11 Replies Latest reply on May 28, 2018 3:45 AM by Intel Corporation

    job submission problem

    munendra

      when i am submitting the job at that job has submitted but when i am doing qstat then at that time it is not showing any job id

        • 1. Re: job submission problem
          Intel Corporation
          This message was posted on behalf of Intel Corporation

          Hi,

          qstat provides the status of jobs that are running/queued.

          Since, you have given a simple echo command, your job would have been executed instantly. You can check for the corresponding(jobid) error and output files in the current folder.


          Hope this helps.

          Thanks & Regards,
          Sandhiya

          • 2. Re: job submission problem
            munendra

            when i run the same file in jupyter notebook with batch size = 1 then it is running but expected time for single epoch is around 30 hrs, if i increase the batch size then it gives kernel dead, then i tried it using "qsub" then in STDIN.e86408 file it gives Permission denied. what should i do??

            • 3. Re: job submission problem
              Intel Corporation
              This message was posted on behalf of Intel Corporation

              Hi,

              Kindly provide the following details:

              1. Are you trying to execute your experiment from the Login node or the Compute node?
              2. If possible, could you please share your code so that we can try out from our end.
              Thanks & Regards,
              Sandhiya
              • 4. Re: job submission problem
                munendra

                my training dataset size is approximately 25000 images

                and i have attached a file which shows the number of trainable and non-trainable parameters

                 

                from jupyter notebook i simply login and run using ctrl+enter  i think this is login node,

                i don't know how to use compute node and the permission denied error is given when i run the file from jupyter notebook terminal using echo filename | qsub

                • 5. Re: job submission problem
                  Intel Corporation
                  This message was posted on behalf of Intel Corporation
                  Hi Munendra,

                   

                  1. Firstly, you are directly working on the compute node since you launched the terminal from Jupyter notebook.
                  2. You are getting the Permission denied error, since you are trying to access the file from the root. i.e When you refer the complete path using the '~' symbol, you get the error.
                  3. echo is not the right way to execute your .ipynb file. Kindly convert your file to a python file using the command:

                   

                  ipython nbconvert --to python CapsNet_Keras.ipynb

                   

                  and run the create python file as below:

                   

                  python <CREATED_PYTHON_FILE.PY> | qsub

                   

                  For your case, python CapsNet_Keras.py | qsub

                   


                  You can refer the following links for any issues regarding submitting a job:

                   

                   

                  Your kernel gets killed faster when you launch the terminal from Jupyter notebook. Since you are having a large dataset, I would recommend you to login using ssh with the help of putty or Linux terminal (Steps are available in the link provided with the welcome mail of DevCloud).

                   

                  Also, you can refer the following threads, for more information on using Jupyter notebook Terminal

                   

                   

                  Hope this helps!

                   

                   

                  Thanks & Regards,
                  Sandhiya
                  • 6. Re: job submission problem
                    munendra

                    Hi Sandhiya,

                     

                        after converting to .py file when i tried to run it by login using ssh then it gives error of  "No module named keras" then i try to install keras using pip but it gives error.

                        How to install keras??

                    • 7. Re: job submission problem
                      Intel Corporation
                      This message was posted on behalf of Intel Corporation
                      Hi,
                       
                      You can install keras in either of the two ways:
                       
                      1.  Use the command: pip install keras --user.
                       
                      In general, you can use the command pip install <PACKAGE_NAME> --user
                       
                      2.  Create a virtual environment:
                      conda create -n <your_env_name> python=<PYTHON_VERSION>
                      source activate <your_env_name>

                       
                      Then do : conda install <PACKAGE_NAME>

                      I would recommend the 2nd method since it provides more modularity.

                      Regards,
                      Sandhiya
                      • 8. Re: job submission problem
                        munendra

                        Hi Sandhiya,

                         

                        i have tried to install keras using conda as you said but after installing few packages the process got killed

                        i tried these two commands

                        1. conda install keras

                        2. conda install -c conda-forge keras

                        but both these commands did not work, the process killed in between

                         

                        conda create -n <env_name> intelpython3_full python=3  run successfully

                        • 9. Re: job submission problem
                          Intel Corporation
                          This message was posted on behalf of Intel Corporation

                          Hi,

                          Could you please try again installing keras in your activated conda environment since the issue you faced may be due to session time out. We tried from our end and didn't face any issues. If you again face the issue please let us know.

                          Regards,
                          Dona.

                          • 10. Re: job submission problem
                            munendra

                            Hi,

                             

                            i have installed keras successfully but i am not able to run the .py file , as you suggested earlier by converting from .ipynb file to .py file to run using ssh.

                            I am attaching the STDIN.o87674 file

                            • 11. Re: job submission problem
                              Intel Corporation
                              This message was posted on behalf of Intel Corporation

                              Hi,

                              As your initial job id problem got resolved, please open a new thread for current issue. This helps other users for their future reference. Kindly share the STDIN.e87674  with  the new thread.

                              Regards,
                              Dona