propose a new Feature and implement it
|
|
15
12
|
- Post about your intended feature, and we shall discuss the design and
|
|
16
13
|
implementation.
|
|
@@ -26,18 +23,16 @@ into three categories:
|
|
|
26
23
|
### Your first issue
|
|
27
24
|
|
|
28
25
|
1. Read the project's [README.md](https://github.com/NVIDIA/structured-data-models/blob/main/README.md)
|
|
29
|
-
to learn how to
|
|
26
|
+
to learn how to install the package.
|
|
30
27
|
2. Find an issue to work on. The best way is to look for the [good first issue](https://github.com/NVIDIA/structured-data-models/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22)
|
|
31
28
|
or [help wanted](https://github.com/NVIDIA/structured-data-models/issues?q=is%3Aissue+is%3Aopen+label%3A%22help+wanted%22) labels
|
|
32
29
|
3. Comment on the issue saying you are going to work on it.
|
|
33
|
-
4. Get familiar with the
|
|
34
|
-
- For C++ developers it is available here [DEVELOPER_GUIDE.md](https://github.com/jarmak-nv/rapids-repo-template/blob/main/Maintainers/DEVELOPER_GUIDE.md)
|
|
35
|
-
- For Python developers, a [cuDF SAMPLE Python DEVELOPER_GUIDE.md](https://docs.rapids.ai/api/cudf/stable/developer_guide/index.html) is available as well.
|
|
30
|
+
4. Get familiar with the [project guidelines](AGENTS.md).
|
|
36
31
|
5. Code! Make sure to update unit tests!
|
|
37
|
-
6. When done, [create your pull request](https://github.com/NVIDIA/structured-data-models/compare).
|
|
32
|
+
6. When done, [create your pull request](https://github.com/NVIDIA/structured-data-models/compare?template=default.md) and fill in the template it loads.
|
|
38
33
|
7. Verify that CI passes all [status checks](https://help.github.com/articles/about-status-checks/), or fix if needed.
|
|
39
34
|
8. Wait for other developers to review your code and update code as needed.
|
|
40
|
-
9. Once reviewed and approved, a
|
|
35
|
+
9. Once reviewed and approved, a project maintainer will merge your pull request.
|
|
41
36
|
|
|
42
37
|
Remember, if you are unsure about anything, don't hesitate to comment on issues and ask for clarifications!
|
|
43
38
|
|
|
@@ -103,8 +98,7 @@ Remember, if you are unsure about anything, don't hesitate to comment on issues
|
|
|
103
98
|
Each PR must be labeled according to whether it is a "breaking" or "non-breaking" change (using GitHub labels). This is used to highlight changes that users should know about when upgrading.
|
|
104
99
|
|
|
105
100
|
For structured-data-models, a "breaking" change is one that modifies the public, non-experimental, Python API in a
|
|
106
|
-
non-backward-compatible way.
|
|
107
|
-
time, so changes to it are not typically considered breaking. Backward-compatible API changes to the Python
|
|
101
|
+
non-backward-compatible way. Backward-compatible API changes to the Python
|
|
108
102
|
API (such as adding a new keyword argument to a function) do not need to be labeled.
|
|
109
103
|
|
|
110
104
|
Additional labels must be applied to indicate whether the change is a feature, improvement, bugfix, or documentation change. See the shared RAPIDS documentation for these labels: https://github.com/NVIDIA/kb/issues/42.
|
|
@@ -124,19 +118,12 @@ implementation of the issue, ask them in the issue instead of the PR.
|
|
|
124
118
|
|
|
125
119
|
### Branches and Versions
|
|
126
120
|
|
|
127
|
-
|
|
128
|
-
|
|
129
|
-
1. `main` branch: it contains the last released version. Only hotfixes are targeted and merged into it.
|
|
130
|
-
2. `branch-x.y`: it is the development branch which contains the upcoming release. All the new features should be based on this branch and Merge/Pull request should target this branch (with the exception of hotfixes).
|
|
121
|
+
Development targets `main` by default. Maintainers may request a `release/X.Y` branch for release-specific changes.
