stelliferous

Search uses release 0.0.2.0 · 2,086 definitions

Releases (2)

def packed
def machine
def machine_support
def supported
def dense_call_validWhat a dense call needs of its three buffers: lengths that fit the shape and the capacities that hold them.
def checked
def status
def replanned
def existing_layout
def execute_existing_layout
def band_machine
def band_supported
def band_call_validWhat a call on bands needs: bands that cover the images of the geometry, and parameters whose lengths fit the shape and their capacities.
def band_inspected
def band_checked
def band_status
def input_layout
def input_selected
def conv2d_selectedParameter representations select one public path before any element work.
def conv2d_bias_layout
def conv2d

type Axis
def include_axis
def axes
type LevelA level with no remaining count never moves; it stands for an absent axis.
def absent
def restoreOuter axes: reset axes are kept reversed and restored after the carry stops.
def carry
def advance_outer
def reset
def step
def Depth0
def Depth1
def Levels
def advance_levels
type Cursor
def moved
def advance
def split_level
def outer
def split
def prepare
def read_value
def read
type StrideOne run of the innermost axis reads address, address + stride, ...
def stride_value
def read_stride
def targetThe address a scatter stores at, with the cursor of the next element.
def target_stride
type CursorPair

def Tensor
def Problem
def StorageStorage exchanged with foreign I/O; see from_storage and into_storage.
def full
def zeros
def ones
def from_list
def from_storage
def load_rawRaw little-endian FP32 files: the values of the shape in row-major order.
def save_raw
def load_npyNumPy .npy files of FP32 values; the file carries the shape.
def save_npy
def WeightsModels: tensors by name, loaded once.
def Model
def Model.pure
def Model.bind
def take
def keep
def load_weightsfolder/names.txt lists one name per line; folder/<name>.npy holds its tensor.
def copy_weights
def run
def run_keepingThe result with what remains of the weights, kept tensors included.
def save_weights
def with_workersExecution context: operations may use count workers, like a tensor's device.
def workers
def shape
def check
def describe
def expect
def to_list
def item
def to_list_orReads with a fallback for a failed tensor, for callers that do not inspect the problem.
def item_or
def into_storage
def expect_storage
def then
def then_shape
def copy
def addBinary elementwise operations reuse the left tensor's storage.
def sub
def mul
def divBase division: a zero divisor gives an infinity or NaN, never a failure.
def maximum
def minimum
def add_scalar
def sub_scalar
def mul_scalar
def div_scalar
def less
def less_equal