[
https://issues.apache.org/jira/browse/SINGA-397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16661941#comment-16661941
]
Junyou Chen commented on SINGA-397:
-----------------------------------
Hi Ngin!
I have another problem about making predictions. The attached is a screen shot
for that. !2018-10-24 13-03-56 的屏幕截图.png!
Also, sometimes I run client-usage.py and find out that the training process
lasts for a very long time and gives out no error information. I tried to
attribute this to the poor network connecting. Would it be okay for me to
download the datasets(MNIST) on my local machine and directly use it in
training process?( Such as change the original code(in client-usage.py, line14):
{color:#ff0000}TRAIN_DATASET_URI =
'[https://github.com/cadmusthefounder/mnist_data/blob/master/output/fashion_train.zip?raw=true'|https://github.com/cadmusthefounder/mnist_data/blob/master/output/fashion_train.zip?raw=true%27]{color}
to the new code
TRAIN_DATASET_URI = ‘home/mnist/train_data/XXXXX’ )
By the way, it seems that
{color:#ff0000}[https://github.com/cadmusthefounder/mnist_data/blob/master/output/fashion_train.zip?raw=true'|https://github.com/cadmusthefounder/mnist_data/blob/master/output/fashion_train.zip?raw=true%27]{color}
{color:#ff0000}{color:#333333}doesn't links to the original dataset of
fashion-MNIST. What are the difference between them?{color} {color}
> Deployment Issues for Rafiki
> ----------------------------
>
> Key: SINGA-397
> URL: https://issues.apache.org/jira/browse/SINGA-397
> Project: Singa
> Issue Type: Test
> Environment: ubuntu 18, local machine
> Reporter: Junyou Chen
> Priority: Major
> Attachments: 1.png, 2.png, 2018-10-24 13-03-56 的屏幕截图.png
>
>
> After I deployed Rafiki on my own machine by following the Quick Setup
> section in the Developer Guide, I tried to run the $client-usage.py$ but
> failed(as picture 1 shows. system halted with no feedback). Also, I followed
> several examples in the User Guide, but there are still some problems(shown
> in picture 2).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)