diff --git a/src/finn/custom_op/fpgadataflow/vectorvectoractivation.py b/src/finn/custom_op/fpgadataflow/vectorvectoractivation.py
index 10ee30f89ab6313a4e5863890483e1da3f76e511..a5806748361e42fc4c24cf1db90ee898ce13ccf4 100644
--- a/src/finn/custom_op/fpgadataflow/vectorvectoractivation.py
+++ b/src/finn/custom_op/fpgadataflow/vectorvectoractivation.py
@@ -106,75 +106,95 @@ class VectorVectorActivation(HLSCustomOp):
def minimize_accumulator_width(self, model):
"""Minimize the accumulator bit width according to the weight values,
input data types, and size of dot product"""
- if not self.get_nodeattr("runtime_writeable_weights"):
- weights = model.get_initializer(self.onnx_node.input[1])
- k_h, k_w = self.get_nodeattr("Kernel")
- fm = self.get_nodeattr("Channels")
- # put weights into the shape expected by calculate_matvec_accumulator_range
- weights = weights.reshape(fm, k_h * k_w).transpose()
- if len(self.onnx_node.input) > 2:
- thresholds = model.get_initializer(self.onnx_node.input[2])
- else:
- thresholds = None
- idt = self.get_input_datatype()
- # calculate minimum and maximum values of accumulator according to the
- # weight values using the bounds derived in https://arxiv.org/abs/2301.13376
+ weights = model.get_initializer(self.onnx_node.input[1])
+ k_h, k_w = self.get_nodeattr("Kernel")
+ fm = self.get_nodeattr("Channels")
+ # put weights into the shape expected by calculate_matvec_accumulator_range
+ weights = weights.reshape(fm, k_h * k_w).transpose()
+ # since in the calculation the values of the weight matrix are used,
+ # for the bipolar case they need to be converted to bipolar
+ if self.get_nodeattr("binaryXnorMode"):
+ weights = 2 * weights - 1
+ if len(self.onnx_node.input) > 2:
+ thresholds = model.get_initializer(self.onnx_node.input[2])
+ else:
+ thresholds = None
+ idt = self.get_input_datatype()
+ # if runtime-writeable weights, then the values of the weights can
+ # change and we need to use the worst-case values from the datatypes
+ if self.get_nodeattr("runtime_writeable_weights"):
+ wdt = self.get_weight_datatype()
+ lower_worst = wdt.min() * np.ones_like(weights)
+ lower_range = calculate_matvec_accumulator_range(lower_worst, idt)
+ upper_worst = wdt.min() * np.ones_like(weights)
+ upper_range = calculate_matvec_accumulator_range(upper_worst, idt)
+ acc_min = min(min(lower_range), min(upper_range))
+ acc_max = max(max(upper_range), max(upper_range))
+ thresholds = None # range of thresholds are also runtime-writeable
+ # if not runtime-writeable weights, then we can calculate the min
+ # and max values of the accumulation range using knowledge of the
+ # weights and input data types since they are fixed
+ else:
(acc_min, acc_max) = calculate_matvec_accumulator_range(weights, idt)
- if thresholds is not None:
- threshold_tensor = self.get_hls_compatible_threshold_tensor(thresholds)
- # set threshold datatype (and accumulator datatype implicitly)
+ # if the thresholds can be used to determine range, then adjust the range
+ # according to the known values of the thresholds
+ if thresholds is not None:
+ threshold_tensor = self.get_hls_compatible_threshold_tensor(thresholds)
+ # set threshold datatype (and accumulator datatype implicitly)
+ min_threshold = thresholds.min()
+ max_threshold = thresholds.max()
+ # clip threshold values
+ clip_upper = None
+ clip_lower = None
+ if max_threshold > acc_max + 1:
+ clip_upper = acc_max + 1
+ if min_threshold < acc_min:
