Utility Functions

Utility Functions#

This module contains simple utility functions to facilitate conversions between NumPy arrays and PyTorch tensors, and shape validation.

Functions#

objectrl.utils.utils.totorch(x: ndarray, dtype: dtype = torch.float32, device: str | device = 'cuda') → Tensor[source]#

Converts a NumPy array or other compatible object to a PyTorch tensor.

Parameters:

  • x (array-like) – Input data to convert.

  • dtype (torch.dtype, optional) – Desired data type of the output tensor. Default is torch.float32.

  • device (str or torch.device, optional) – Device to store the tensor on. Default is “cuda”.

Returns:

A tensor containing the same data as x, on the specified device and with the specified dtype.

Return type:

torch.Tensor

objectrl.utils.utils.tonumpy(x: Tensor) → ndarray[source]#

Converts a PyTorch tensor to a NumPy array. Automatically moves the tensor to CPU and detaches it from the computation graph.

Parameters:

x (torch.Tensor) – Input tensor.

Returns:

NumPy array with the same data as the input tensor.

Return type:

numpy.ndarray

objectrl.utils.utils.toint(x: Tensor) → int[source]#

Converts a PyTorch tensor to an integer.

Parameters:

x (torch.Tensor) – Input tensor.

Returns:

Integer value of the input tensor.

Return type:

int

objectrl.utils.utils.dim_check(tensor1: Tensor, tensor2: Tensor) → None[source]#

Asserts that two tensors have the same shape. Useful for debugging shape mismatches in model inputs/outputs.

Parameters:

  • tensor1 (torch.Tensor) – First tensor.

  • tensor2 (torch.Tensor) – Second tensor.

Raises:

AssertionError – If the shapes of the two tensors do not match.

Function Details#

totorch(x, dtype=torch.float32, device=”cuda”) -> torch.Tensor

Converts a NumPy array or compatible object into a PyTorch tensor with specified dtype and device.

tonumpy(x) -> numpy.ndarray

Converts a PyTorch tensor to a NumPy ndarray, moving it to CPU and detaching from the computation graph.

toint(x) -> int

Converts a PyTorch tensor containing a single value to a Python integer.

dim_check(tensor1, tensor2) -> None

Asserts that two tensors have the same shape. Raises AssertionError if they differ.

Example Usage#

import numpy as np
import torch

arr = np.array([1, 2, 3])
tensor = totorch(arr, dtype=torch.float32, device="cpu")
back_to_np = tonumpy(tensor)
integer_value = toint(torch.tensor(5))
dim_check(tensor, torch.ones_like(tensor))  # no error
Contents