Merge pull request #537 from JochiSt/master

returning integer from _compute_n_samples

hls4ml/model/graph.py

@@ -632,7 +632,7 @@ def _compute_n_samples(self, x):
         if not all([n_samples[i] == n_samples[i+1] for i in range(len(xlist)-1)]):
             raise Exception('Input size mismatch, not all inputs match')
 
-        return n_sample
+        return int(n_sample)
 
     def predict(self, x):
         top_function, ctype = self._get_top_function(x)
@@ -651,9 +651,9 @@ def predict(self, x):
             for i in range(n_samples):
                 predictions = [np.zeros(yj.size(), dtype=ctype) for yj in self.get_output_variables()]
                 if n_inputs == 1:
-                    inp = [x[i]]
+                    inp = [np.asarray(x[i])]
                 else:
-                    inp = [xj[i] for xj in x]
+                    inp = [np.asarray(xj[i]) for xj in x]
                 argtuple = inp
                 argtuple += predictions
                 argtuple = tuple(argtuple)
@@ -683,6 +683,7 @@ def trace(self, x):
         top_function, ctype = self._get_top_function(x)
         n_samples = self._compute_n_samples(x)
         n_inputs = len(self.get_input_variables())
+        n_outputs = len(self.get_output_variables())
 
         class TraceData(ctypes.Structure):
             _fields_ = [('name', ctypes.c_char_p),
@@ -721,16 +722,15 @@ class TraceData(ctypes.Structure):
             alloc_func(ctypes.sizeof(ctype))
 
             for i in range(n_samples):
-                predictions = np.zeros(self.get_output_variables()[0].size(), dtype=ctype)
+                predictions = [np.zeros(yj.size(), dtype=ctype) for yj in self.get_output_variables()]
                 if n_inputs == 1:
-                    top_function(x[i], predictions, ctypes.byref(ctypes.c_ushort()), ctypes.byref(ctypes.c_ushort()))
+                    inp = [np.asarray(x[i])]
                 else:
-                    inp = [xj[i] for xj in x]
-                    argtuple = inp
-                    argtuple += [predictions]
-                    argtuple += [ctypes.byref(ctypes.c_ushort()) for k in range(len(inp)+1)]
-                    argtuple = tuple(argtuple)
-                    top_function(*argtuple)
+                    inp = [np.asarray(xj[i]) for xj in x]
+                argtuple = inp
+                argtuple += predictions
+                argtuple = tuple(argtuple)
+                top_function(*argtuple)
                 output.append(predictions)
                 collect_func(trace_data)
                 for trace in trace_data:
@@ -742,15 +742,19 @@ class TraceData(ctypes.Structure):
             for key in trace_output.keys():
                 trace_output[key] = np.asarray(trace_output[key])
 
-            #Convert to numpy array
-            output = np.asarray(output)
+            # Convert to list of numpy arrays (one for each output)
+            output = [np.asarray([output[i_sample][i_output] for i_sample in range(n_samples)]) for i_output in range(n_outputs)]
 
             free_func()
         finally:
             os.chdir(curr_dir)
 
-        if n_samples == 1:
+        if n_samples == 1 and n_outputs == 1:
+            return output[0][0], trace_output
+        elif n_outputs == 1:
             return output[0], trace_output
+        elif n_samples == 1:
+            return [output_i[0] for output_i in output], trace_output
         else:
             return output, trace_output
 

test/pytest/test_graph.py

@@ -200,6 +200,8 @@ def test_multiple_outputs(batch):
   X1 = np.random.randint(0, 100, size=(batch, 10)).astype(float)
   y = model.predict(X1)
   y_hls = hls_model.predict(X1)
+  # test trace as well
+  y_hls, hls_trace = hls_model.trace(X1)
   for y_i, y_hls_i in zip(y, y_hls):
     y_hls_i = y_hls_i.reshape(y_i.shape)
-    np.testing.assert_allclose(y_i, y_hls_i, rtol=0)
+    np.testing.assert_allclose(y_i, y_hls_i, rtol=0)