|
|
131
122
|
|
|
132
123
|
### Additional details
|
|
133
124
|
|
|
134
|
-
For every new version `x.y` of structured-data-models there is a corresponding branch called `branch-x.y`, from where new feature development starts and PRs will be targeted and merged before its release. The exceptions to this are the 'hotfixes' that target the `main` branch, which target critical issues raised by GitHub users and are directly merged to `main` branch, and create a new subversion of the project. While trying to patch an issue which requires a 'hotfix', please state the intent in the PR.
|
|
135
|
-
|
|
136
125
|
For all development, your changes should be pushed into a branch (created using the naming instructions below) in your own fork of structured-data-models and then create a pull request when the code is ready.
|
|
137
126
|
|
|
138
|
-
A few days before releasing version `x.y` the code of the current development branch (`branch-x.y`) will be frozen and a new branch, 'branch-x+1.y' will be created to continue development.
|
|
139
|
-
|
|
140
127
|
### Branch naming
|
|
141
128
|
|
|
142
129
|
Branches used to create PRs should have a name of the form `<type>-<name>`
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: structured-data-models
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.3.0
|
|
4
4
|
Summary: Foundation Models for Structured Data
|
|
5
5
|
Author: NVIDIA
|
|
6
6
|
Maintainer: NVIDIA
|
|
@@ -16,14 +16,20 @@ Classifier: Programming Language :: Python :: 3.11
|
|
|
16
16
|
Classifier: Programming Language :: Python :: 3.12
|
|
17
17
|
Classifier: Programming Language :: Python :: 3.13
|
|
18
18
|
Classifier: Programming Language :: Python :: 3.14
|
|
19
|
+
Classifier: Programming Language :: Python :: 3.15
|
|
19
20
|
Requires-Python: >=3.11
|
|
20
21
|
Description-Content-Type: text/markdown
|
|
21
22
|
License-File: LICENSE
|
|
22
23
|
License-File: THIRD_PARTY_LICENSES.md
|
|
24
|
+
License-File: sdm/models/chronos2/LICENSE
|
|
23
25
|
License-File: sdm/models/tabiclv2/LICENSE
|
|
24
26
|
License-File: sdm/models/tabfm/LICENSE
|
|
27
|
+
License-File: sdm/models/timesfm3/LICENSE
|
|
28
|
+
License-File: sdm/models/timestral/LICENSE
|
|
25
29
|
License-File: sdm/models/kumo/relational/NOTICE
|
|
30
|
+
License-File: sdm/models/kumo/timeseries/anomaly/NOTICE
|
|
26
31
|
License-File: third_party/contributor-covenant/LICENSE
|
|
32
|
+
License-File: third_party/dpm-solver/LICENSE
|
|
27
33
|
License-File: third_party/pytorch/LICENSE
|
|
28
34
|
License-File: third_party/tabpfn-extensions/LICENSE
|
|
29
35
|
Requires-Dist: torch>=2.7
|
|
@@ -32,6 +38,8 @@ Requires-Dist: pyarrow
|
|
|
32
38
|
Requires-Dist: huggingface_hub
|
|
33
39
|
Requires-Dist: safetensors
|
|
34
40
|
Requires-Dist: typing_extensions
|
|
41
|
+
Provides-Extra: prior
|
|
42
|
+
Requires-Dist: scikit-learn; extra == "prior"
|
|
35
43
|
Dynamic: license-file
|
|
36
44
|
|
|
37
45
|
<p align="center">
|
|
@@ -53,9 +61,10 @@ Dynamic: license-file
|
|
|
53
61
|
|
|
54
62
|
**A GPU-native library of foundation models, tensor subclasses, and data processors for structured data.**
|
|
55
63
|
|
|
56
|
-
- **Models:** Reference implementations of structured data foundation models, including tabular models ([`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2), [`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular), *etc*),
|
|
64
|
+
- **Models:** Reference implementations of structured data foundation models, including tabular models ([`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2), [`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular), *etc*), relational models ([`KumoRelational`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoRelational)), and time series models ([`Chronos2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.Chronos2), [`KumoForecast`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoForecast), [`TimesFM3`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TimesFM3), [`KumoAnomaly`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoAnomaly), *etc*), built on a unified interface with room for future model families.
