tensorflow - NotFoundError：没有为 GPU/tf.eager 和 keras 注册的“ZerosLike”OpKernel
问题描述
我写下了我自己的网络复制pix2pixHD。这是对pix2pix的简单修改，只是将生成器复制为两个，将鉴别器复制为三个，它们处理不同级别的分辨率。
但是当我尝试训练时，它不断返回，Opkernel_zeroslike unfound error但找不到发生此错误的原因。
我通常修改了 pix2pix以制作此代码，并且pix2pix不会返回错误并且成功地进行了训练。
对此有任何提示Opkernel unfound error吗？
#### this sciprt based on tensorflow v.1.10
#!pip install --upgrade tensorflow-gpu==1.10.0 #normally the default provided by aws is 1.8.0
# Keras Implementation of pix2pix
import os
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"]="0"
import sys
import time
import glob
import random
from functools import reduce
import cv2
import pandas as pd
import numpy as np
from PIL import Image
import matplotlib.pyplot as plt
from IPython.display import clear_output
import scipy
import tensorflow as tf
tf.enable_eager_execution()
from tensorflow.python.data import Dataset
from tensorflow.python.data import Iterator
from tensorflow.python.framework import ops
from tensorflow.python.ops import math_ops
import collections
import math
import time
import util
now = time.strftime("%Y%m%d-%H%M%S")
# Paths
data = './data/ipx_2'
PATH = os.path.join(data, 'train')
image_paths = sorted(glob.glob(PATH+"/*"))
# Hypers
BUFFER_SIZE = len(image_paths)
IMG_WIDTH = 256
IMG_HEIGHT = 256
BATCH_SIZE = {IMG_WIDTH == 256 : 16, IMG_WIDTH == 512: 10}.get(True, "IMG_WIDTH must be either 256 or 512")
TIME_SPACE = 11
print(BUFFER_SIZE, BATCH_SIZE, IMG_HEIGHT, TIME_SPACE) #for sanity check
# I/O Utils
def load_image(image_path, diff = True, corr =True, time_space = TIME_SPACE):
image = tf.read_file(image_path)
image = tf.image.decode_png(image)
w = tf.shape(image)[0]
print(w)
target_image = image[:, -w:, :]
input_image = image[:, :-w, :]
input_image = tf.reshape(input_image, [256,256,66])
target_image = tf.reshape(target_image, [256,256,3])
target_image = tf.cast(target_image, tf.float32)
input_image = tf.cast(input_image, tf.float32)
# normalizing the images to [-1, 1]
target_image = (target_image / 127.5) - 1
input_image = (input_image / 127.5) - 1
return input_image, target_image
EPOCHS = int(1e+3)
train_dataset = tf.data.Dataset.list_files(PATH+'/*.png')
train_dataset = train_dataset.shuffle(BUFFER_SIZE)
train_dataset = train_dataset.map(lambda x: load_image(x))
train_dataset = train_dataset.batch(BATCH_SIZE)
print(train_dataset)
# Generator
OUTPUT_CHANNELS = 3
class Downsample(tf.keras.Model): #conv2d && Lrelu
def __init__(self, filters, size=4, apply_batchnorm=True, first = False):
super(Downsample, self).__init__()
self.apply_batchnorm = apply_batchnorm
self.first = first
initializer = tf.random_normal_initializer(0., 0.02)
self.conv1 = tf.keras.layers.Conv2D(filters,
(size, size),
strides=2,
padding='same',
kernel_initializer=initializer,
use_bias=False)
if self.apply_batchnorm:
self.batchnorm = tf.keras.layers.BatchNormalization()
def call(self, x, training):
x = self.conv1(x)
if self.apply_batchnorm:
x = self.batchnorm(x, training=training)
if not self.first:
x = tf.nn.leaky_relu(x)
return x
class Upsample(tf.keras.Model):
def __init__(self, filters, size, apply_dropout=False, first = False):
super(Upsample, self).__init__()
self.apply_dropout = apply_dropout
self.first = first
initializer = tf.random_normal_initializer(0., 0.02)
self.up_conv = tf.keras.layers.Conv2DTranspose(filters,
(size, size),
strides=2,
padding='same',
kernel_initializer=initializer,
use_bias=False)
self.refine = tf.keras.layers.Conv2D(filters,
(3, 3),
strides=1,
padding='same',
kernel_initializer=initializer,
use_bias=False)
self.batchnorm = tf.keras.layers.BatchNormalization()
if self.apply_dropout:
self.dropout = tf.keras.layers.Dropout(0.5) #drop_out rate here
def call(self, x1, x2, training):
x = self.up_conv(x1)
x = self.batchnorm(x, training=training)
if self.apply_dropout:
x = self.dropout(x, training=training)
x = self.refine(x)
x = tf.nn.relu(x)
x = self.refine(x)
x = tf.nn.relu(x)
