def _register_flows(self):
initializers = self._get_layer_initializers()
init_flow = register_flow('init_layers',
initializers,
requires=['optimize'],
backend=self.name)
quartus_types = [
'quartus:transform_types',
]
quartus_types_flow = register_flow('specific_types',
quartus_types,
requires=[init_flow],
backend=self.name)
templates = self._get_layer_templates()
template_flow = register_flow('apply_templates',
templates,
requires=[init_flow],
backend=self.name)
writer_passes = ['quartus:write_hls']
writer_flow_requirements = [
'optimize', quartus_types_flow, template_flow
]
self._writer_flow = register_flow('write',
writer_passes,
requires=writer_flow_requirements,
backend=self.name)
all_passes = get_backend_passes(self.name)
extras = [
# Ideally this should be empty
opt_pass for opt_pass in all_passes
if opt_pass not in initializers + quartus_types + templates +
writer_passes
]
if len(extras) > 0:
extras_flow = register_flow('extras',
extras,
requires=[init_flow],
backend=self.name)
else:
extras_flow = None
ip_flow_requirements = [
'optimize', init_flow, quartus_types_flow, extras_flow,
template_flow
]
ip_flow_requirements = list(filter(None, ip_flow_requirements))
self._default_flow = register_flow('ip',
None,
requires=ip_flow_requirements,
backend=self.name)
def _get_layer_initializers(self):
all_initializers = {
name: get_optimizer(name)
for name in get_backend_passes(self.name)
if isinstance(get_optimizer(name), LayerOptimizerPass)
}
# Sort through the initializers based on the base class (e.g., to apply 'Layer' optimizers before 'Dense')
sorted_initializers = sorted(all_initializers.items(),
key=lambda x: len(x[1].layer_class.mro()))
# Return only the names of the initializers
return [opt[0] for opt in sorted_initializers]
def _get_layer_templates(self):
return [
name for name in get_backend_passes(self.name)
if isinstance(get_optimizer(name), Template)
]
def _register_flows(self):
initializers = self._get_layer_initializers()
init_flow = register_flow('init_layers',
initializers,
requires=['optimize'],
backend=self.name)
streaming_passes = [
'vivado:remove_final_reshape',
'vivado:reshape_stream',
'vivado:clone_output',
'vivado:insert_zero_padding_before_conv1d',
'vivado:insert_zero_padding_before_conv2d',
'vivado:broadcast_stream',
]
streaming_flow = register_flow('streaming',
streaming_passes,
requires=[init_flow],
backend=self.name)
quantization_passes = [
'vivado:merge_batch_norm_quantized_tanh',
'vivado:quantize_dense_output',
'fuse_consecutive_batch_normalization',
]
quantization_flow = register_flow('quantization',
quantization_passes,
requires=[init_flow],
backend=self.name)
optimization_passes = [
'vivado:optimize_pointwise_conv',
]
optimization_flow = register_flow('optimize',
optimization_passes,
requires=[init_flow],
backend=self.name)
vivado_types = [
'vivado:register_bram_weights',
'vivado:transform_types',
'vivado:generate_conv_streaming_instructions',
'vivado:apply_resource_strategy',
]
vivado_types_flow = register_flow('specific_types',
vivado_types,
requires=[init_flow],
backend=self.name)
templates = self._get_layer_templates()
template_flow = register_flow('apply_templates',
templates,
requires=[init_flow],
backend=self.name)
writer_passes = ['vivado:write_hls']
writer_flow_requirements = [
'optimize', vivado_types_flow, template_flow
]
self._writer_flow = register_flow('write',
writer_passes,
requires=writer_flow_requirements,
backend=self.name)
all_passes = get_backend_passes(self.name)
extras = [
# Ideally this should be empty
opt_pass for opt_pass in all_passes
if opt_pass not in initializers + streaming_passes +
quantization_passes + optimization_passes + vivado_types +
templates + writer_passes
]
if len(extras) > 0:
extras_flow = register_flow('extras',
extras,
requires=[init_flow],
backend=self.name)
else:
extras_flow = None
ip_flow_requirements = [
'optimize', init_flow, streaming_flow, quantization_flow,
optimization_flow, vivado_types_flow, extras_flow, template_flow
]
ip_flow_requirements = list(filter(None, ip_flow_requirements))
self._default_flow = register_flow('ip',
None,
requires=ip_flow_requirements,
backend=self.name)