TensorFlow

Native TensorFlow bindings for Node.JS

Install

npm install tensorflow

Features

• Fully exposes the TensorFlow C API
• Fast native bindings (no FFI)
• Composable, intuitive wrapper API
• No TensorFlow install/build required

Globals

tf.Graph()

tf.Session()

tf.Operation()

tf.Tensor(tf.DataType, [tf.TensorShape])

Instance methods

Tensor.batchToSpace(block_shape, crops)

Tensor.cast(type)

Tensor.broadcast(tensor)

Tensor.isNumeric()

Returns true if the tensor doesn't contain NaN or Infinity values.

Tensor.concat(tensors[], axis) // ops::Concat, ops::ParallelConcat

Tensor.copy([name])

Tensor.copyHost([name])

Tensor.debugID()

Tensor.countNAN()

Tensor.summary()

Tensor.depthToSpace(block_size)

Tensor.dequantize(minRange, maxRange, quantize)

quantize = true || { signed: // If the quantization is signed or unsigned. bits: // The bitwidth of the quantization. minRange: maxRange: }

Tensor.diagonal([diagonal]) // ops::Diag and ops::MatrixBandPart

opts = { upper: true || number of superdiagonals to keep lower: true || number of subdiagonals to keep } if given a diagonal, this method sets the diagonal values

Tensor.leven(tensor, { normalized: false })

Tensor.expand(axis)

Tensor.quantize(min, max, [type], [gradient])

Tensor.fill(value, [dimensions])

Tensor.gather(indices)

Tensor.clone() // ops::Identity, ops::ZerosLike

Tensor.freeze() // ops::ImmutableConst

Tensor.invert() // ops::InvertPermutation

Tensor.pad(paddings, mode) // ops::Pad, ops::MirrorPad

Tensor.quantize(minRange, maxRange, type)

Tensor.quantizedInstanceNorm(min, max, opts)

opts = { output_range_given: If True, given_y_min and given_y_min and given_y_max are used as the output range. Otherwise, the implementation computes the output range. given_y_min: Output in y_min if output_range_given is True. given_y_max: Output in y_max if output_range_given is True. variance_epsilon: A small float number to avoid dividing by 0. min_separation: Minimum value of y_max - y_min }

Tensor.reverse(axis)

Tensor.reverseSequence(sequenceLengths, sequenceDimension, reverseDimension)

Tensor.scatter(indices, values, [shape])

Tensor.diff(tensor)

Tensor.shape([shape])

Return the shape of the tensor.

If shape is provided, sets the shape of the tensor.

Tensor.size()

Tensor.slice(offset, length, strides)

Tensor.spaceToBatch(block_shape, paddings)

Tensor.squeeze([dimensionIndices])

Static methods

Tensor.diagonal(diagonal)

Tensor.oneHot(indices, value, [opts])

Tensor.quantize(tensors[], dimensions, min[], max[])

Tensor.shapes(tensors[])

Tensor.split(axis, count, byValue (ops::SplitV))

Tensor.stack(tensors[], [axis])

Tensor.tile(multiples)

Tensor.transpose(permutation)

Tensor.unique(counts == false)

Tensor.unstack([axis])

Tensor.where()

opts = { axis: 1, depth: 1, zeros: 0.1 }

tf.Image()

Static methods

Image.extractPatches([images], windowSize, centerStride, rates, padding)

Global methods

tf.IsGoogleCudaEnabled()

tf.LogAllRegisteredKernels()