FIX bidirectional layers name

Author: Enrico Lupi

hls4ml/backends/vivado/passes/recurrent_templates.py

@@ -1,10 +1,6 @@
 from hls4ml.backends.backend import get_backend
 from hls4ml.backends.template import FunctionCallTemplate, LayerConfigTemplate
-<<<<<<< HEAD
-from hls4ml.model.layers import GRU, LSTM, BLSTM, BGRU, TimeDistributed
-=======
-from hls4ml.model.layers import BGRU, BLSTM, GRU, LSTM
->>>>>>> d2d3b452 (ADD fixes)
+from hls4ml.model.layers import GRU, LSTM, BidirectionalLSTM, BidirectionalGRU, TimeDistributed
 
 # recurrent multiplication template
 
@@ -247,7 +243,7 @@ def format(self, node):
 
 class BidirectionalRecurrentConfigTemplate(LayerConfigTemplate):
     def __init__(self):
-        super().__init__((BLSTM, BGRU))
+        super().__init__((BidirectionalLSTM, BidirectionalGRU))
         self.template = bidir_recr_config_template
         self.act_template = activ_config_template
         self.recr_act_template = recr_activ_config_template
@@ -275,11 +271,11 @@ def format(self, node):
         params['static'] = 'true' if node.attributes['static'] else 'false'
         params['pytorch'] = 'true' if node.get_attr('pytorch', False) else 'false'
         params['recr_type'] = node.class_name.lower()
-        params['RECR_TYPE'] = node.class_name[1:]
+        params['RECR_TYPE'] = node.class_name[13:]
 
-        if node.class_name == 'BLSTM':
+        if node.class_name == 'BidirectionalLSTM':
             n_recr_mult = 4
-        else:  # BGRU
+        else:  # BidirectionalGRU
             n_recr_mult = 3
 
         recr_config = self.template.format(**params)
@@ -458,7 +454,7 @@ def format(self, node):
 
 class BidirectionalRecurrentFunctionTemplate(FunctionCallTemplate):
     def __init__(self):
-        super().__init__((BLSTM, BGRU), include_header=recr_include_list)
+        super().__init__((BidirectionalLSTM, BidirectionalGRU), include_header=recr_include_list)
 
     def format(self, node):
         params = self._default_function_params(node)

hls4ml/backends/vivado/vivado_backend.py

@@ -12,6 +12,8 @@
 from hls4ml.model.layers import (
     GRU,
     LSTM,
+    BidirectionalGRU,
+    BidirectionalLSTM,
     Conv1D,
     Conv2D,
     Dense,
@@ -46,11 +48,7 @@ def __init__(self):
 
     def _register_layer_attributes(self):
         # Add RNN-specific attributes, recurrent_reuse_factor and static implementation
-        rnn_layers = [
-            SimpleRNN,
-            LSTM,
-            GRU,
-        ]
+        rnn_layers = [SimpleRNN, LSTM, GRU, BidirectionalLSTM, BidirectionalGRU]
 
         for layer in rnn_layers:
             attrs = self.attribute_map.get(layer, [])

hls4ml/model/layers.py

@@ -1402,8 +1402,8 @@ def initialize(self):
             self.add_weights_variable(name='recurrent_bias', var_name='br{index}', data=recurrent_bias)
 
 
-class BLSTM(LSTM):
-    _expected_attributes = LSTM._expected_attributes + [
+class BidirectionalLSTM(LSTM):
+    _expected_attributes = [
         WeightAttribute('weight_b'),
         WeightAttribute('bias_b'),
         WeightAttribute('recurrent_weight_b'),
@@ -1510,8 +1510,8 @@ def initialize(self):
         self.add_output_variable(shape, dims)
 
 
-class BGRU(GRU):
-    _expected_attributes = GRU._expected_attributes + [
+class BidirectionalGRU(GRU):
+    _expected_attributes = [
         WeightAttribute('weight_b'),
         WeightAttribute('bias_b'),
         WeightAttribute('recurrent_weight_b'),
@@ -1826,8 +1826,8 @@ def initialize(self):
     'SimpleRNN': SimpleRNN,
     'LSTM': LSTM,
     'GRU': GRU,
-    'BLSTM': BLSTM,
-    'BGRU': BGRU,
+    'BidirectionalLSTM': BidirectionalLSTM,
+    'BidirectionalGRU': BidirectionalGRU,
     'QSimpleRNN': SimpleRNN,
     'QLSTM': LSTM,
     'QGRU': GRU,