still troubleshooting

selfdrive/camerad/cameras/camera_webcam.cc

@@ -52,8 +52,7 @@ void open_gl_stream_def(CameraState * s, char* camera_id, int width, int height,
   if (flip == 1) {
     flip_command = "videoflip method=rotate-180 ! ";
   }
-  //1.5 is a fixed factor that came from Comma implementation
-  sprintf(*strm_def,strm_template.c_str(),camera_id, width, height, s->fps, (int)(s->ci.frame_width*fx*1.5), (int)(s->ci.frame_height*fx*1.5), flip_command.c_str(), s->ci.frame_width, s->ci.frame_height);
+  sprintf(*strm_def,strm_template.c_str(),camera_id, width, height, s->fps, (int)(s->ci.frame_width*fx), (int)(s->ci.frame_height*fx), flip_command.c_str(), s->ci.frame_width, s->ci.frame_height);
   printf(" GL Stream :[%s]\n",*strm_def);
 }
 

selfdrive/common/SConscript

@@ -18,7 +18,7 @@ files = [
 
 
 if is_tbp:
-  defines = {"CLU_NO_CACHE": None}   
+  defines = {} #{"CLU_NO_CACHE": None}   
   files += [                         
     'visionbuf_cl.c',                
   ]                                                                     

selfdrive/modeld/models/driving.h

@@ -4,9 +4,9 @@
 // gate this here
 #define TEMPORAL
 #define DESIRE
+#define TRAFFIC_CONVENTION
 
 #ifdef QCOM
-#define TRAFFIC_CONVENTION
 #include <eigen3/Eigen/Dense>
 #else
 #include <Eigen/Dense>

selfdrive/modeld/runners/keras_runner.py

@@ -24,16 +24,25 @@ def read(sz):
 def write(d):
   os.write(1, d.tobytes())
 
-def run_loop(m):
+def run_loop(m,name):
   isize = m.inputs[0].shape[1]
   osize = m.outputs[0].shape[1]
+  int 
   print("ready to run keras model %d -> %d" % (isize, osize), file=sys.stderr)
+  cnt = 0
   while 1:
+    if name == "supercombo":
+      cnt = (cnt + 1 ) % 100
+    if (cnt % 20 == 1):
+        print("Processing frame %d " % cnt, file=sys.stderr)
     # check parent process, if ppid is 1, then modeld is no longer running and the runner should exit.
     if os.getppid() == 1:
-      break
+      print("exiting due to Parent PID", file=sys.stderr)  
+      #break
     idata = read(isize).reshape((1, isize))
     ret = m.predict_on_batch(idata)
+    if (cnt % 20 == 1):
+        print(ret, file=sys.stderr)
     write(ret)
 
 if __name__ == "__main__":
@@ -42,16 +51,26 @@ def run_loop(m):
   gpus = tf.config.experimental.list_physical_devices('GPU')
   model_path = Path(sys.argv[1]).parents[0]
   name = Path(sys.argv[1]).stem
-  print("\n\nRunning [%s] with [%s]\n\n" % (sys.argv[1],name))
-  if name == "supercombo":
-    if len(gpus) > 0:
+  print("\n\nRunning [%s] with path [%s]\n\n" % (name,sys.argv[1]),file=sys.stderr)
+  if len(gpus) > 0:
+    if name == "supercombo":
       tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=2048)])
-  else:
-    if len(gpus) > 0:
-      tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=1524)])
-  with open(f"{model_path / name}.model.keras", "r") as json_file:
-    m = model_from_json(json_file.read())
-  m.load_weights(f"{model_path / name}.weights.keras")
+    else:
+      tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=300)])
+  m = load_model(sys.argv[1])
   print(m, file=sys.stderr)
   bs = [int(np.product(ii.shape[1:])) for ii in m.inputs]
   ri = keras.layers.Input((sum(bs),))
+
+  tii = []
+  acc = 0
+  for i, ii in enumerate(m.inputs):
+    print(ii, file=sys.stderr)
+    ti = keras.layers.Lambda(lambda x: x[:,acc:acc+bs[i]], output_shape=(1, bs[i]))(ri)
+    acc += bs[i]
+    tr = keras.layers.Reshape(ii.shape[1:])(ti)
+    tii.append(tr)
+  no = keras.layers.Concatenate()(m(tii))
+  m = Model(inputs=ri, outputs=[no])
+  run_loop(m,name)
+

