skainet-backend-cpu/sk.ainet.java/SKaiNET

SKaiNET

Primary Java entry point for SKaiNET tensor operations.

Provides static factory methods for creating execution contexts, tensors, and performing common operations without requiring Kotlin-specific knowledge.

Example usage from Java:

var ctx = SKaiNET.context();
var a = SKaiNET.tensor(ctx, new int[]{2, 3}, DType.fp32(), new float[]{1,2,3,4,5,6});
var b = SKaiNET.ones(ctx, new int[]{2, 3}, DType.fp32());
var c = TensorJavaOps.add(a, b);

Functions

context

@JvmStatic

fun context(): ExecutionContext

Creates a new CPU execution context with default settings (eval mode).

full

@JvmStatic

fun full(ctx: ExecutionContext, shape: IntArray, dtype: DType, value: Number): Tensor<*, *>

Creates a tensor filled with a constant value.

ones

@JvmStatic

@JvmOverloads

fun ones(ctx: ExecutionContext, shape: IntArray, dtype: DType = FP32): Tensor<*, *>

Creates a tensor filled with ones.

randn

@JvmStatic

@JvmOverloads

fun randn(ctx: ExecutionContext, shape: IntArray, dtype: DType = FP32): Tensor<*, *>

Creates a tensor filled with random values from a normal distribution.

tensor

@JvmStatic

fun tensor(ctx: ExecutionContext, shape: IntArray, dtype: DType, data: FloatArray): Tensor<*, *>

Creates a tensor from a float array with the given shape and dtype.

tensorFromInts

@JvmStatic

fun tensorFromInts(ctx: ExecutionContext, shape: IntArray, dtype: DType, data: IntArray): Tensor<*, *>

Creates a tensor from an int array with the given shape and dtype.

zeros

@JvmStatic

@JvmOverloads

fun zeros(ctx: ExecutionContext, shape: IntArray, dtype: DType = FP32): Tensor<*, *>

Creates a tensor filled with zeros.