From 8850581ba4ce620156bf6777964bb7e53b77f9b4 Mon Sep 17 00:00:00 2001
From: Yaman Umuroglu <maltanar@gmail.com>
Date: Tue, 3 Dec 2019 01:23:40 +0000
Subject: [PATCH] [StreamingFC] check num inputs before trying to grab
thresholds
---
.../fpgadataflow/streamingfclayer_batch.py | 59 ++++++++++---------
1 file changed, 30 insertions(+), 29 deletions(-)
diff --git a/src/finn/custom_op/fpgadataflow/streamingfclayer_batch.py b/src/finn/custom_op/fpgadataflow/streamingfclayer_batch.py
index e375e239e..27272bc10 100644
--- a/src/finn/custom_op/fpgadataflow/streamingfclayer_batch.py
+++ b/src/finn/custom_op/fpgadataflow/streamingfclayer_batch.py
@@ -185,35 +185,36 @@ class StreamingFCLayer_Batch(HLSCustomOp):
f_weights.write(weight_hls_code)
f_weights.close()
# thresholds
- thresholds = model.get_initializer(self.onnx_node.input[2])
- if thresholds is not None:
- threshold_tensor = self.get_hls_compatible_threshold_tensor(thresholds)
- tdt = DataType.INT32
- # use UINT32 threshold export for bipolar times bipolar
- inp_is_bipolar = self.get_input_datatype() == DataType.BIPOLAR
- wt_is_bipolar = self.get_weight_datatype() == DataType.BIPOLAR
- if inp_is_bipolar and wt_is_bipolar:
- tdt = DataType.UINT32
- thresholds_hls_code = numpy_to_hls_code(
- threshold_tensor, tdt, "thresholds", False, True
- )
- # write thresholds into thresh.h
- code_gen_dir = self.get_nodeattr("code_gen_dir")
- f_thresh = open("{}/thresh.h".format(code_gen_dir), "w")
- tdt_hls = tdt.get_hls_datatype_str()
- odt_hls = self.get_output_datatype().get_hls_datatype_str()
- f_thresh.write(
- "static ThresholdsActivation<{},{},{},{},{},{}> threshs = ".format(
- self.get_nodeattr("TMEM"),
- self.get_nodeattr("PE"),
- threshold_tensor.shape[-1],
- tdt_hls,
- odt_hls,
- self.get_nodeattr("ActVal"),
+ if len(self.onnx_node.input) > 2:
+ thresholds = model.get_initializer(self.onnx_node.input[2])
+ if thresholds is not None:
+ threshold_tensor = self.get_hls_compatible_threshold_tensor(thresholds)
+ tdt = DataType.INT32
+ # use UINT32 threshold export for bipolar times bipolar
+ inp_is_bipolar = self.get_input_datatype() == DataType.BIPOLAR
+ wt_is_bipolar = self.get_weight_datatype() == DataType.BIPOLAR
+ if inp_is_bipolar and wt_is_bipolar:
+ tdt = DataType.UINT32
+ thresholds_hls_code = numpy_to_hls_code(
+ threshold_tensor, tdt, "thresholds", False, True
)
- )
- f_thresh.write(thresholds_hls_code)
- f_thresh.close()
+ # write thresholds into thresh.h
+ code_gen_dir = self.get_nodeattr("code_gen_dir")
+ f_thresh = open("{}/thresh.h".format(code_gen_dir), "w")
+ tdt_hls = tdt.get_hls_datatype_str()
+ odt_hls = self.get_output_datatype().get_hls_datatype_str()
+ f_thresh.write(
+ "static ThresholdsActivation<{},{},{},{},{},{}> threshs = ".format(
+ self.get_nodeattr("TMEM"),
+ self.get_nodeattr("PE"),
+ threshold_tensor.shape[-1],
+ tdt_hls,
+ odt_hls,
+ self.get_nodeattr("ActVal"),
+ )
+ )
+ f_thresh.write(thresholds_hls_code)
+ f_thresh.close()
def execute_node(self, context, graph):
node = self.onnx_node
@@ -247,7 +248,7 @@ class StreamingFCLayer_Batch(HLSCustomOp):
# execute the precompiled model
super().exec_precompiled_singlenode_model()
# load output npy file
- super().npy_to_dynamic_output()
+ super().npy_to_dynamic_output(context)
def global_includes(self):
self.code_gen_dict["$GLOBALS$"] = ['#include "weights.hpp"']
--
GitLab