selfdrive/modeld/runners/keras_runner_split.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python3
+# TODO: why are the keras models saved with python 2?
+from __future__ import print_function
+
+import tensorflow as tf
+import os
+import sys
+import tensorflow.keras as keras
+import numpy as np
+from tensorflow.keras.models import Model
+from tensorflow.keras.models import model_from_json
+from pathlib import Path
+
+def read(sz):
+  dd = []
+  gt = 0
+  while gt < sz*4:
+    st = os.read(0, sz*4 - gt)
+    assert(len(st) > 0)
+    dd.append(st)
+    gt += len(st)
+  return np.fromstring(b''.join(dd), dtype=np.float32)
+
+def write(d):
+  os.write(1, d.tobytes())
+
+def run_loop(m,name):
+  isize = m.inputs[0].shape[1]
+  osize = m.outputs[0].shape[1]
+  int 
+  print("ready to run keras model %d -> %d" % (isize, osize), file=sys.stderr)
+  cnt = 0
+  while 1:
+    if name == "supercombo":
+      cnt = (cnt + 1 ) % 100
+    if (cnt % 20 == 1):
+        print("Processing frame %d " % cnt, file=sys.stderr)
+    # check parent process, if ppid is 1, then modeld is no longer running and the runner should exit.
+    if os.getppid() == 1:
+      print("exiting due to Parent PID", file=sys.stderr)  
+      #break
+    idata = read(isize).reshape((1, isize))
+    ret = m.predict_on_batch(idata)
+    if (cnt % 20 == 1):
+        print(ret, file=sys.stderr)
+    write(ret)
+
+if __name__ == "__main__":
+  print(tf.__version__, file=sys.stderr)
+  # limit gram alloc
+  gpus = tf.config.experimental.list_physical_devices('GPU')
+  model_path = Path(sys.argv[1]).parents[0]
+  name = Path(sys.argv[1]).stem
+  print("\n\nRunning [%s] with path [%s] and name [%s] from [%s]\n\n" % (sys.argv[1],model_path,name,f"{model_path / name}.weights.keras"),file=sys.stderr)
+  if name == "supercombo":
+    if len(gpus) > 0:
+      tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=2548)])
+  else:
+    if len(gpus) > 0:
+      tf.config.experimental.set_virtual_device_configuration(gpus[0], [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=256)])
+  with open(f"{model_path / name}.model.keras", "r") as json_file:
+    m = model_from_json(json_file.read())
+  m.load_weights(f"{model_path / name}.weights.keras")
+  print(m, file=sys.stderr)
+  bs = [int(np.product(ii.shape[1:])) for ii in m.inputs]
+  ri = keras.layers.Input((sum(bs),))
+  tii = []
+  acc = 0
+  for i, ii in enumerate(m.inputs):
+    if name == "supercombo":
+      print("SuperCombo Loading i=%d" % i, file=sys.stderr)
+    ti = keras.layers.Lambda(lambda x: x[:,acc:acc+bs[i]], output_shape=(1, bs[i]))(ri)
+    acc += bs[i]
+    tr = keras.layers.Reshape(ii.shape[1:])(ti)
+    tii.append(tr)
+  no = keras.layers.Concatenate()(m(tii))
+  m = Model(inputs=ri, outputs=[no])
+  run_loop(m,name)
+

tools/keras/split_keras.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python3
+#!/usr/bin/env python3
 
 import tensorflow as tf
 import os