Construct an estimator that uses the given version of the model. In addition to selecting the model, this also configures the estimator with

Predict regression target for X. Parameters ---------- X : array-like of shape (n_samples, n_features) The input

Predict class labels for samples in X. Args: X: The input samples. Returns: The predicted class labels.

Initializes the TabPFN client and handles user authentication. This function checks for existing credentials, prompts for login/registration

Predict class probabilities for X. Args: X: The input samples. Returns: The class probabilities of the input

Attempts to perform browser-based login Returns (success: bool, token: str | None)

Fetch and cache dataset limits. The cache expires after 1 hour. Not thread-safe, but concurrent calls are benign: duplicates fetch the

Resets the client state and clears local authentication caches. Use this function if you need to log out or clear stored session data fr

Reduce a type hint to a comparable core: unwrap ``Annotated[T, ...]`` and drop ``None`` from unions, so the server's all-optional ``X | None`` fie

Prompt for login/registration. Returns True if successful, False if interrupted.

Helper method to simulate the callback response

Check if server is reachable and accepts the connection.

Check if the provided access token is valid and return True if successful.

Prompt with validation and optional example.

Simplified terms prompt for registration flow.

Register a new user with the provided credentials. Parameters ---------- email : str password : str

Send entered email to server that checks if it is valid and not already in use. Parameters ---------- email : str

Prompt text and give user infinitely many attempts to select one of the possible choices. If valid choice is selected, return choice

Coerce numpy / pandas inputs to plain Python lists for JSON.

Coerce numpy / pandas inputs to plain Python lists for JSON.

Delete all datasets uploaded by the user from the server. Returns ------- deleted_dataset_uids : list[str]

Delete the dataset with the provided UID from the server. Note that deleting a train set with lead to deleting all associated test se

Download all data uploaded by the user from the server. Returns ------- save_path : Path | None The path

Convert password requirement strings like "Length(8)" into a PasswordPolicy.

Retrieve greeting messages that are new for the user.

Let the server send an email for resetting the password.

Let the server send an email for verifying the email.

Check the integrity of the test data.

Check the integrity of the training data.

Assemble a Foundry `/predict` JSON body. When `cached_model_id` is provided, the body targets the V3 cache-hit path: `X_train` / `y_train` ar

Top-level wire fields for thinking-mode. Empty when inactive.

Computes the CRC32C checksum and returns it as a base64 encoded string.

Async Inference path: stage input through S3, invoke_endpoint_async, poll the OutputLocation S3 object. Required for payloads > 6 MB or

Thinking is server-side supported only for v3 models (or the auto sentinel, which lets the server pick — currently a v3 model).