Azure Foundry

Looking for usage documentation? Check out Classification, Regression and KV cache.

`foundry.TabPFNClassifier`

TabPFN classifier backed by an Azure AI Foundry endpoint.

foundry.TabPFNClassifier(
    *args: Any,
    task: str = "classification",
    **kwargs: Any,
)

Parameters

Parameter	Type	Default	Description
`*args`	`Any`	—	—
`task`	`str`	`"classification"`	—
`**kwargs`	`Any`	—	See the accepted constructor options in this table.
`endpoint_url`	`str`	Required	Accepted through `**kwargs`.
`api_key`	`str`	Required	Accepted through `**kwargs`.
`n_estimators`	`int`	`8`	Accepted through `**kwargs`.
`softmax_temperature`	`float`	`0.9`	Accepted through `**kwargs`.
`balance_probabilities`	`bool`	`False`	Accepted through `**kwargs`.
`average_before_softmax`	`bool`	`False`	Accepted through `**kwargs`.
`ignore_pretraining_limits`	`bool`	`True`	Accepted through `**kwargs`.
`inference_precision`	`Literal[“autocast”, “auto”]`	`"auto"`	Accepted through `**kwargs`.
`random_state`	`Optional[int]`	`0`	Accepted through `**kwargs`.
`inference_config`	`Optional[Dict[str, Any]]` `InferenceConfig` options	`None`	Accepted through `**kwargs`.
`paper_version`	`bool`	`False`	Accepted through `**kwargs`.
`thinking_mode`	`bool`	`False`	Accepted through `**kwargs`.
`thinking_effort`	`Optional[ThinkingEffort]`	`None`	Accepted through `**kwargs`.
`thinking_timeout_s`	`Optional[float]`	`None`	Accepted through `**kwargs`.
`thinking_metric`	`Optional[str]`	`None`	Accepted through `**kwargs`.
`use_kv_cache`	`bool`	`False`	Accepted through `**kwargs`.
`fit_mode`	`Optional[FitModeLiteral]`	`None`	Accepted through `**kwargs`.
`timeout_s`	`float`	`300.0`	Accepted through `**kwargs`.

Example

from tabpfn_client.foundry import TabPFNClassifier
clf = TabPFNClassifier(
    endpoint_url="https://<your-endpoint>.<region>.inference.ml.azure.com/predict",
    api_key="<your-foundry-bearer-token>",
)
clf.fit(X_train, y_train)
clf.predict(X_test)
clf.predict_proba(X_test)

`foundry.TabPFNClassifier.fit`

View source

Store the training data; the endpoint fits on the next predict* call.

foundry.TabPFNClassifier.fit(
    X: Any,
    y: Any,
) -> "_FoundryBase"

Parameters

Parameter	Type	Default	Description
`X`	`Any`	Required	Training features.
`y`	`Any`	Required	Training targets.

Returns

Type	Description
`_FoundryBase`	The estimator itself.

Raises ValueError If X and y have different numbers of samples.

`foundry.TabPFNClassifier.predict`

View source

Predict class labels for X.

foundry.TabPFNClassifier.predict(
    X: Any,
) -> np.ndarray

Parameters

Parameter	Type	Default	Description
`X`	`Any`	Required	—

Returns

Type	Description
`np.ndarray`	—

`foundry.TabPFNClassifier.predict_proba`

View source

Predict class probabilities for X.

foundry.TabPFNClassifier.predict_proba(
    X: Any,
) -> np.ndarray

Parameters

Parameter	Type	Default	Description
`X`	`Any`	Required	—

Returns

Type	Description
`np.ndarray`	—

`foundry.TabPFNRegressor`

View source

TabPFN regressor backed by an Azure AI Foundry endpoint.

foundry.TabPFNRegressor(
    *args: Any,
    task: str = "regression",
    **kwargs: Any,
)