|
|
57
65
|
- **Tensor semantics:** PyTorch-compatible tensor types for numerical, categorical, datetime, text, and relational data.
|
|
58
66
|
- **Data processing:** Composable, extensible, and GPU-accelerated preprocessing and postprocessing for structured data workflows.
|
|
67
|
+
- **Synthetic data generation:** Distributions over synthetic training datasets, including SCM priors for `KumoTabular`.
|
|
59
68
|
|
|
60
69
|
## Installation
|
|
61
70
|
|
|
@@ -66,6 +75,12 @@ Install from PyPI:
|
|
|
66
75
|
pip install structured-data-models
|
|
67
76
|
```
|
|
68
77
|
|
|
78
|
+
Alternatively, install from `main` via:
|
|
79
|
+
|
|
80
|
+
```bash
|
|
81
|
+
pip install git+https://github.com/NVIDIA/structured-data-models.git
|
|
82
|
+
```
|
|
83
|
+
|
|
69
84
|
> [!NOTE]
|
|
70
85
|
> For CUDA workloads, we highly recommend installing [`cudf`](https://docs.rapids.ai/install) as an additional dependency to keep dataframe-style operations on GPU and avoid unnecessary data movement.
|
|
71
86
|
|
|
@@ -74,13 +89,23 @@ pip install structured-data-models
|
|
|
74
89
|
**Tabular Foundation Models:**
|
|
75
90
|
|
|
76
91
|
- **[`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2)** from Qu *et al.*: [TabICLv2: A Better, Faster, Scalable, and Open Tabular Foundation Model](https://arxiv.org/abs/2602.11139) (ICML '26)
|
|
77
|
-
- **[`
|
|
78
|
-
- **[`
|
|
92
|
+
- **[`TabFM`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabFM)** from Kong *et al.*: [TabFM: A Zero-Shot Foundation Model for Tabular Data](https://arxiv.org/abs/2609.37959) (CoRR '26)
|
|
93
|
+
- **[`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular)** from Qu *et al.*: [NVIDIA Kumo Tabular: Pushing the Accuracy–Efficiency Frontier for Tabular Prediction](https://arxiv.org/abs/2611.23001) (CoRR '26)
|
|
79
94
|
|
|
80
95
|
**Relational Foundation Models:**
|
|
81
96
|
|
|
82
97
|
- **[`KumoRelational`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoRelational)** from Hudovernik *et al.*: [KumoRFM-2: Scaling Foundation Models for Relational Learning](https://arxiv.org/abs/2604.12596) (CoRR '26)
|
|
83
98
|
|
|
99
|
+
**Time Series Forecasting Foundation Models:**
|
|
100
|
+
|
|
101
|
+
- **[`Chronos2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.Chronos2)** from Ansari *et al.*: [Chronos-2: From Univariate to Universal Forecasting](https://arxiv.org/abs/2510.15821) (CoRR '25)
|
|
102
|
+
- **[`KumoForecast`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoForecast)** from Gautam *et al.*: [Memory-Augmented Forecasting: Scalability and Generalization Across Temporal Domains](https://ieeexplore.ieee.org/abstract/document/11402442) (IEEE BigData '25)
|
|
103
|
+
- **[`TimesFM3`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TimesFM3)** from Jain *et al.*: [TimesFM-3: A Zero-Shot Foundation Model for Multivariate Forecasting](https://research.google/blog/timesfm-3-a-zero-shot-foundation-model-for-multivariate-forecasting) ('26)
|
|
104
|
+
|
|
105
|
+
**Time Series Anomaly Detection Foundation Models:**
|
|
106
|
+
|
|
107
|
+
- **[`KumoAnomaly`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoAnomaly)** from Gautam *et al.*: [NV-Tesseract-AD: Curriculum-Enhanced Conditional Diffusion for Multivariate Time-Series Anomaly Detection](https://arxiv.org/abs/2610.12754) (CoRR '26)
|
|
108
|
+
|
|
84
109
|
## Quick Tour
|
|
85
110
|
|
|
86
111
|
```python
|
|
@@ -98,7 +123,7 @@ table = sdm.TableTensor.from_pandas(
|
|
|
98
123
|
)
|
|
99
124
|
|
|
100
125
|
# Access to a variety of pre-trained structured data models:
|
|
101
|
-
model = sdm.models.
|
|
126
|
+
model = sdm.models.KumoTabular(device="cuda")
|
|
102
127
|
|
|
103
128
|
# Default in-context learning forward pass:
|
|
104
129
|
model(
|
|
@@ -137,6 +162,10 @@ Third-party software and separately distributed model assets are documented in [
|
|
|
137
162
|
|
|
138
163
|
- [**`sdm/models/tabiclv2/`**](sdm/models/tabiclv2/) contains code derived from [`TabICLv2`](https://github.com/soda-inria/tabicl) under the BSD 3-Clause License; its terms are distributed in [`sdm/models/tabiclv2/LICENSE`](sdm/models/tabiclv2/LICENSE).
|
|
139
164
|
- [**`sdm/models/tabfm/`**](sdm/models/tabfm/) contains code derived from [`TabFM`](https://github.com/google-research/tabfm) under the Apache License 2.0; its terms are distributed in [`sdm/models/tabfm/LICENSE`](sdm/models/tabfm/LICENSE).
|
|
165
|
+
- [**`sdm/models/chronos2/`**](sdm/models/chronos2/) contains code derived from [`Chronos2`](https://github.com/amazon-science/chronos-forecasting) under the Apache License 2.0; its terms are distributed in [`sdm/models/chronos2/LICENSE`](sdm/models/chronos2/LICENSE).
|
|
166
|
+
- [**`sdm/models/timesfm3/`**](sdm/models/timesfm3/) contains code derived from [`TimesFM`](https://github.com/google-research/timesfm) under the Apache License 2.0; its terms are distributed in [`sdm/models/timesfm3/LICENSE`](sdm/models/timesfm3/LICENSE).
|
|
167
|
+
- [**`sdm/models/timestral/`**](sdm/models/timestral/) contains code derived from [`Timestral`](https://github.com/mistralai/timestral) under the Apache License 2.0; its terms are distributed in [`sdm/models/timestral/LICENSE`](sdm/models/timestral/LICENSE).
|
|
140
168
|
- [**`sdm/models/kumo/relational/NOTICE`**](sdm/models/kumo/relational/NOTICE) documents [`KumoRelational`](sdm/models/kumo/relational/)'s reuse of [`TabICLv2`](sdm/models/tabiclv2/)-derived components.
|
|
169
|
+
- [**`sdm/models/kumo/timeseries/anomaly/NOTICE`**](sdm/models/kumo/timeseries/anomaly/NOTICE) documents [`KumoAnomaly`](sdm/models/kumo/timeseries/anomaly)'s attribution to [DPM-Solver](https://github.com/LuChengTHU/dpm-solver).
|
|
141
170
|
- [**`third_party/pytorch/`**](third_party/pytorch/) contains the BSD 3-Clause License for material adapted from [PyTorch](https://github.com/pytorch/pytorch) in [`CONTRIBUTING.md`](CONTRIBUTING.md); its terms are distributed in [`third_party/pytorch/LICENSE`](third_party/pytorch/LICENSE).
|
|
142
171
|
- [**`third_party/contributor-covenant/`**](third_party/contributor-covenant/) contains the MIT License for [Contributor Covenant version 1.4](https://www.contributor-covenant.org/version/1/4/code-of-conduct); its terms are distributed in [`third_party/contributor-covenant/LICENSE`](third_party/contributor-covenant/LICENSE).
|
|
@@ -17,9 +17,10 @@
|
|
|
17
17
|
|
|
18
18
|
**A GPU-native library of foundation models, tensor subclasses, and data processors for structured data.**
|
|
19
19
|
|
|
20
|
-
- **Models:** Reference implementations of structured data foundation models, including tabular models ([`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2), [`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular), *etc*),
|
|
20
|
+
- **Models:** Reference implementations of structured data foundation models, including tabular models ([`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2), [`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular), *etc*), relational models ([`KumoRelational`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoRelational)), and time series models ([`Chronos2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.Chronos2), [`KumoForecast`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoForecast), [`TimesFM3`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TimesFM3), [`KumoAnomaly`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoAnomaly), *etc*), built on a unified interface with room for future model families.
|
|
21
21
|
- **Tensor semantics:** PyTorch-compatible tensor types for numerical, categorical, datetime, text, and relational data.
|
|
22
22
|
- **Data processing:** Composable, extensible, and GPU-accelerated preprocessing and postprocessing for structured data workflows.
|
|
23
|
+
- **Synthetic data generation:** Distributions over synthetic training datasets, including SCM priors for `KumoTabular`.
|
|
23
24
|
|
|
24
25
|
## Installation
|
|
25
26
|
|
|
@@ -30,6 +31,12 @@ Install from PyPI:
|
|
|
30
31
|
pip install structured-data-models
|
|
31
32
|
```
|
|
32
33
|
|
|
34
|
+
Alternatively, install from `main` via:
|
|
35
|
+
|
|
36
|
+
```bash
|
|
37
|
+
pip install git+https://github.com/NVIDIA/structured-data-models.git
|
|
38
|
+
```
|
|
39
|
+
|
|
33
40
|
> [!NOTE]
|
|
34
41
|
> For CUDA workloads, we highly recommend installing [`cudf`](https://docs.rapids.ai/install) as an additional dependency to keep dataframe-style operations on GPU and avoid unnecessary data movement.
|
|
35
42
|
|
|
@@ -38,13 +45,23 @@ pip install structured-data-models
|
|
|
38
45
|
**Tabular Foundation Models:**
|
|
39
46
|
|
|
40
47
|
- **[`TabICLv2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabICLv2)** from Qu *et al.*: [TabICLv2: A Better, Faster, Scalable, and Open Tabular Foundation Model](https://arxiv.org/abs/2602.11139) (ICML '26)
|
|
41
|
-
- **[`
|
|
42
|
-
- **[`
|
|
48
|
+
- **[`TabFM`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TabFM)** from Kong *et al.*: [TabFM: A Zero-Shot Foundation Model for Tabular Data](https://arxiv.org/abs/2609.37959) (CoRR '26)
|
|
49
|
+
- **[`KumoTabular`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoTabular)** from Qu *et al.*: [NVIDIA Kumo Tabular: Pushing the Accuracy–Efficiency Frontier for Tabular Prediction](https://arxiv.org/abs/2611.23001) (CoRR '26)
|
|
43
50
|
|
|
44
51
|
**Relational Foundation Models:**
|
|
45
52
|
|
|
46
53
|
- **[`KumoRelational`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoRelational)** from Hudovernik *et al.*: [KumoRFM-2: Scaling Foundation Models for Relational Learning](https://arxiv.org/abs/2604.12596) (CoRR '26)
|
|
47
54
|
|
|
55
|
+
**Time Series Forecasting Foundation Models:**
|
|
56
|
+
|
|
57
|
+
- **[`Chronos2`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.Chronos2)** from Ansari *et al.*: [Chronos-2: From Univariate to Universal Forecasting](https://arxiv.org/abs/2510.15821) (CoRR '25)
|
|
58
|
+
- **[`KumoForecast`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoForecast)** from Gautam *et al.*: [Memory-Augmented Forecasting: Scalability and Generalization Across Temporal Domains](https://ieeexplore.ieee.org/abstract/document/11402442) (IEEE BigData '25)
|
|
59
|
+
- **[`TimesFM3`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.TimesFM3)** from Jain *et al.*: [TimesFM-3: A Zero-Shot Foundation Model for Multivariate Forecasting](https://research.google/blog/timesfm-3-a-zero-shot-foundation-model-for-multivariate-forecasting) ('26)
|
|
60
|
+
|
|
61
|
+
**Time Series Anomaly Detection Foundation Models:**
|
|
62
|
+
|
|
63
|
+
- **[`KumoAnomaly`](https://nvidia.github.io/structured-data-models/api/generated/sdm.models.KumoAnomaly)** from Gautam *et al.*: [NV-Tesseract-AD: Curriculum-Enhanced Conditional Diffusion for Multivariate Time-Series Anomaly Detection](https://arxiv.org/abs/2610.12754) (CoRR '26)
|
|
64
|
+
|
|
48
65
|
## Quick Tour
|
|
49
66
|
|
|
50
67
|
```python
|
|
@@ -62,7 +79,7 @@ table = sdm.TableTensor.from_pandas(
|
|
|
62
79
|
)
|
|
63
80
|
|
|
64
81
|
# Access to a variety of pre-trained structured data models:
|
|
65
|
-
model = sdm.models.