if not self.first:
x = tf.concat([x, x2], axis=-1)
return x
class Generator(tf.keras.Model):
def __init__(self):
super(Generator, self).__init__()
initializer = tf.random_normal_initializer(0., 0.02)
self.down1 = Downsample(64, 4, first = True)
self.down2 = Downsample(128, 4)
self.down3 = Downsample(256, 4)
self.down4 = Downsample(512, 4)
self.down5 = Downsample(512, 4)
self.down6 = Downsample(512, 4)
self.down7 = Downsample(512, 4)
self.down8 = Downsample(512, 4)
self.up1 = Upsample(512, 4, apply_dropout=True, first = True)
self.up2 = Upsample(512, 4, apply_dropout=True)
self.up3 = Upsample(512, 4, apply_dropout=True)
self.up4 = Upsample(512, 4)
self.up5 = Upsample(256, 4)
self.up6 = Upsample(128, 4)
self.up7 = Upsample(64, 4)
self.last = tf.keras.layers.Conv2DTranspose(OUTPUT_CHANNELS,
(4, 4),
strides=2,
padding='same',
kernel_initializer=initializer)
@tf.contrib.eager.defun
def call(self, x, training):
# x shape == (bs, 256, 256, 3)
x1 = self.down1(x, training=training) # (bs, 128, 128, 64)
x2 = self.down2(x1, training=training) # (bs, 64, 64, 128)
x3 = self.down3(x2, training=training) # (bs, 32, 32, 256)
x4 = self.down4(x3, training=training) # (bs, 16, 16, 512)
x5 = self.down5(x4, training=training) # (bs, 8, 8, 512)
x6 = self.down6(x5, training=training) # (bs, 4, 4, 512)
x7 = self.down7(x6, training=training) # (bs, 2, 2, 512)
x8 = self.down8(x7, training=training) # (bs, 1, 1, 512)
x9 = self.up1(x8, x7, training=training) # (bs, 2, 2, 1024)
x10 = self.up2(x9, x6, training=training) # (bs, 4, 4, 1024)
x11 = self.up3(x10, x5, training=training) # (bs, 8, 8, 1024)
x12 = self.up4(x11, x4, training=training) # (bs, 16, 16, 1024)
x13 = self.up5(x12, x3, training=training) # (bs, 32, 32, 512)
x14 = self.up6(x13, x2, training=training) # (bs, 64, 64, 256)
x15 = self.up7(x14, x1, training=training) # (bs, 128, 128, 128)
x16 = self.last(x15) # (bs, 256, 256, 3)
x16 = tf.nn.tanh(x16)
return x16
# Discriminator
class DiscDownsample(tf.keras.Model):
def __init__(self, filters, size, apply_batchnorm=True, first = False):
super(DiscDownsample, self).__init__()
self.first = first
self.apply_batchnorm = apply_batchnorm
initializer = tf.random_normal_initializer(0., 0.02)
self.conv1 = tf.keras.layers.Conv2D(filters,
(size, size),
strides=2,
padding='same',
kernel_initializer=initializer,
use_bias=False)
if self.apply_batchnorm:
self.batchnorm = tf.keras.layers.BatchNormalization()
def call(self, x, training):
x = self.conv1(x)
if self.apply_batchnorm:
x = self.batchnorm(x, training=training)
x = tf.nn.leaky_relu(x)
return x
class Discriminator(tf.keras.Model):
def __init__(self):
super(Discriminator, self).__init__()
initializer = tf.random_normal_initializer(0., 0.02)
self.down1 = DiscDownsample(64, 4, False)
self.down2 = DiscDownsample(128, 4)
self.down3 = DiscDownsample(256, 4)
# we are zero padding here with 1 because we need our shape to
# go from (batch_size, 32, 32, 256) to (batch_size, 31, 31, 512)
self.zero_pad1 = tf.keras.layers.ZeroPadding2D()
self.conv = tf.keras.layers.Conv2D(512,
(4, 4),
strides=1,
kernel_initializer=initializer,
use_bias=False)
self.batchnorm1 = tf.keras.layers.BatchNormalization()
# shape change from (batch_size, 31, 31, 512) to (batch_size, 30, 30, 1)
self.zero_pad2 = tf.keras.layers.ZeroPadding2D()
self.last = tf.keras.layers.Conv2D(1,
(4, 4),
strides=1,
kernel_initializer=initializer)
@tf.contrib.eager.defun
def call(self, inp, tar, training):
# concatenating the input and the target
x = tf.concat([inp, tar], axis=-1) # (bs, 256, 256, channels*2)
x = self.down1(x, training=training) # (bs, 128, 128, 64)
x = self.down2(x, training=training) # (bs, 64, 64, 128)
x = self.down3(x, training=training) # (bs, 32, 32, 256)
x = self.zero_pad1(x) # (bs, 34, 34, 256)
x = self.conv(x) # (bs, 31, 31, 512)
x = self.batchnorm1(x, training=training)
x = tf.nn.leaky_relu(x)
x = self.zero_pad2(x) # (bs, 33, 33, 512)
