-
Notifications
You must be signed in to change notification settings - Fork 943
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
I will train VDSR on the tensorflowOnSpark,which data should I will use? #74
Comments
@bobo2001281 You have several options:
|
In the first way: How can I ship my data? There doesnot seem to have the label data in my data. |
Shipping the data to the executors can be done with |
now
17/05/03 16:39:30 INFO yarn.Client: Application report for application_1493036076768_0092 (state: ACCEPTED) Container exited with a non-zero exit code 1 |
Please grab the yarn logs via: |
hadoop@u10-121-135-150:~/hadoop-2.7.1/logs$ yarn logs -applicationId application_1493036076768_0106 The log is not exist both in the middle of the running of the spark submit and after the command. |
I have modified the yarn-site.xml with below and now I can see my logs.
|
😢 |
17/05/05 11:40:07 INFO memory.MemoryStore: MemoryStore started with capacity 14.2 GB |
but when I python VDSR.py , and no error occured. |
It looks like you haven't installed tensorflow into your Python distribution that is shipped to the executors via
|
I install tensorflow in my local env.
How can I install tensorflow into my Python distribution? hadoop@u10-121-135-150:~$ Python/bin/python
|
In the instruction: Install and compile TensorFlow w/ RDMA Support git clone git@github.com:yahoo/tensorflow.git
In the last line,How can I install tensorflow into Python? |
Actually, if you do not need RDMA support, you should be able to just run something like: If you need specific versions (e.g. Python 2.7 vs. 3.x, CPU/GPU, etc), you can adapt these instructions from TensorFlow |
When I execute python VDSR.py and cost 10s . |
2017-05-11 20:46:35,619 INFO org.apache.hadoop.yarn.server.resourcemanager.rmapp.attempt.RMAppAttemptImpl: appattempt_1494227397798_0029_000001 State change from SCHEDULED to ALLOCATED_SAVING |
Those logs don't reveal much... Please grab the yarn application logs and look for errors on the executors. |
If I cancel (Ctrl + C)the application that is running,where the application log will saved? |
You will need to do the following:
|
hadoop@u10-121-135-150:~$ Python/bin/python
[4]+ Stopped Python/bin/python
What is the difference of the above 2 tensorflow s? |
I have install scipy in Python that is distributed to the Spark executors. |
A couple notes:
So, with that all said, I'd recommend picking the versions of TensorFlow and Python that you wish to move forward with, then create a custom python distribution (with all necessary dependencies), and then use ONLY this distribution to test your code going forward (for "local" and "distributed" testing). |
origin resource is here
https://github.com/Jongchan/tensorflow-vdsr
1.the data is MATLAB 5.0 MAT-file.
according the https://github.com/yahoo/TensorFlowOnSpark/wiki/Conversion,I change the code.
but on condition that the data is valid for tensorflowOnSpark to use on HDFS.
2.Need I change the data to TFRecords?
The text was updated successfully, but these errors were encountered: