Layer attributes

Input

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Constant

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • value: ndarray

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Activation

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

ParametrizedActivation

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

  • n_in: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • param_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • param_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

PReLU

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

  • n_in: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • param_t: NamedType

Weight attributes

  • param: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • param_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Softmax

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • n_outer: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • n_inner: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • implementation: list [latency,stable,argmax,legacy] (Default: stable)

    • Choice of implementation of softmax function. “latency” provides good latency at the expense of extra resources. It performs well with a small number of classes. “stable” may require extra clock cycles but has better accuracy. “legacy” is the older implementation which has bad accuracy, but is fast and has low resource use. It is superseded by the “latency” implementation for most applications. “argmax” is a special implementation that can be used if only the output with the highest probability is important. Using this implementation will save resources and clock cycles.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • skip: bool (Default: False)

    • If enabled, skips the softmax node and returns the raw outputs.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • exp_table_t: NamedType (Default: fixed<18,8,RND,SAT,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • inv_table_t: NamedType (Default: fixed<18,8,RND,SAT,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • inv_inp_t: NamedType (Default: fixed<18,8,RND,SAT,0>)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • accum_t: NamedType (Default: fixed<18,8,RND,SAT,0>)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

IFNeuron

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_out: int

  • window_size: int (Default: 0)

  • threshold: float

  • threshold_mode: str (Default: scalar)

  • reset_mechanism: list [subtract,zero] (Default: subtract)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • threshold_t: NamedType

  • membrane_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • threshold_t: NamedType

  • membrane_t: NamedType

LIFNeuron

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_out: int

  • window_size: int (Default: 0)

  • threshold: float

  • threshold_mode: str (Default: scalar)

  • beta: float

  • beta_mode: str (Default: scalar)

  • reset_mechanism: list [subtract,zero] (Default: subtract)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • beta_t: NamedType

  • threshold_t: NamedType

  • membrane_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • beta_t: NamedType

  • threshold_t: NamedType

  • membrane_t: NamedType

SNNReadout

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_classes: int

  • window_size: int (Default: 1)

  • class_threshold: int (Default: 1)

  • beta: float (Default: 1.0)

  • state_reset_policy: list [fixed_window,tlast,host_pulse,never] (Default: fixed_window)

  • decision_rule: list [argmax_spike_count,first_to_threshold,threshold_then_argmax,binary_logit,argmax_membrane] (Default: argmax_membrane)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • membrane_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • output_mode: list [spike,membrane] (Default: spike)

  • membrane_t: NamedType

TernaryTanh

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

HardActivation

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • activation: str

  • slope: float (Default: 0.2)

  • shift: float (Default: 0.5)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • slope_t: NamedType

  • shift_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • slope_t: NamedType

  • shift_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Reshape

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • target_shape: Sequence

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Dense

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_out: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • strategy: list [latency,resource] (Default: latency)

    • Available in: Libero

Conv

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Conv1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

Conv2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

Conv2DBatchnorm

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

SeparableConv1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • depth_multiplier: int (Default: 1)

  • filt_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • depthwise_t: NamedType

  • pointwise_t: NamedType

  • bias_t: NamedType

Weight attributes

  • depthwise: WeightVariable

  • pointwise: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • depthwise_t: NamedType

  • pointwise_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • depthwise_accum_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • pointwise_accum_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • depthwise_result_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • depthwise_reuse_factor: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • pointwise_reuse_factor: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

  • dw_output_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • Available in: Catapult

DepthwiseConv1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

  • in_width: int

  • out_width: int

  • n_chan: int

  • depth_multiplier: int (Default: 1)

  • n_filt: int

  • filt_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

SeparableConv2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • depth_multiplier: int (Default: 1)

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • depthwise_t: NamedType

  • pointwise_t: NamedType

  • bias_t: NamedType

Weight attributes

  • depthwise: WeightVariable

  • pointwise: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • depthwise_t: NamedType

  • pointwise_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • depthwise_accum_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • pointwise_accum_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • depthwise_result_t: NamedType

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • depthwise_reuse_factor: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • pointwise_reuse_factor: int (Default: 1)

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

  • dw_output_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • Available in: Catapult

DepthwiseConv2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • depth_multiplier: int (Default: 1)

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

Conv1DTranspose

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_width: int

  • stride_width: int

  • dilation_width: int (Default: 1)

  • crop_left: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • slot_groups: tuple (Default: ())

    • Set by the backend. The output is computed in blocks of stride positions, one block for each input position. For each kernel element, in the order of the rows of the weight matrix, the position in the block that it adds to.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • slot_shifts: tuple (Default: ())

    • Set by the backend. For each kernel element, in the order of the rows of the weight matrix, how many blocks before the current block its input is, as (height, width).

