New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
KeyError: 'input' error #94
Comments
Hi @xiaoyongzhu, this error generally occurs when a data feeding task is assigned to the executor running the PS node. And, this can only occur if you're configured to run more than one task per executor. So, for example, you have two hosts, and you started a PS node on one executor and a worker node on the other, per:
However, the data feeding job reports:
So you will need to configure spark to run one task per executor, e.g. setting |
Hi @leewyang |
@renato2099 yes. Keep in mind that the |
Hello,
I am trying TFoS, and successfully converted the MNIST zip files into HDFS files following the link here. However, when I try to run the sample named "Run distributed MNIST training (using feed_dict)", I encountered the following error saying "KeyError: 'input' error" and I have no clue on what's going on.
My environment is:
Spark 2.1.1 + YARN/Hadoop 2.6 + latest TFoS (master branch) + latest TF (1.2.0). I am using YARN-client mode. Below is the command line I am using. I have a small cluster of 2 nodes (I remember some issue mentioned that the executor number should be better be set to the cluster node number, not quite sure though) and I am sure that libhdfs.so and libjvm.so is in the LD_LIBRARY_PATH.
Also I did a bit research and found a similar issue #32 with a couple of other issues - However since that issue was on Feb, I thought all the fixes should be merged in the latest branches. That's why I am raising this issue again.
I am also attaching the full logs Untitled-1.txt in case you need more information.
Thanks for looking at it!
The text was updated successfully, but these errors were encountered: