Expand description
Backend implementations of tensor operations.
Modules§
- numeric
- Numeric utility functions for jit backends
Functions§
- empty_
qtensor - Create an empty quantized tensor.
- empty_
qtensor_ optimized - Create an empty quantized tensor.
- into_
data - into_
data_ sync - Read data from a
RudaTensorsynchronously - permute
- Permute a tensor’s dimensions
- permute_
nchw_ to_ nhwc - Permute a tensor’s dimensions from NCHW to NHWC, or the N-dimensional equivalent
- permute_
nchw_ to_ nhwc_ shape - Permute a shape’s dimensions from NCHW to NHWC, or the N-dimensional equivalent
- permute_
nhwc_ to_ nchw - Permute a tensor’s dimensions from NHWC to NCHW, or the N-dimensional equivalent
- permute_
nhwc_ to_ nchw_ shape - Permute a shape’s dimensions from NHWC to NCHW, or the N-dimensional equivalent
- q_
reshape - Reshape a quantized tensor without recalibrating its values.
- reshape
- Reshape a jit tensor to a new shape
- unfold
- Unfold windows along a dimension.