MCPcopy
hub / github.com/hanzhanggit/StackGAN / build_model

Method build_model

stageI/trainer.py:279–300  ·  view source on GitHub ↗
(self, sess)

Source from the content-addressed store, hash-verified

277 return img_summary
278
279 def build_model(self, sess):
280 self.init_opt()
281 sess.run(tf.initialize_all_variables())
282
283 if len(self.model_path) > 0:
284 print("Reading model parameters from %s" % self.model_path)
285 restore_vars = tf.all_variables()
286 # all_vars = tf.all_variables()
287 # restore_vars = [var for var in all_vars if
288 # var.name.startswith('g_') or
289 # var.name.startswith('d_')]
290 saver = tf.train.Saver(restore_vars)
291 saver.restore(sess, self.model_path)
292
293 istart = self.model_path.rfind('_') + 1
294 iend = self.model_path.rfind('.')
295 counter = self.model_path[istart:iend]
296 counter = int(counter)
297 else:
298 print("Created model with fresh parameters.")
299 counter = 0
300 return counter
301
302 def train(self):
303 config = tf.ConfigProto(allow_soft_placement=True)

Callers 1

trainMethod · 0.95

Calls 1

init_optMethod · 0.95

Tested by

no test coverage detected