elasticai.creator.nn.fixed_point.lstm.layer#

Module Contents#

Classes#

LSTMNetwork

FixedPointLSTMWithHardActivations

Use only together with the above LSTMNetwork. There is no single hw design corresponding to this sw layer. Instead, the design of the LSTMNetwork handles most of the tasks, that are performed by FixedPointLSTMWithHardActivations

API#

class elasticai.creator.nn.fixed_point.lstm.layer.LSTMNetwork(layers: list[torch.nn.Module])[source]#

Bases: elasticai.creator.nn.design_creator_module.DesignCreatorModule

create_design(name: str) → elasticai.creator.nn.fixed_point.lstm.design.lstm.LSTMNetworkDesign[source]#
create_testbench(test_bench_name, uut: elasticai.creator.vhdl.design.design.Design) → elasticai.creator.nn.fixed_point.lstm.design.testbench.LSTMTestBench[source]#
forward(x)[source]#
class elasticai.creator.nn.fixed_point.lstm.layer.FixedPointLSTMWithHardActivations(total_bits: int, frac_bits: int, input_size: int, hidden_size: int, bias: bool)[source]#

Bases: elasticai.creator.nn.design_creator_module.DesignCreatorModule, elasticai.creator.base_modules.lstm.LSTM

Use only together with the above LSTMNetwork. There is no single hw design corresponding to this sw layer. Instead, the design of the LSTMNetwork handles most of the tasks, that are performed by FixedPointLSTMWithHardActivations

Initialization

Initialize internal Module state, shared by both nn.Module and ScriptModule.

property fixed_point_config: elasticai.creator.arithmetic.FxpArithmetic#
create_design(name: str = 'lstm_cell') → elasticai.creator.vhdl.design.design.Design[source]#