# don't add a sigmoid activation here since
# the loss function expects raw logits.
x = self.last(x) # (bs, 30, 30, 1)
return x
generator = Generator()
discriminator = Discriminator()
# Loss and Opter
LAMBDA = 100 #param for l1_loss
def discriminator_loss(disc_real_output, disc_generated_output):
real_loss = tf.losses.sigmoid_cross_entropy(multi_class_labels = tf.ones_like(disc_real_output),
logits = disc_real_output)
generated_loss = tf.losses.sigmoid_cross_entropy(multi_class_labels = tf.zeros_like(disc_generated_output),
logits = disc_generated_output)
total_disc_loss = real_loss + generated_loss
return total_disc_loss
def generator_loss(disc_generated_output, gen_output, target):
gan_loss = tf.losses.sigmoid_cross_entropy(multi_class_labels = tf.ones_like(disc_generated_output),
logits = disc_generated_output)
# mean absolute error
l1_loss = tf.reduce_mean(tf.abs(target - gen_output))
total_gen_loss = gan_loss + (LAMBDA * l1_loss)
return total_gen_loss
generator_optimizer = tf.train.AdamOptimizer(2e-4, beta1=0.5)
discriminator_optimizer = tf.train.AdamOptimizer(2e-4, beta1=0.5)
# ckpt
checkpoint_dir = os.path.join('./log/', now)
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
checkpoint = tf.train.Checkpoint(generator_optimizer=generator_optimizer,
discriminator_optimizer=discriminator_optimizer,
generator=generator,
discriminator=discriminator)
# Train
def generate_images(epoch, step, model, test_input, tar, logDir):
# the training=True is intentional here since
# we want the batch statistics while running the model
# on the test dataset. If we use training=False, we will get
# the accumulated statistics learned from the training dataset
# (which we don't want)
prediction = model(test_input, training=True)
## for disply the plots, it takes 1.2sec
display_list = [tar[0], prediction[0]]
img1 = (tar[0] + 1.) * 127.5
img2 = (prediction[0] + 1.) * 127.5
img3 = np.append(img1, img2, axis = 1)
img3 = cv2.cvtColor(img3, cv2.COLOR_BGR2RGB)
image_name = "step-" + str(epoch+1) + "-"+ str(step) + ".jpg"
mkdirs(logDir + "/images/")
cv2.imwrite(logDir + "/images/" + image_name, img3)
def train(dataset, epochs):
for epoch in range(epochs):
if (epoch + 1) % 6 == 0:
epoch_start = time.time()
mkdirs(checkpoint_dir)
for step , (input_image, target) in enumerate(dataset):
if step % 100 == 0:
step_start = time.time()
with tf.GradientTape() as gen_tape, tf.GradientTape() as disc_tape:
gen_output = generator(input_image, training=True)
disc_real_output = discriminator(input_image, target, training=True)
disc_generated_output = discriminator(input_image, gen_output, training=True)
gen_loss = generator_loss(disc_generated_output, gen_output, target)
disc_loss = discriminator_loss(disc_real_output, disc_generated_output)
generator_gradients = gen_tape.gradient(gen_loss,
generator.variables)
discriminator_gradients = disc_tape.gradient(disc_loss,
discriminator.variables)
generator_optimizer.apply_gradients(zip(generator_gradients,
generator.variables))
discriminator_optimizer.apply_gradients(zip(discriminator_gradients,
discriminator.variables))
if step % 100 == 0:
print('Time taken for step {} of epoch {} is {} sec\n'.format(step + 1, epoch+1,
time.time()-step_start))
clear_output(wait=True)
for inp, tar in dataset.take(1):
generate_images(epoch, step, generator, inp, tar, checkpoint_dir)
# saving (checkpoint) the model every 6 epochs
if (epoch + 1) % 6 == 0:
print('Time taken for epoch {} is {} sec\n'.format(epoch + 1,
time.time()-epoch_start))
checkpoint.save(file_prefix = checkpoint_prefix)
train(train_dataset, EPOCHS)
错误
NotFoundError: No registered 'ZerosLike' OpKernel for GPU devices compatible with node Func/_191 = ZerosLike[T=DT_RESOURCE](downsample_conv2d_conv2d_readvariableop_resource)
(OpKernel was found, but attributes didn't match)
. Registered: device='CPU'; T in [DT_VARIANT]
device='CPU'; T in [DT_STRING]
device='CPU'; T in [DT_BOOL]
device='CPU'; T in [DT_COMPLEX128]
device='CPU'; T in [DT_COMPLEX64]
device='CPU'; T in [DT_DOUBLE]
device='CPU'; T in [DT_FLOAT]
device='CPU'; T in [DT_BFLOAT16]
device='CPU'; T in [DT_HALF]
device='CPU'; T in [DT_INT8]
device='CPU'; T in [DT_UINT8]
device='CPU'; T in [DT_INT16]
device='CPU'; T in [DT_UINT16]
device='CPU'; T in [DT_INT32]
device='CPU'; T in [DT_INT64]
device='GPU'; T in [DT_INT32]
device='GPU'; T in [DT_VARIANT]
device='GPU'; T in [DT_INT64]
device='GPU'; T in [DT_COMPLEX128]
device='GPU'; T in [DT_COMPLEX64]
device='GPU'; T in [DT_DOUBLE]
device='GPU'; T in [DT_FLOAT]
device='GPU'; T in [DT_BFLOAT16]
device='GPU'; T in [DT_HALF]
device='GPU'; T in [DT_BOOL]
[[Node: Func/_191 = ZerosLike[T=DT_RESOURCE](downsample_conv2d_conv2d_readvariableop_resource)]]
[[Node: gradients/generator_1/StatefulPartitionedCall_grad/SymbolicGradient = SymbolicGradient[Tin=[DT_FLOAT, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT], Tout=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT], f=__forward___inference_call_1463_1839[Tin=[DT_FLOAT, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, ..., DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE, DT_RESOURCE], Tout=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT], f=__forward___inference_call_1463_1839[]], _device="/job:localhost/replica:0/task:0/device:GPU:0"](resize_images_resizebilinear, generator_1_statefulpartitionedcall_args_1, generator_1_statefulpartitionedcall_args_2, generator_1_statefulpartitionedcall_args_3, generator_1_statefulpartitionedcall_args_4, generator_1_statefulpartitionedcall_args_5, generator_1_statefulpartitionedcall_args_6, generator_1_statefulpartitionedcall_args_7, generator_1_statefulpartitionedcall_args_8, generator_1_statefulpartitionedcall_args_9, generator_1_statefulpartitionedcall_args_10, generator_1_statefulpartitionedcall_args_11, generator_1_statefulpartitionedcall_args_12, generator_1_statefulpartitionedcall_args_13, generator_1_statefulpartitionedcall_args_14, generator_1_statefulpartitionedcall_args_15, generator_1_statefulpartitionedcall_args_16, generator_1_statefulpartitionedcall_args_17, generator_1_statefulpartitionedcall_args_18, generator_1_statefulpartitionedcall_args_19, generator_1_statefulpartitionedcall_args_20, generator_1_statefulpartitionedcall_args_21, generator_1_statefulpartitionedcall_args_22, generator_1_statefulpartitionedcall_args_23, generator_1_statefulpartitionedcall_args_24, generator_1_statefulpartitionedcall_args_25, generator_1_statefulpartitionedcall_args_26, generator_1_statefulpartitionedcall_args_27, generator_1_statefulpartitionedcall_args_28, generator_1_statefulpartitionedcall_args_29, generator_1_statefulpartitionedcall_args_30, generator_1_statefulpartitionedcall_args_31, generator_1_statefulpartitionedcall_args_32, generator_1_statefulpartitionedcall_args_33, generator_1_statefulpartitionedcall_args_34, generator_1_statefulpartitionedcall_args