Skip to main content

Common queries that Upgrad students might have

How to download or upload datasets in the JarvisCloud platform?

We provide you with a customized Linux machine with several popular libraries preloaded. So you can do most of the things that you do with a Linux machine.

You interact with the platform through JupyterLab which is a popular open-source tool used across the data science communities. Most of the common functionalities that you would need like downloading, uploading, and so many others are well documented in their official documentation.

Disk is not getting freed up even after deleting files.

The data you delete from Jupyter goes to the Trash folder. You can run the below command to delete any files in the Trash

rm -rf ~/.local/share/Trash/*

Ohh My disk is full!

Every instance is allocated with 20GB of storage. All Linux machines start creating various problems if the disk gets filled. So avoid filling the storage.

When I spoke to some of you, I realized the code you are executing is saving the model weights for each epoch. You can avoid that by changing the save_best_only to True in the ModelCheckpoint callback. Doing so will result only in the best model weights getting saved and also not utilizing the full disk.

What do I do when the disk is full

A simple way to do this is to delete the given folder from the terminal using

rm -rf 'folder_name'

JupyterLab stopped working

We launched a new feature a few weeks back which will pause any instances running longer than 4 hours. If you do not want the feature, then you can disable it. Click Auto Pause from the user menu available on the top right corner of the cloud.jarvislabs.ai website.

Why launching a machine is taking longer or slow

Sometimes resuming an instance takes longer than usual. Which is about 4 to 10 minutes. We believe we should convey this in a better way on the UI, we will do that in the coming weeks. This problem is not specific to Jarvislabs.ai but also to other major cloud providers.

My program takes a lot of time to run

We observed this issue with some of the students and it is caused due to installing extra libraries. Tensorflow has 2 variants TF1 and TF2. Installing them together causes a lot of conflicts. And there is also a very high chance that your code will stop utilizing GPU. Fixing this could be a nightmare. Hence it is strongly advised that you do not install any variants of TF. You can always open a terminal and check the GPU utilization to ensure that the code is using the GPU.

watch nvidia-smi

The instance fails to launch

Very rarely the instance fails to launch and it is not recoverable. From our earlier experiences, we observed that it can be caused due to the below reasons.

  • Disk space is full and hence instance launch gets failed.
  • Installation of conflicting libraries can prevent the instance from getting launched.

So please ensure utmost care, when installing additional libraries. Also, avoid filling up the disk.

Unable to import Keras

On the internet, you can find 2 different code versions of TF. TF2 comes preloaded on the instance, so the code you use should be in TF2. The UpGrad team has provided various sample codes, take a look at how a particular code is called.

For example:

In TF2 - Valid Code

from tensorflow.keras.models import Sequential

In TF2 - In Valid Code will not work

from keras... import ...

Unable to save files

Please check if you are saving it to a read-only folder (/datasets is a read-only folder) or if the instance is paused.

What are the folders that come at the launch of an instance

You get 2 types of folders.

  • Datasets is a read-only folder that contains all the data provided by UpGrad for the course. The size of the folder is not considered in your overall storage.
  • Dl_content contains the code provided by UpGrad required for the course.

Both of these folders come every time you launch a new instance.

GPU is not being utilized

GPU may not be utilized for several reasons. Check with a notebook that works. We would recommend running this notebook

/dl_content/Upgrad DL//dl_content/Upgrad DL/Transfer Learning/Transfer_Learning_Notebook.ipynb

to observe the GPU utilization.

  1. If the GPU is not being utilized and
  2. If you have not done extra installations Then something went wrong with the instance. This occurrence is highly unlikely and has not occurred in the past.

If the GPU is being utilized, then you have to tweak your code. If you need help in that, you should reach out to Upgrad TA.

When to reach us

At Jarvislabs.ai we provide the platform required for you to practice DL skills. If you face any challenges with

  • Launching a new instance
  • Starting Jupyter Lab
  • Pushing the instance
  • Deleting the instances

Then please reach out to us. In simple words, you can reach to us to ask about the platform or anything in general.

When to reach UpGrad support

Although all the JarvisLabs.ai team members have varying levels of knowledge in Data Science and Deep Learning, we are neither specialized in the TensorFlow framework nor aware of the course content nor that is being offered by UpGrad.

So many questions like

  • Why some code is not working?
  • I am unable to import Keras
  • Any TF-related issues.

are best answered by experts available at UpGrad.