Tensor Creation

Creating Tensors from Data

from_array - Create from existing data

Create tensors from C arrays.

CC++

#include <tensr/tensr_array.h>

/* 1D array */
float data[] = {1, 2, 3, 4, 5};
size_t shape[] = {5};
Tensor* t = tensr_from_array(shape, 1, TENSR_FLOAT32, TENSR_CPU, data);

/* 2D array (matrix) */
float matrix[] = {1, 2, 3, 4, 5, 6};
size_t shape2d[] = {2, 3};
Tensor* mat = tensr_from_array(shape2d, 2, TENSR_FLOAT32, TENSR_CPU, matrix);

/* C++ uses factory methods */
auto t = tensr::Tensor::zeros({2, 3});

Initialization Functions

zeros - All zeros

CC++

size_t shape[] = {3, 3};
Tensor* t = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);
tensr_print(t);
tensr_free(t);

auto t = tensr::Tensor::zeros({3, 3});
t.print();

ones - All ones

CC++

size_t shape[] = {2, 4};
Tensor* t = tensr_ones(shape, 2, TENSR_FLOAT32, TENSR_CPU);

auto t = tensr::Tensor::ones({2, 4});

full - Fill with value

CC++

size_t shape[] = {3, 3};
Tensor* t = tensr_full(shape, 2, 5.0, TENSR_FLOAT32, TENSR_CPU);

auto t = tensr::Tensor::full({3, 3}, 5.0);

Range Functions

arange - Evenly spaced values

CC++

/* Create [0, 2, 4, 6, 8] */
Tensor* t = tensr_arange(0.0, 10.0, 2.0, TENSR_FLOAT32, TENSR_CPU);

auto t = tensr::Tensor::arange(0.0, 10.0, 2.0);

linspace - Linearly spaced values

CC++

/* Create 5 values from 0 to 1 */
Tensor* t = tensr_linspace(0.0, 1.0, 5, TENSR_FLOAT32, TENSR_CPU);
/* Result: [0.0, 0.25, 0.5, 0.75, 1.0] */

auto t = tensr::Tensor::linspace(0.0, 1.0, 5);

Special Matrices

eye - Identity matrix

CC++

/* Create 3x3 identity matrix */
Tensor* t = tensr_eye(3, TENSR_FLOAT32, TENSR_CPU);
/* Result:
   [[1, 0, 0],
    [0, 1, 0],
    [0, 0, 1]] */

auto t = tensr::Tensor::eye(3);

Complete Example

#include <tensr/tensr.h>
#include <tensr/tensr_array.h>

int main() {
    /* Create from data */
    float data[] = {1, 2, 3, 4, 5, 6};
    size_t shape[] = {2, 3};
    Tensor* a = tensr_from_array(shape, 2, TENSR_FLOAT32, TENSR_CPU, data);

    /* Create initialized tensors */
    Tensor* zeros = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);
    Tensor* ones = tensr_ones(shape, 2, TENSR_FLOAT32, TENSR_CPU);

    /* Create ranges */
    Tensor* range = tensr_arange(0.0, 10.0, 1.0, TENSR_FLOAT32, TENSR_CPU);

    /* Create identity */
    Tensor* eye = tensr_eye(3, TENSR_FLOAT32, TENSR_CPU);

    /* Print and cleanup */
    tensr_print(a);
    tensr_free(a);
    tensr_free(zeros);
    tensr_free(ones);
    tensr_free(range);
    tensr_free(eye);

    return 0;
}

Memory Management

Always Free Tensors

Always call tensr_free() when done with a tensor to prevent memory leaks.

Tensor* t = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);
/* Use tensor */
tensr_free(t);  /* Don't forget! */

Data Types

Available data types:

• TENSR_FLOAT32 - 32-bit float (default)
• TENSR_FLOAT64 - 64-bit double
• TENSR_INT32 - 32-bit integer
• TENSR_INT64 - 64-bit integer
• TENSR_UINT8 - 8-bit unsigned integer
• TENSR_BOOL - Boolean

Device Selection

Create tensors on different devices:

/* CPU */
Tensor* t_cpu = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);

/* CUDA GPU */
Tensor* t_gpu = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CUDA);