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • n_buf: int (Default: 0)

    • Set by the backend. The number of inputs multiplied for each block of outputs, which is the number of kernel elements times the number of input channels.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • first_block_width: int (Default: 0)

    • Set by the backend. The index of the first block of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • n_blocks_width: int (Default: 0)

    • Set by the backend. The number of blocks of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

Conv2DTranspose

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • dilation_height: int (Default: 1)

  • dilation_width: int (Default: 1)

  • crop_top: int

  • crop_left: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • slot_groups: tuple (Default: ())

    • Set by the backend. The output is computed in blocks of stride positions, one block for each input position. For each kernel element, in the order of the rows of the weight matrix, the position in the block that it adds to.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • slot_shifts: tuple (Default: ())

    • Set by the backend. For each kernel element, in the order of the rows of the weight matrix, how many blocks before the current block its input is, as (height, width).

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • n_buf: int (Default: 0)

    • Set by the backend. The number of inputs multiplied for each block of outputs, which is the number of kernel elements times the number of input channels.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • first_block_height: int (Default: 0)

    • Set by the backend. The index of the first block of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • n_blocks_height: int (Default: 0)

    • Set by the backend. The number of blocks of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • first_block_width: int (Default: 0)

    • Set by the backend. The index of the first block of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • n_blocks_width: int (Default: 0)

    • Set by the backend. The number of blocks of outputs along this axis that cropping does not remove entirely.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

BatchNormalization

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_filt: int (Default: -1)

  • use_gamma: bool (Default: True)

  • use_beta: bool (Default: True)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Weight attributes

  • scale: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Pooling1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_out: int

  • n_filt: int

  • pool_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

  • count_pad: bool (Default: False)

  • pool_op: list [Max,Average]

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

Pooling2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_filt: int

  • pool_height: int

  • pool_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

  • count_pad: bool (Default: False)

  • pool_op: list [Max,Average]

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

GlobalPooling1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_filt: int

  • keepdims: bool (Default: False)

  • pool_op: list [Max,Average]

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

GlobalPooling2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • n_filt: int

  • keepdims: bool (Default: False)

  • pool_op: list [Max,Average]

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

ZeroPadding1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

ZeroPadding2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Cropping1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • crop_left: int

  • crop_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Cropping2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • crop_top: int

  • crop_bottom: int

  • crop_left: int

  • crop_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Merge

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

MatMul

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Dot

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, Vivado, VivadoAccelerator, VivadoAccelerator, Vitis, Vitis, VitisUnified, VitisUnified, Catapult, Catapult, SymbolicExpression, SymbolicExpression, Altera, Altera, Coyote, Coyote, Libero, Libero, XLS, XLS

Concatenate

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

Resize

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • align_corners: bool (Default: False)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • algorithm: list [nearest,bilinear] (Default: nearest)

Transpose

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Embedding

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_out: int

  • vocab_size: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • embeddings_t: NamedType

Weight attributes

  • embeddings: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • embeddings_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

SimpleRNN

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_out: int

  • activation: str

  • return_sequences: bool (Default: False)

  • return_state: bool (Default: False)

  • pass_initial_states: bool (Default: False)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

  • recurrent_weight: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • direction: list [forward,backward] (Default: forward)

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • static: bool (Default: True)

    • If set to True, will reuse the same recurrent block for computation, resulting in lower resource usage at the expense of serialized computation and higher latency/II. Only applies to io_parallel, as io_stream always reuses the same block. On Vivado and Vitis, the Resource strategy also always reuses the same block.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • recurrent_reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Catapult

LSTM

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_out: int

  • activation: str

  • recurrent_activation: str

  • return_sequences: bool (Default: False)

  • return_state: bool (Default: False)

  • pass_initial_states: bool (Default: False)

  • direction: list [forward,backward] (Default: forward)

  • time_major: bool (Default: False)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