_35, generator_1_statefulpartitionedcall_args_36, generator_1_statefulpartitionedcall_args_37, generator_1_statefulpartitionedcall_args_38, generator_1_statefulpartitionedcall_args_39, generator_1_statefulpartitionedcall_args_40, generator_1_statefulpartitionedcall_args_41, generator_1_statefulpartitionedcall_args_42, generator_1_statefulpartitionedcall_args_43, generator_1_statefulpartitionedcall_args_44, generator_1_statefulpartitionedcall_args_45, generator_1_statefulpartitionedcall_args_46, generator_1_statefulpartitionedcall_args_47, generator_1_statefulpartitionedcall_args_48, generator_1_statefulpartitionedcall_args_49, generator_1_statefulpartitionedcall_args_50, generator_1_statefulpartitionedcall_args_51, generator_1_statefulpartitionedcall_args_52, generator_1_statefulpartitionedcall_args_53, generator_1_statefulpartitionedcall_args_54, generator_1_statefulpartitionedcall_args_55, generator_1_statefulpartitionedcall_args_56, generator_1_statefulpartitionedcall_args_57, generator_1_statefulpartitionedcall_args_58, generator_1_statefulpartitionedcall_args_59, generator_1_statefulpartitionedcall_args_60, generator_1_statefulpartitionedcall_args_61, generator_1_statefulpartitionedcall_args_62, generator_1_statefulpartitionedcall_args_63, generator_1_statefulpartitionedcall_args_64, generator_1_statefulpartitionedcall_args_65, generator_1_statefulpartitionedcall_args_66, generator_1_statefulpartitionedcall_args_67, generator_1_statefulpartitionedcall_args_68, generator_1_statefulpartitionedcall_args_69, generator_1_statefulpartitionedcall_args_70, gradients/downsample_6/conv2d_24/Conv2D_grad/Conv2DBackpropInput, gradients/zeros_like_88, gradients/zeros_like_89, gradients/zeros_like_90, gradients/zeros_like_91, gradients/zeros_like_92, gradients/zeros_like_93, gradients/zeros_like_94, gradients/zeros_like_95, gradients/zeros_like_96, gradients/zeros_like_97, gradients/zeros_like_98, gradients/zeros_like_99, gradients/zeros_like_100, gradients/zeros_like_101, gradients/zeros_like_102, gradients/zeros_like_103, gradients/zeros_like_104, gradients/zeros_like_105, gradients/zeros_like_106, gradients/zeros_like_107, gradients/zeros_like_108, gradients/zeros_like_109, gradients/zeros_like_110, gradients/zeros_like_111, gradients/zeros_like_112, gradients/zeros_like_113, gradients/zeros_like_114, gradients/zeros_like_115, gradients/zeros_like_116, gradients/zeros_like_117, gradients/zeros_like_118, gradients/zeros_like_119, gradients/zeros_like_120, gradients/zeros_like_121, gradients/zeros_like_122, gradients/zeros_like_123, gradients/zeros_like_124, gradients/zeros_like_125, gradients/zeros_like_126, gradients/zeros_like_127, gradients/zeros_like_128, gradients/zeros_like_129, gradients/zeros_like_130, gradients/zeros_like_131, gradients/zeros_like_132, gradients/zeros_like_133, gradients/zeros_like_134, gradients/zeros_like_135, gradients/zeros_like_136, gradients/zeros_like_137, gradients/zeros_like_138, gradients/zeros_like_139, gradients/zeros_like_140, gradients/zeros_like_141, gradients/zeros_like_142, gradients/zeros_like_143, gradients/zeros_like_144, gradients/zeros_like_145, gradients/zeros_like_146, gradients/zeros_like_147, gradients/zeros_like_148, gradients/zeros_like_149, gradients/zeros_like_150, gradients/zeros_like_151, gradients/zeros_like_152, gradients/zeros_like_153, gradients/zeros_like_154, gradients/zeros_like_155, gradients/zeros_like_