TTEncoder#

class TTEncoder#

Encoder for tile tensors. Can encode and/or encrypt a tensor into a tile tensor and back again.

assert_equals(self: pyhelayers.TTEncoder, c: pyhelayers.CTileTensor, title: str, expected_vals: numpy.ndarray[numpy.float64], eps: float = 1e-06, percent: bool = False) → float#

decode_double(self: pyhelayers.TTEncoder, src: pyhelayers.PTileTensor) → numpy.ndarray[numpy.float64]#

decrypt(self: pyhelayers.TTEncoder, res: pyhelayers.PTileTensor, src: pyhelayers.CTileTensor) → None#

Decrypt.

Parameters:: res (PTileTensor) – Resulting PTileTensor.
Paramy src:: Input CTileTensor.

decrypt_decode_double(self: pyhelayers.TTEncoder, src: pyhelayers.CTileTensor) → numpy.ndarray[numpy.float64]#

encode(self: pyhelayers.TTEncoder, shape: pyhelayers.TTShape, vals: numpy.ndarray[numpy.float64], chain_index: int = -1) → pyhelayers.PTileTensor#

encode_encrypt(self: pyhelayers.TTEncoder, shape: pyhelayers.TTShape, vals: numpy.ndarray[numpy.float64], chain_index: int = -1) → pyhelayers.CTileTensor#

encode_encrypt_generic_packing(self: pyhelayers.TTEncoder, vals: numpy.ndarray[numpy.float64], generic_packing_config: pyhelayers.GenericPackingConfig = <pyhelayers.GenericPackingConfig object at 0x7f6023140f30>) → pyhelayers.CTileTensor#

encrypt(self: pyhelayers.TTEncoder, res: pyhelayers.CTileTensor, src: pyhelayers.PTileTensor) → None#

Encrypt.

Parameters:: res (CTileTensor) – Resulting CTileTensor.
Paramy src:: Input PTileTensor.

get_encoder(self: pyhelayers.TTEncoder) → pyhelayers.Encoder#

Returns a reference to the basic encoder used by the TTEncoder.

Return type:: bool

class TTEncoder#

Encoder for tile tensors.

Can encode and/or encrypt a tensor into a tile tensor and back again.

Public Functions

TTEncoder(const HeContext &he, LazyMode lazyMode = NONE)#

Constructs a ready to use object.

Parameters:

he – [in] the underlying context.
lazyMode – Type of lazy encoding to use (default: NONE).

~TTEncoder()#

TTEncoder &operator=(const TTEncoder &src) = delete#: Assignment deleted. Construct new objects when needed.

inline Encoder &getEncoder()#: Returns a reference to the basic encoder used by the TTEncoder.

void encode(PTileTensor &res, const TTShape &shape, const DoubleTensor &vals, int chainIndex = -1) const#

Encodes an input double tensor to a PTileTensor.

Parameters:

res – Output PTileTensor
shape – Tile tensor shape
vals – Input tensor
chainIndex – Chain index used for encoding (when applicable)

void encodeOneTile(PTile &res, const TTShape &shape, const DoubleTensor &vals, int valsId, int flatIndex, const std::vector<int> &mask = {}, int chainIndex = -1, std::shared_ptr<const TTTileMapping> ttTileMapping = nullptr) const#

Encode one tile of a tile tensor into a PTile object.

Parameters:

res – Output PTile
shape – shape of tile tensor
vals – Original tensor of tile tensor
valsId – Unique ID of original tensor of tile tensor
flatIndex – Flat index of tile
mask – Optional mask to multiply tile with. Default: none (empty).
chainIndex – Optional chain-index for encoding (where applicable). Default: highest.
ttTileMapping – Optional pre-processed tile mapping. If exists encoding works faster.

void encodeOneTile(PTile &res, const TTShape &shape, const ComplexTensor &vals, int valsId, int flatIndex, const std::vector<int> &mask = {}, int chainIndex = -1, std::shared_ptr<const TTTileMapping> ttTileMapping = nullptr) const#

Encode one tile of a complex packed tile tensor into a PTile object.

