Python class

TensorValue

Library for the graph TensorValue class.

TensorValue

class max.graph.TensorValue(value: Value | Value | BufferValue | TensorValue | Shape | Dim | int | float | integer | floating | ndarray)

Bases: Value

Represents a value semantic tensor within a Graph. It provides various methods and properties to manipulate and query tensor attributes such as shape, data type (dtype), device placement (device), and more.

For example:

import numpy as np
from max.dtype import DType
from max.graph import Graph, ops

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("tensor_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    # Perform a simple operation: transpose the tensor
    transposed_tensor = tensor.T  # Output: Tensor representing [[1, 3], [2, 4]]

import numpy as np
from max.dtype import DType
from max.graph import Graph, ops

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("tensor_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    # Perform a simple operation: transpose the tensor
    transposed_tensor = tensor.T  # Output: Tensor representing [[1, 3], [2, 4]]

T

property T*: TensorValue*

Returns the transposed tensor. T is the shorthand notation for transposing. For more information, see transpose().

• Returns:

  A new TensorValue with swapped dimensions.

broadcast_to()

broadcast_to(shape: Iterable[int | str | Dim | integer]) → TensorValue

Broadcasts the tensor to a new shape.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("broadcast_to_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    broadcasted_tensor = tensor.broadcast_to((3, 2, 2)) # Output: Tensor with shape (3, 2, 2)

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("broadcast_to_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    broadcasted_tensor = tensor.broadcast_to((3, 2, 2)) # Output: Tensor with shape (3, 2, 2)

• Parameters:

  shape – An iterable of integers or symbolic dimensions.

• Returns:

  A new TensorValue with the broadcasted shape.

cast()

cast(dtype: DType) → TensorValue

Casts a symbolic tensor to a different data type.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("cast_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    casted_tensor = tensor.cast(DType.int32) # Output: Tensor representing [[1, 2], [3, 4]] with dtype=int32

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("cast_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    casted_tensor = tensor.cast(DType.int32) # Output: Tensor representing [[1, 2], [3, 4]] with dtype=int32

• Parameters:

  dtype – The target data type (e.g., DType.int32, DType.float64).

• Returns:

  A new TensorValue with the casted data type.

device

property device*: DeviceRef | None*

Returns the device of the TensorValue.

dtype

property dtype*: DType*

Returns the tensor data type.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("dtype_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(f"Data type: {tensor.dtype}")  # Output: Data type: DType.float32

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("dtype_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(f"Data type: {tensor.dtype}")  # Output: Data type: DType.float32

flatten()

flatten(start_dim: int = 0, end_dim: int = -1) → TensorValue

Flattens the specified dims of a symbolic tensor.

The number and order of the elements in the tensor is unchanged. All dimensions from start_dim to end_dim (inclusive) are merged into a single output dim.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("flatten_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    # Flatten the tensor
    flattened_tensor = tensor.flatten() # Output: Tensor representing [1, 2, 3, 4]

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("flatten_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    # Flatten the tensor
    flattened_tensor = tensor.flatten() # Output: Tensor representing [1, 2, 3, 4]

• Parameters:

  • start_dim – The starting dimension to flatten. Defaults to 1.
  • end_dim – The ending dimension to flatten. Defaults to -1.

• Returns:

  A new TensorValue with the broadcasted shape.

from_dim()

static from_dim(dim: int | str | Dim | integer) → TensorValue

Creates a new tensor based on provided MLIR dimension type.

• Parameters:

  dim – The dimension value.

• Returns:

  A new TensorValue.

from_shape()

static from_shape(shape: Iterable[int | str | Dim | integer]) → TensorValue

Creates a new tensor with the specified shape.

• Parameters:

  shape – An iterable of integers or symbolic dimensions.

• Returns:

  A new TensorValue.

permute()

permute(dims: list[int]) → TensorValue

Permutes the tensor’s dimensions based on provided indices.

• Parameters:

  dims – A list of integers specifying the new order of dimensions.

• Returns:

  A new TensorValue with permuted dimensions.

print()

print(label: str = 'debug_tensor')

Prints detailed information about the tensor.

• Parameters:

  label – A string label for the printed output. Defaults debug_tensor.

rank

property rank*: int*

Returns the rank (number of dims) of the buffer.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("rank_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(f"Rank (number of dimensions): {tensor.rank}")  # Output: Rank: 2

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("rank_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(f"Rank (number of dimensions): {tensor.rank}")  # Output: Rank: 2

rebind()

rebind(shape: Iterable[int | str | Dim | integer], message: str = '') → TensorValue

Rebinds the tensor to a new shape with error handling.

• Parameters:

  • shape – The new shape as an iterable of integers or symbolic dimensions.
  • message – (optional) A message for logging or debugging.

• Returns:

  A new TensorValue with the updated shape.

reshape()

reshape(shape: Iterable[int | str | Dim | integer]) → TensorValue

Creates a new tensor with the same data but reshaped.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("reshape_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    reshaped_tensor = tensor.reshape((1, 4)) # Output: Tensor representing [[1, 2, 3, 4]]

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("reshape_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    reshaped_tensor = tensor.reshape((1, 4)) # Output: Tensor representing [[1, 2, 3, 4]]

• Parameters:

  shape – The new shape as an iterable of integers or symbolic dimensions.

• Returns:

  A new TensorValue with the reshaped dimensions.

shape

property shape*: Shape*

Returns the shape of the TensorValue.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("shape_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    # Access tensor properties
    print(f"Shape: {tensor.shape}")  # Output: Shape: (2, 2)

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

# Create a Graph context to work with tensors
with Graph("shape_demo") as graph:
    # Create a constant tensor from the matrix
    tensor = ops.constant(matrix, dtype=DType.float32)

    # Access tensor properties
    print(f"Shape: {tensor.shape}")  # Output: Shape: (2, 2)

to()

to(device: DeviceRef) → TensorValue

Transfers the tensor to a specified device without mutation.

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

with Graph("to_example") as graph:
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(tensor.device)

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)

with Graph("to_example") as graph:
    tensor = ops.constant(matrix, dtype=DType.float32)
    print(tensor.device)

• Parameters:

  device – A DeviceRef object specifying the target device.

• Returns:

  A new TensorValue on the specified device.

transpose()

transpose(dim_1: int, dim_2: int) → TensorValue

Swaps two dimensions of the tensor.

import numpy as np
from max.dtype import DType
from max.graph import Graph, ops

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)
with Graph("transpose_demo") as graph:
    tensor = ops.constant(matrix, dtype=DType.float32)
    transposed_tensor = tensor.transpose(dim_1=0, dim_2=1)
    print(transposed_tensor)

import numpy as np
from max.dtype import DType
from max.graph import Graph, ops

matrix = np.array([[1, 2], [3, 4]], dtype=np.float32)
with Graph("transpose_demo") as graph:
    tensor = ops.constant(matrix, dtype=DType.float32)
    transposed_tensor = tensor.transpose(dim_1=0, dim_2=1)
    print(transposed_tensor)

• Parameters:

  • dim_1 – The first dimension to swap.
  • dim_2 – The second dimension to swap.

• Returns:

  A new TensorValue with swapped dimensions.

type

property type*: TensorType*

Returns the type of the TensorValue as a TensorType.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit