kriskorrel-cw

Kris Korrel

kriskorrel-cw / saved_model_custom_layer.py

Created April 9, 2020 09:37

Saving a model not possible with custom layer, using different signature

	from tensorflow import keras
	from tensorflow.keras.layers import BatchNormalization, Dense


	class BatchNormalization1(BatchNormalization):
	def call(self, inputs, **kwargs):
	return super(BatchNormalization1, self).call(inputs, **kwargs)


	class BatchNormalization2(BatchNormalization):

kriskorrel-cw / copy_tensor_bug.py

Last active June 10, 2020 07:50

MWE copy tensor bug

	import numpy as np
	import os
	import tensorflow as tf
	from tensorflow.keras.layers import Dense, Flatten
	from tensorflow.keras.optimizers import Adam

	os.environ['CUDA_VISIBLE_DEVICES'] = '0'
	os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true'

kriskorrel-cw / copy_tensor_bug.out

Last active June 10, 2020 07:51

	ssh://[email protected]:22/home/username/.local/share/virtualenvs/project-1BQtdZDZ/bin/python -u /data/personal/username/deployed/project/project/bin/copy_tensor_bug.py
	2020-06-10 09:49:40.096725: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
	2020-06-10 09:49:40.178485: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
	2020-06-10 09:49:40.178972: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1561] Found device 0 with properties:
	pciBusID: 0000:01:00.0 name: GeForce GTX 1080 computeCapability: 6.1
	coreClock: 1.86GHz coreCount: 20 deviceMemorySize: 7.93GiB deviceMemoryBandwidth: 298.32GiB/s
	2020-06-10 09:49:40.179163: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
	2020-06-10 09:49:40.180759: I tensorflow/stream_executor/platfo