xihao_

Same problem. GPU utilization rate keep 0% almost all of the time.

I figure out the problem. there is something wrong with the `inference` function in model.py : ``` def inference(self): if self.architecture.lower()=='alexnet': return alexnet(self.inputs, self.num_classes, self.wd, tf.where(self.is_training, 0.5, 1.0), self.is_training) elif...