|
|
82
|
+
model = sdm.models.KumoTabular(device="cuda")
|
|
66
83
|
|
|
67
84
|
# Default in-context learning forward pass:
|
|
68
85
|
model(
|
|
@@ -101,6 +118,10 @@ Third-party software and separately distributed model assets are documented in [
|
|
|
101
118
|
|
|
102
119
|
- [**`sdm/models/tabiclv2/`**](sdm/models/tabiclv2/) contains code derived from [`TabICLv2`](https://github.com/soda-inria/tabicl) under the BSD 3-Clause License; its terms are distributed in [`sdm/models/tabiclv2/LICENSE`](sdm/models/tabiclv2/LICENSE).
|
|
103
120
|
- [**`sdm/models/tabfm/`**](sdm/models/tabfm/) contains code derived from [`TabFM`](https://github.com/google-research/tabfm) under the Apache License 2.0; its terms are distributed in [`sdm/models/tabfm/LICENSE`](sdm/models/tabfm/LICENSE).
|
|
121
|
+
- [**`sdm/models/chronos2/`**](sdm/models/chronos2/) contains code derived from [`Chronos2`](https://github.com/amazon-science/chronos-forecasting) under the Apache License 2.0; its terms are distributed in [`sdm/models/chronos2/LICENSE`](sdm/models/chronos2/LICENSE).
|
|
122
|
+
- [**`sdm/models/timesfm3/`**](sdm/models/timesfm3/) contains code derived from [`TimesFM`](https://github.com/google-research/timesfm) under the Apache License 2.0; its terms are distributed in [`sdm/models/timesfm3/LICENSE`](sdm/models/timesfm3/LICENSE).
|
|
123
|
+
- [**`sdm/models/timestral/`**](sdm/models/timestral/) contains code derived from [`Timestral`](https://github.com/mistralai/timestral) under the Apache License 2.0; its terms are distributed in [`sdm/models/timestral/LICENSE`](sdm/models/timestral/LICENSE).
|
|
104
124
|
- [**`sdm/models/kumo/relational/NOTICE`**](sdm/models/kumo/relational/NOTICE) documents [`KumoRelational`](sdm/models/kumo/relational/)'s reuse of [`TabICLv2`](sdm/models/tabiclv2/)-derived components.
|
|
125
|
+
- [**`sdm/models/kumo/timeseries/anomaly/NOTICE`**](sdm/models/kumo/timeseries/anomaly/NOTICE) documents [`KumoAnomaly`](sdm/models/kumo/timeseries/anomaly)'s attribution to [DPM-Solver](https://github.com/LuChengTHU/dpm-solver).
|
|
105
126
|
- [**`third_party/pytorch/`**](third_party/pytorch/) contains the BSD 3-Clause License for material adapted from [PyTorch](https://github.com/pytorch/pytorch) in [`CONTRIBUTING.md`](CONTRIBUTING.md); its terms are distributed in [`third_party/pytorch/LICENSE`](third_party/pytorch/LICENSE).
|
|
106
127
|
- [**`third_party/contributor-covenant/`**](third_party/contributor-covenant/) contains the MIT License for [Contributor Covenant version 1.4](https://www.contributor-covenant.org/version/1/4/code-of-conduct); its terms are distributed in [`third_party/contributor-covenant/LICENSE`](third_party/contributor-covenant/LICENSE).
|