  • recurrent_bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

  • recurrent_weight: WeightVariable

  • recurrent_bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

  • recurrent_bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • static: bool (Default: True)

    • If set to True, will reuse the same recurrent block for computation, resulting in lower resource usage at the expense of serialized computation and higher latency/II. Only applies to io_parallel, as io_stream always reuses the same block. On Vivado and Vitis, the Resource strategy also always reuses the same block.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • recurrent_reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Catapult

GRU

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_out: int

  • activation: str

  • recurrent_activation: str

  • return_sequences: bool (Default: False)

  • return_state: bool (Default: False)

  • pass_initial_states: bool (Default: False)

  • direction: list [forward,backward] (Default: forward)

  • time_major: bool (Default: False)

  • apply_reset_gate: list [before,after] (Default: after)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

  • recurrent_bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

  • recurrent_weight: WeightVariable

  • recurrent_bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • recurrent_weight_t: NamedType

  • recurrent_bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • static: bool (Default: True)

    • If set to True, will reuse the same recurrent block for computation, resulting in lower resource usage at the expense of serialized computation and higher latency/II. Only applies to io_parallel, as io_stream always reuses the same block. On Vivado and Vitis, the Resource strategy also always reuses the same block.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • recurrent_reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Catapult

Bidirectional

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_out: int

  • return_sequences: bool (Default: False)

  • return_state: bool (Default: False)

  • pass_initial_states: bool (Default: False)

  • time_major: bool (Default: False)

  • forward_activation: str

  • forward_recurrent_activation: str

  • backward_activation: str

  • backward_recurrent_activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • forward_weight_t: NamedType

  • forward_bias_t: NamedType

  • forward_recurrent_weight_t: NamedType

  • forward_recurrent_bias_t: NamedType

  • backward_weight_t: NamedType

  • backward_bias_t: NamedType

  • backward_recurrent_weight_t: NamedType

  • backward_recurrent_bias_t: NamedType

Weight attributes

  • forward_weight: WeightVariable

  • forward_bias: WeightVariable

  • forward_recurrent_weight: WeightVariable

  • forward_recurrent_bias: WeightVariable

  • backward_weight: WeightVariable

  • backward_bias: WeightVariable

  • backward_recurrent_weight: WeightVariable

  • backward_recurrent_bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • forward_weight_t: NamedType

  • forward_bias_t: NamedType

  • forward_recurrent_weight_t: NamedType

  • forward_recurrent_bias_t: NamedType

  • backward_weight_t: NamedType

  • backward_bias_t: NamedType

  • backward_recurrent_weight_t: NamedType

  • backward_recurrent_bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • static: bool (Default: True)

    • If set to True, will reuse the same recurrent block for computation, resulting in lower resource usage at the expense of serialized computation and higher latency/II. Only applies to io_parallel, as io_stream always reuses the same block. On Vivado and Vitis, the Resource strategy also always reuses the same block.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • table_size: int (Default: 1024)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

TimeDistributed

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • wrapped_layer: None

  • n_time_steps: int

  • output_shape: list

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • time_step_loop_parallelism: list [Off,Unroll,Pipeline] (Default: Off)

    • Controls the amount and type of parallelism in the loop over time steps. If set to “off”, no parallelism will be used. If set to “unroll”, the loop will be unrolled. This may result in excessive resource use and cannot be used in “io_stream” mode. If set to “pipeline”, the loop will be pipelined.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

GarNet

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, Vivado, VivadoAccelerator, VivadoAccelerator, Vitis, Vitis, VitisUnified, VitisUnified, Catapult, Catapult, SymbolicExpression, SymbolicExpression, Altera, Altera, Coyote, Coyote, Libero, Libero, XLS, XLS

GarNetStack

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, Vivado, VivadoAccelerator, VivadoAccelerator, Vitis, Vitis, VitisUnified, VitisUnified, Catapult, Catapult, SymbolicExpression, SymbolicExpression, Altera, Altera, Coyote, Coyote, Libero, Libero, XLS, XLS

Quant

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • narrow: bool

  • rounding_mode: str

  • signed: bool

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

BipolarQuant

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

ApplyAlpha

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_filt: int (Default: -1)

  • use_gamma: bool (Default: True)

  • use_beta: bool (Default: True)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Weight attributes