Parameters

Parameter	Type	Default	Description
`*args`	`Any`	—	—
`task`	`str`	`"regression"`	—
`**kwargs`	`Any`	—	See the accepted constructor options in this table.
`endpoint_url`	`str`	Required	Accepted through `**kwargs`.
`api_key`	`str`	Required	Accepted through `**kwargs`.
`n_estimators`	`int`	`8`	Accepted through `**kwargs`.
`softmax_temperature`	`float`	`0.9`	Accepted through `**kwargs`.
`balance_probabilities`	`bool`	`False`	Accepted through `**kwargs`.
`average_before_softmax`	`bool`	`False`	Accepted through `**kwargs`.
`ignore_pretraining_limits`	`bool`	`True`	Accepted through `**kwargs`.
`inference_precision`	`Literal[“autocast”, “auto”]`	`"auto"`	Accepted through `**kwargs`.
`random_state`	`Optional[int]`	`0`	Accepted through `**kwargs`.
`inference_config`	`Optional[Dict[str, Any]]` `InferenceConfig` options	`None`	Accepted through `**kwargs`.
`paper_version`	`bool`	`False`	Accepted through `**kwargs`.
`thinking_mode`	`bool`	`False`	Accepted through `**kwargs`.
`thinking_effort`	`Optional[ThinkingEffort]`	`None`	Accepted through `**kwargs`.
`thinking_timeout_s`	`Optional[float]`	`None`	Accepted through `**kwargs`.
`thinking_metric`	`Optional[str]`	`None`	Accepted through `**kwargs`.
`use_kv_cache`	`bool`	`False`	Accepted through `**kwargs`.
`fit_mode`	`Optional[FitModeLiteral]`	`None`	Accepted through `**kwargs`.
`timeout_s`	`float`	`300.0`	Accepted through `**kwargs`.

Example

from tabpfn_client.foundry import TabPFNRegressor
reg = TabPFNRegressor(
    endpoint_url="https://<your-endpoint>.<region>.inference.ml.azure.com/predict",
    api_key="<your-foundry-bearer-token>",
)
reg.fit(X_train, y_train)
reg.predict(X_test)
reg.predict(X_test, output_type="quantiles", quantiles=[0.1, 0.5, 0.9])

`foundry.TabPFNRegressor.fit`

View source

Store the training data; the endpoint fits on the next predict* call.

foundry.TabPFNRegressor.fit(
    X: Any,
    y: Any,
) -> "_FoundryBase"

Parameters

Parameter	Type	Default	Description
`X`	`Any`	Required	Training features.
`y`	`Any`	Required	Training targets.

Returns

Type	Description
`_FoundryBase`	The estimator itself.

Raises ValueError If X and y have different numbers of samples.

`foundry.TabPFNRegressor.predict`

View source

Predict regression targets for X.

foundry.TabPFNRegressor.predict(
    X: Any,
    output_type: Literal["mean", "median", "mode", "quantiles", "full", "main"] = "mean",
    quantiles: Optional[list] = None,
) -> Union[np.ndarray, list, Dict[str, np.ndarray]]

Parameters

Parameter	Type	Default	Description
`X`	`Any`	Required	Test features.
`output_type`	`Literal[“mean”, “median”, “mode”, “quantiles”, “full”, “main”]`	`"mean"`	Which prediction to return.
`quantiles`	`Optional[list]`	`None`	Quantile levels to compute when `output_type` is “quantiles”.

Returns

Type	Description
`Union[np.ndarray, list, Dict[str, np.ndarray]]`	An array for point predictions; a list with one array per quantile for “quantiles”; a dict of arrays for dict-shaped outputs such as “full” and “main”.

Scoring

The classifier and regressor inherit the corresponding scikit-learn score method. These adapters do not inherit the main hosted client’s save_model, load_model, or model-version selection methods.

Reference

tabpfn

tabpfn-client

tabpfn-extensions

REST API

`foundry.TabPFNClassifier`

`foundry.TabPFNClassifier.fit`

`foundry.TabPFNClassifier.predict`

`foundry.TabPFNClassifier.predict_proba`

`foundry.TabPFNRegressor`

`foundry.TabPFNRegressor.fit`

`foundry.TabPFNRegressor.predict`

Scoring

​foundry.TabPFNClassifier

​foundry.TabPFNClassifier.fit

​foundry.TabPFNClassifier.predict

​foundry.TabPFNClassifier.predict_proba

​foundry.TabPFNRegressor

​foundry.TabPFNRegressor.fit

​foundry.TabPFNRegressor.predict

​Scoring

`foundry.TabPFNClassifier`

`foundry.TabPFNClassifier.fit`

`foundry.TabPFNClassifier.predict`

`foundry.TabPFNClassifier.predict_proba`

`foundry.TabPFNRegressor`

`foundry.TabPFNRegressor.fit`

`foundry.TabPFNRegressor.predict`

Scoring