156, gradients/zeros_like_157, gradients/zeros_like_158, gradients/zeros_like_159, gradients/zeros_like_160, gradients/zeros_like_161, gradients/zeros_like_162, gradients/zeros_like_163, gradients/zeros_like_164, gradients/zeros_like_165, gradients/zeros_like_166, gradients/zeros_like_167, gradients/zeros_like_168, gradients/zeros_like_169, gradients/zeros_like_170, gradients/zeros_like_171, gradients/zeros_like_172, gradients/zeros_like_173, gradients/zeros_like_174, gradients/zeros_like_175, gradients/zeros_like_176, gradients/zeros_like_177, gradients/zeros_like_178, gradients/zeros_like_179, gradients/zeros_like_180, gradients/zeros_like_181, gradients/zeros_like_182, gradients/zeros_like_183, gradients/zeros_like_184, gradients/zeros_like_185, gradients/zeros_like_186, gradients/zeros_like_187, gradients/zeros_like_188, gradients/zeros_like_189, gradients/zeros_like_190, gradients/zeros_like_191, gradients/zeros_like_192, gradients/zeros_like_193, gradients/zeros_like_194, gradients/zeros_like_195, gradients/zeros_like_196, gradients/zeros_like_197, gradients/zeros_like_198, gradients/zeros_like_199, gradients/zeros_like_200, gradients/zeros_like_201, gradients/zeros_like_202, gradients/zeros_like_203, gradients/zeros_like_204, gradients/zeros_like_205, gradients/zeros_like_206, gradients/zeros_like_207, gradients/zeros_like_208, gradients/zeros_like_209, gradients/zeros_like_210, gradients/zeros_like_211, gradients/zeros_like_212, gradients/zeros_like_213, gradients/zeros_like_214, gradients/zeros_like_215, gradients/zeros_like_216, gradients/zeros_like_217, gradients/zeros_like_218, gradients/zeros_like_219, gradients/zeros_like_220, gradients/zeros_like_221, gradients/zeros_like_222, gradients/zeros_like_223, gradients/zeros_like_224, gradients/zeros_like_225, gradients/zeros_like_226, gradients/zeros_like_227, gradients/zeros_like_228, gradients/zeros_like_229, gradients/zeros_like_230, gradients/zeros_like_231, gradients/zeros_like_232, gradients/zeros_like_233, gradients/zeros_like_234, gradients/zeros_like_235, gradients/zeros_like_236, gradients/zeros_like_237, gradients/zeros_like_238, gradients/zeros_like_239, gradients/zeros_like_240, gradients/zeros_like_241, gradients/zeros_like_242, gradients/zeros_like_243, gradients/zeros_like_244, gradients/zeros_like_245, gradients/zeros_like_246, gradients/zeros_like_247, gradients/zeros_like_248, gradients/zeros_like_249, gradients/zeros_like_250, gradients/zeros_like_251, gradients/zeros_like_252, gradients/zeros_like_253, gradients/zeros_like_254, gradients/zeros_like_255, gradients/zeros_like_256, gradients/zeros_like_257, gradients/zeros_like_258, gradients/zeros_like_259, gradients/zeros_like_260, gradients/zeros_like_261)]] [Op:PartitionedCall]
解决方案
我也遇到过同样的问题，目前这个问题似乎没有明确的答案。您可以查看 Tensorflow 的 github 专用问题的演变： https ://github.com/tensorflow/tensorflow/issues/20369
但也有人指出，这可能与int32tensorflow 中的某些数据类型（例如 ）不是由 GPU 处理而是由 CPU 处理有关。因此，可以通过将有问题的变量的类型更改为uint32or来解决您的问题int64。
推荐阅读
- ios - 如何使 pdfView 缩放=100% 以适应屏幕尺寸
- encryption - 建立https连接，解密消息
- html - Bootstrap - 具有响应宽度的自动大小列
- java - 从字节中读取高位和低位在 Java 中未按预期工作
- ajax - Yii2. 使用 Captcha + Ajax 时如何避免“错误代码”？
- google-data-studio - 在谷歌数据工作室的 1 个字段中替换多个 ID
- android - How to hide the view of last clicked item on the recyclerview
- javascript - 将拼写检查属性设置为 false 后删除下划线
- javascript - React：基于 embla carousel 索引的数组的唯一 ID 或类名在样式化的组件中做出反应
- python - 通过合并替换熊猫数据框中的NAN