环境************

Paddle version: 2.0.0-rc0
Paddle With CUDA: True
OS: Windows 10
Python version: 3.6.5
CUDA version: 10.1.243
cuDNN version: None.None.None
Nvidia driver version: 441.87

-代码如下：****************

from paddle.inference import Config
from paddle.inference import Predictor
import paddle
from paddle.vision.datasets import MNIST
import paddle.vision.transforms as trans
import numpy as np
from paddle.io import DataLoader
import matplotlib.pyplot as plt

%%

def draw_imgs(imgs):

print(imgs.shape)

imgs=imgs.reshape(imgs.shape[0],imgs.shape[2],imgs.shape[3])
print("draw_imgs--shape:{}".format(imgs.shape))
for i in range(imgs.shape[0]):
plt.subplot(2,4,i+1)
plt.imshow(imgs[i],cmap="gray")
plt.show()
#%%
train_ds=MNIST(mode="train")
train_loader=DataLoader(train_ds,batch_size=1)

train_it=iter(train_loader)
data=next(train_it)
imgs,labels=data

%%

print(imgs)
print(labels)

%%

draw_imgs(imgs.numpy())

%%

imgs
#%%
model_path="./mnist/"
config=Config(model_path+"mnist.pdmodel",model_path+"mnist.pdiparams")
predictor=Predictor(config)
input_names=predictor.get_input_names()
input_tensor=predictor.get_input_handle(input_names[0])

input_tensor.reshape([1,1,28,28])
input_tensor.copy_from_cpu(imgs) #此行报错
predictor.run()

output_names=predictor.get_output_names()
output_tensor=predictor.get_output_handle(output_names[0])
output_data=output_tensor.copy_to_cpu()
print(output_data)

报错：*************

UnimplementedError: Invalid dimension to be accessed. Now only supports access to dimension 0 to 9, but received dimension is 32. (at D:\2.0.0-rc0\paddle\paddle/fluid/framework/ddim.h:54)

！上面用到的模型为以下代码生成：***********************

#%%

import paddle
import paddle.nn as nn
import paddle.vision as vision
import paddle.metric as metric
import paddle.vision.transforms as trans
import paddle.optimizer as opt
import numpy as np
import matplotlib.pyplot as plt
from paddle.io import DataLoader
import paddle.nn.functional as F
import paddle.static
import paddle.jit

%%

transform=trans.Normalize(mean=[0.0],std=[1.0]) #mean /std的维度与数据维度一致
train_ds=vision.datasets.MNIST(mode="train",transform=transform)
test_ds=vision.datasets.MNIST(mode="test",transform=transform)

%%

Batch_size=64
epochs=5
LR=0.001

%%

train_loader=DataLoader(train_ds,batch_size=Batch_size,shuffle=True)
test_loader=DataLoader(test_ds,batch_size=Batch_size,shuffle=False)

%%

class MyNet(nn.Layer):
definit(self):
super(MyNet,self).init()
self.flatten=nn.Flatten()
self.fc1=nn.Linear(784,1024)
self.fc2=nn.Linear(1024,512)
self.fc3=nn.Linear(512,10)

@paddle.jit.to_static(input_spec=[paddle.static.InputSpec(shape=[None,1,28,28],name="inputs"),])  #注意书写格式
def forward(self, inputs):
    out=self.flatten(inputs)
    out=F.relu(self.fc1(out))
    out=F.relu(self.fc2(out))
    out=self.fc3(out)
    return out

%%

model=MyNet()

model.train()

optimizer=opt.Adam(learning_rate=LR,parameters=model.parameters())
loss_fun=nn.CrossEntropyLoss()

%%

for epoch in range(1,epochs+1):
correct_count=0
for idx,data in enumerate(train_loader):
inputs,labels=data
y_preds=model(inputs)
loss=loss_fun(y_preds,labels)
y_preds=y_preds.numpy()
y_max=np.argmax(y_preds,axis=1)
y_max=y_max.reshape(y_max.shape[0],1)
correct_count+=(y_max==labels.numpy()).sum()
loss.backward()
optimizer.step()
optimizer.clear_grad()
if idx%100==0:
print("epoch:%1d idx:%4d loss:%.4f"%(epoch,idx,loss.numpy()))
print("epoch:%1d Accuracy:%.2f"%(epoch,correct_count/len(train_ds)*100))

%%

paddle.jit.save(model,"./mnist/mnist")

【注意】*******************

DataLoader中batchsize设置不生效，总是使用32的batchsize，也请确认