  • scale: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

BatchNormOnnx

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

LayerGroup

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • layer_list: list

  • input_layers: list

  • output_layers: list

  • data_reader: object

  • output_shape: list

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

SymbolicExpression

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • expression: list

  • n_symbols: int

  • lut_functions: list (Default: [])

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

LayerNormalization

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • seq_len: int

  • axis: int (Default: 2)

  • epsilon_power_of_10: int (Default: 3)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Weight attributes

  • scale: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • scale_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • table_range_power2: int (Default: 0)

    • The negative power of 2 that represents the range of the lookup table, e.g. a value of 1 would represent a range of 0.5.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • table_size: int (Default: 4096)

    • The size of the lookup table used to approximate the function.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • table_t: NamedType (Default: ufixed<8,5,RND_CONV,SAT,0>)

    • The datatype (precision) used for the values of the lookup table.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

  • accum_t: NamedType (Default: fixed<14,4,RND_CONV,SAT,0>)

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Coyote

EinsumDense

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • equation: str

  • inp_shape: tuple

  • out_shape: tuple

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • accum_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • accum_t: NamedType

Einsum

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • equation: str

  • inp0_shape: tuple

  • inp1_shape: tuple

  • out_shape: tuple

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

BiasAdd

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Backend-specific attributes

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

DACombinational

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

SparseInputReduce

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • n_chan: int

  • n_sparse: int

  • threshold: float

  • hash_bits: int (Default: 10)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

SparseConv2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_sparse: int

  • n_chan: int

  • n_filt: int

  • kernel_size: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • accum_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

  • accum_t: NamedType

SparseActivation

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_sparse: int

  • n_chan: int

  • activation: str

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

SparsePooling2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_sparse: int

  • n_chan: int

  • in_height: int

  • in_width: int

  • pool_height: int

  • pool_width: int

  • pool_op: str (Default: avg)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

SparseFlatten

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_sparse: int

  • n_chan: int

  • out_height: int

  • out_width: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

FixedPointQuantizer

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

UnaryLUT

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

Weight attributes

  • table: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • table_t: NamedType (Default: fixed<18,8,TRN,WRAP,0>)

Clone

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Repack

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

BatchNormalizationQuantizedTanh

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • n_in: int

  • n_filt: int (Default: 0)

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • accum_t: NamedType

  • reuse_factor: int (Default: 1)

Broadcast

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

PointwiseConv1D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_width: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_width: int

  • stride_width: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote

PointwiseConv2D

Base attributes

  • index: int

    • Internal node counter used for bookkeeping and variable/tensor naming.

  • in_height: int

  • in_width: int

  • out_height: int

  • out_width: int

  • n_chan: int

  • n_filt: int

  • filt_height: int

  • filt_width: int

  • stride_height: int

  • stride_width: int

  • pad_top: int

  • pad_bottom: int

  • pad_left: int

  • pad_right: int

Type attributes

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Weight attributes

  • weight: WeightVariable

  • bias: WeightVariable

Configurable attributes

  • trace: int (Default: False)

    • Enables saving of layer output (tracing) when using hls_model.predict(…) or hls_model.trace(…)

  • result_t: NamedType

    • The datatype (precision) of the output tensor.

  • weight_t: NamedType

  • bias_t: NamedType

Backend-specific attributes

  • accum_t: NamedType

    • The datatype (precision) used to store intermediate results of the computation within the layer.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • reuse_factor: int (Default: 1)

    • The number of times each multiplier is used by controlling the amount of pipelining/unrolling. A lower number results in more parallelism and lower latency at the expense of the resources used. Reuse factor = 1 corresponds to all multiplications executed in parallel, and hence, the lowest possible latency.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, SymbolicExpression, Altera, Coyote, Libero, XLS

  • parallelization_factor: int (Default: 1)

    • The number of outputs computed in parallel. Essentially the number of multiplications of input window with the convolution kernel occurring in parallel. A higher number results in more parallelism (lower latency and II) at the expense of resources used. Currently only supported in io_parallel.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Altera, Coyote

  • conv_implementation: list [LineBuffer,Encoded] (Default: LineBuffer)

    • “LineBuffer” implementation is preferred over “Encoded” for most use cases. This attribute only applies to io_stream.

    • Available in: Vivado, VivadoAccelerator, Vitis, VitisUnified, Catapult, Coyote