Trust tensorflow's thread safety (it wasn't the cause of an earlier bug)

Dependencies

• [2] TCMUFA6E Try some different hyperparameters
• [3] 4IXWCEHM Gamma correction
• [4] SJHJS463 Model trainer: main function to open sdl window and launch second thread for actual training
• [5] E742MTJA Fix segfault (don't pass pointers between functions in different SDL versions), minor refactor
• [6] 2ABZP2KN Model trainer: save map.png after training
• [7] ALMG52BO model trainer: speed up by waiting 15 seconds between monitor renders
• [8] EB3DTD43 Model trainer: use ground altitude instead of crust mass distribution
• [9] ZRPV3GAJ Model trainer: Add function to make tensor of cartesian coordinates for drawing the map (equirectangular projection)
• [10] 7ML3OFE7 Model trainer: initial train and visualize thread pair (total crust mass; todo: altitude instead)
• [11] FL3C6ERZ Assume python is being run with -m switch
• [12] ROQCAPZJ Begin function for showing the map (for now just opens SDL window)
• [*] ZM2EMAZO Start doing python multi-module stuff properly

Change contents

• edit in trainmodel/src/model.py at line 36
  [4.2054]→[4.2054:2238](∅→∅),[4.2238]→[2.139:191](∅→∅),[2.191]→[4.2284:2530](∅→∅),[4.2284]→[4.2284:2530](∅→∅)
  − class Monitor(keras.callbacks.Callback):

  − def __init__(self, target):

  − self.lock = threading.Lock()

  − self.target = target

  − self.stable = clone_my_model(target)

  − def on_train_batch_end(self, batch, logs=None):

  − self.lock.acquire()

  − self.stable.set_weights(self.target.get_weights())

  − self.lock.release()

  − def snapshot(self):

  − self.lock.acquire()

  − result = self.stable

  − self.lock.release()

  − return result
• replacement in trainmodel/src/drawmap.py at line 12
  [4.54]→[4.535:553](∅→∅)
  − def run(monitor):
  [4.54]

  [4.98]
  + def run(m):
• replacement in trainmodel/src/drawmap.py at line 22
  [4.82]→[4.82:127](∅→∅)
  − outputs = monitor.snapshot()(inputs)
  [4.82]

  [3.0]
  + outputs = m(inputs)
• replacement in trainmodel/src/__main__.py at line 9
  [4.30]→[4.50:80](∅→∅)
  − def train(monitor, filename):
  [4.30]

  [4.51]
  + def train(m, filename):
• replacement in trainmodel/src/__main__.py at line 12
  [4.129]→[2.192:303](∅→∅)
  − monitor.target.fit(x=training_data[0], y=training_data[1], batch_size=100, epochs=5000, callbacks=monitor)
  [4.129]

  [4.128]
  + m.fit(x=training_data[0], y=training_data[1], batch_size=100, epochs=5000)
• replacement in trainmodel/src/__main__.py at line 15
  [4.110]→[4.110:175](∅→∅)
  − outputs = tf.reshape(monitor.target(inputs), (1024,2048, 1))
  [4.110]

  [3.101]
  + outputs = tf.reshape(m(inputs), (1024,2048, 1))
• replacement in trainmodel/src/__main__.py at line 28
  [4.390]→[4.390:508](∅→∅)
  − monitor = model.Monitor(m)

  − bg_thread = threading.Thread(target=train, args=(monitor, sys.argv[1]), kwargs={})
  [4.390]

  [4.230]
  + bg_thread = threading.Thread(target=train, args=(m, sys.argv[1]), kwargs={})
• replacement in trainmodel/src/__main__.py at line 30
  [4.252]→[4.509:534](∅→∅)
  − drawmap.run(monitor)
  [4.252]

  [4.270]
  + drawmap.run(m)