+ clip_lower = acc_min
+ if (clip_lower is not None) or (clip_upper is not None):
+ warnings.warn(
+ "Clipping some thresholds in %s" % self.onnx_node.name
+ )
+ thresholds = np.clip(thresholds, clip_lower, clip_upper)
+ model.set_initializer(self.onnx_node.input[2], thresholds)
+ threshold_tensor = self.get_hls_compatible_threshold_tensor(
+ thresholds
+ )
min_threshold = thresholds.min()
max_threshold = thresholds.max()
- # clip threshold values
- clip_upper = None
- clip_lower = None
- if max_threshold > acc_max + 1:
- clip_upper = acc_max + 1
- if min_threshold < acc_min:
- clip_lower = acc_min
- if (clip_lower is not None) or (clip_upper is not None):
- warnings.warn(
- "Clipping some thresholds in %s" % self.onnx_node.name
- )
- thresholds = np.clip(thresholds, clip_lower, clip_upper)
- model.set_initializer(self.onnx_node.input[2], thresholds)
- threshold_tensor = self.get_hls_compatible_threshold_tensor(
- thresholds
- )
- min_threshold = thresholds.min()
- max_threshold = thresholds.max()
- # get range required by threshold values
- tdt_min = min(acc_min, min_threshold)
- tdt_max = max(acc_max, max_threshold)
- if tdt_min < 0:
- if abs(tdt_min) > tdt_max:
- tdt = DataType.get_smallest_possible(tdt_min)
- else:
- tdt = DataType.get_smallest_possible(-tdt_max - 1)
+ # get range required by threshold values
+ tdt_min = min(acc_min, min_threshold)
+ tdt_max = max(acc_max, max_threshold)
+ if tdt_min < 0:
+ if abs(tdt_min) > tdt_max:
+ tdt = DataType.get_smallest_possible(tdt_min)
else:
- tdt = DataType.get_smallest_possible(tdt_max)
- assert np.vectorize(tdt.allowed)(
- threshold_tensor
- ).all(), "Thresholds in %s can't be expressed with type %s" % (
- self.onnx_node.name,
- str(tdt),
- )
- self.set_nodeattr("accDataType", tdt.name)
+ tdt = DataType.get_smallest_possible(-tdt_max - 1)
else:
- if acc_min < 0:
- if abs(acc_min) > acc_max:
- adt = DataType.get_smallest_possible(acc_min)
- else:
- adt = DataType.get_smallest_possible(-acc_max - 1)
+ tdt = DataType.get_smallest_possible(tdt_max)
+ assert np.vectorize(tdt.allowed)(
+ threshold_tensor
+ ).all(), "Thresholds in %s can't be expressed with type %s" % (
+ self.onnx_node.name,
+ str(tdt),
+ )
+ adt = tdt # Set activation datatype to the threshold datatype
+ else:
+ if acc_min < 0:
+ if abs(acc_min) > acc_max:
+ adt = DataType.get_smallest_possible(acc_min)
else:
- adt = DataType.get_smallest_possible(acc_max)
- # ensure a datatype divisible by 8-bits in case this is the last node
- bw = roundup_to_integer_multiple(adt.bitwidth(), 8)
- new_adt_name = adt.name.replace(str(adt.bitwidth()), str(bw))
- adt = DataType[new_adt_name]
- self.set_nodeattr("accDataType", adt.name)
- # for no-activation nodes, output dt = acc dt
- self.set_nodeattr("outputDataType", adt.name)
+ adt = DataType.get_smallest_possible(-acc_max - 1)
+ else:
+ adt = DataType.get_smallest_possible(acc_max)
+ # if this is the last node in the graph, then ensure the datatype is
+ # divisibly by 8 bits
+ if model.find_direct_successors(self.onnx_node) is None:
+ bw = roundup_to_integer_multiple(adt.bitwidth(), 8)
+ new_adt_name = adt.name.replace(str(adt.bitwidth()), str(bw))
+ adt = DataType[new_adt_name]
+ # for no-activation nodes, output dt = acc dt
+ self.set_nodeattr("outputDataType", adt.name)
+ self.set_nodeattr("accDataType", adt.name)
return DataType[self.get_nodeattr("accDataType")]
def minimize_weight_bit_width(self, model):