Parameters:

res – Output PTile
shape – shape of tile tensor
vals – Original (complex packed) tensor of tile tensor
valsId – Unique ID of original tensor of tile tensor
flatIndex – Flat index of tile
mask – Optional mask to multiply tile with. Default: none (empty).
chainIndex – Optional chain-index for encoding (where applicable). Default: highest.
ttTileMapping – Optional pre-processed tile mapping. If exists encoding works faster.

void encode(PTileTensor &res, const TTShape &shape, double val, int chainIndex = -1) const#

Encodes a PTileTensor filled with a single value.

Parameters:

res – Output PTileTensor
shape – Tile tensor shape
val – The value to encode
chainIndex – Chain index used for encoding (when applicable)

DoubleTensor decodeDouble(const PTileTensor &src) const#

Returns the decoding result of input PTileTensor into a tensor of doubles.

Parameters:: src – Input PTileTensor

void encrypt(CTileTensor &res, const PTileTensor &src) const#

Encrypts a PTileTensor to a CTileTensor.

Parameters:

res – Output CTileTensor
src – Input PTileTensor

void decrypt(PTileTensor &res, const CTileTensor &src) const#

Decrypts a CTileTensor to a PTileTensor.

Parameters:

res – Output PTileTensor
src – Input CTileTensor caller responsibility to call src.wakeup()

void encodeEncrypt(CTileTensor &res, const TTShape &shape, const DoubleTensor &vals, int chainIndex = -1) const#

Encrypts an input double tensor to a CTileTensor.

Parameters:

res – Output CTileTensor
shape – Tile tensor shape
vals – Input tensor
chainIndex – Chain index used for encoding (when applicable)

void encodeEncrypt(CTileTensor &res, const TTShape &shape, double val, int chainIndex = -1) const#

Encrypts a CTileTensor filled with a single value.

Parameters:

res – Output CTileTensor
shape – Tile tensor shape
val – The value to encode
chainIndex – Chain index used for encoding (when applicable)

void encodeEncryptGenericPacking(CTileTensor &res, const DoubleTensor &vals, const GenericPackingConfig &config = GenericPackingConfig()) const#

Encrypts an input double tensor to a CTileTensor using generic packing.

Parameters:

res – Output CTileTensor
vals – Input tensor
config – The generic packing configuration (optional).

DoubleTensor decryptDecodeDouble(const CTileTensor &src) const#

Returns the result of decryption and decoding of input CTileTensor.

Parameters:: src – Input CTileTensor

double assertEquals(const CTileTensor &c, const std::string &title, const DoubleTensor &expectedVals, double eps, bool percent = false) const#

Asserts that the content of given CTileTensor is equal to given double tensor.

Parameters:

c – CTileTensor input to compare
title – Text to include in error message
expectedVals – tensor input to compare
eps – Tolerance
percent – Relative diff indictor

double testMse(const CTileTensor &c, const DoubleTensor &expectedVals, const std::string &title = "", double eps = -1) const#

Computes the MSE (Mean Square Error) with expected values.

If title!=””, prints results to cout. If eps>=0, asserts MSE<=eps

Parameters:

c – CTileTensor input to compare
expectedVals – tensor input to compare
title – If not empty, will be output to cout along with MSE
eps – If non-negative, an exception will be thrown if MSE>eps

inline const HeContext &getHeContext()#: Returns the context.

Public Static Functions

static void extractTileValues(std::vector<double> &res, const DoubleTensor &tensor, const TTShape &shape, int tileFlatIndex, std::shared_ptr<const TTTileMapping> ttTileMapping = nullptr)#

Extracts and populates the values of a single tile from an original tensor of values.

Parameters:

res – Output vector of values.
tensor – Original tensor.
shape – Shape of the tile tensor.
tileFlatIndex – Flat index of tile in tile tensor.
ttTileMapping – Optional pre-processed tile mapping. If exists extraction works faster.

static void extractTileValues(std::vector<std::complex<double>> &res, const ComplexTensor &tensor, const TTShape &shape, int tileFlatIndex, std::shared_ptr<const TTTileMapping> ttTileMapping = nullptr)#

Extracts and populates the complex values of a single tile from an original tensor of values.

Parameters:

res – Output vector of complex values.
tensor – Original (complex packed) tensor.
shape – Shape of the tile tensor.
tileFlatIndex – Flat index of tile in tile tensor.
ttTileMapping – Optional pre-processed tile mapping. If exists extraction works faster.