Free GPU credits for Fast.ai Courses


#62

It also could be that snark jupyter fails to bind to the port because it’s already busy. Was there any error/warning message displayed before

Copy/paste this URL into your browser when you connect for the first time,
to login with a token:
http://(52246a307858 or 127.0.0.1):8888/?token=2283e2d20e1d2e80655062db782a17273b58faab458a4695

(kojima) #63

No web browser found: could not locate runnable browser


#64

Hey @kojima are you running snark jupyter on any cloud server or docker container? If that’s the case you need some port tunneling for it to work. Reference: Opening Jupyter notebook: server not responding


(zhou fan) #65

why I always get a 403


(Thalanayar Muthukumar) #66

Where can I get the code? I am an online student


(Davit) #67

Sure, could you please email us at support@snark.ai or write on intercom messenger to help you fix the problem.


#68

The promo code is FastAI2018


(Igor Kasianenko) #69

Thanks for this product!
I’m going to run it from MacBook, it was never used for coding.
To begin with I installed python following this instructions https://docs.python-guide.org/starting/install3/osx/

Then I followed steps in main post and got notebook up and running!
Proceeding to dl1 lesson4-imdb notebook, I was going over cells and faced this issue

!pip install spacy && python -m spacy download en

Requirement already satisfied: spacy in /opt/conda/envs/fastai/lib/python3.6/site-packages (2.0.11)
Requirement already satisfied: numpy>=1.7 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (1.14.5)
Requirement already satisfied: murmurhash<0.29,>=0.28 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (0.28.0)
Requirement already satisfied: cymem<1.32,>=1.30 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (1.31.2)
Requirement already satisfied: preshed<2.0.0,>=1.0.0 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (1.0.1)
Requirement already satisfied: thinc<6.11.0,>=6.10.1 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (6.10.2)
Requirement already satisfied: plac<1.0.0,>=0.9.6 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (0.9.6)
Collecting pathlib (from spacy)
  Downloading https://files.pythonhosted.org/packages/ac/aa/9b065a76b9af472437a0059f77e8f962fe350438b927cb80184c32f075eb/pathlib-1.0.1.tar.gz (49kB)
    100% |################################| 51kB 3.6MB/s ta 0:00:011
Requirement already satisfied: ujson>=1.35 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (1.35)
Requirement already satisfied: dill<0.3,>=0.2 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from spacy) (0.2.8.2)
Collecting regex==2017.4.5 (from spacy)
  Downloading https://files.pythonhosted.org/packages/36/62/c0c0d762ffd4ffaf39f372eb8561b8d491a11ace5a7884610424a8b40f95/regex-2017.04.05.tar.gz (601kB)
    100% |################################| 604kB 7.2MB/s eta 0:00:01
Requirement already satisfied: wrapt in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (1.10.11)
Requirement already satisfied: tqdm<5.0.0,>=4.10.0 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (4.23.4)
Requirement already satisfied: cytoolz<0.9,>=0.8 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (0.8.2)
Requirement already satisfied: six<2.0.0,>=1.10.0 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (1.11.0)
Requirement already satisfied: termcolor in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (1.1.0)
Requirement already satisfied: msgpack-python in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (0.4.8)
Requirement already satisfied: msgpack-numpy==0.4.1 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from thinc<6.11.0,>=6.10.1->spacy) (0.4.1)
Requirement already satisfied: toolz>=0.8.0 in /opt/conda/envs/fastai/lib/python3.6/site-packages (from cytoolz<0.9,>=0.8->thinc<6.11.0,>=6.10.1->spacy) (0.9.0)
Building wheels for collected packages: pathlib, regex
  Running setup.py bdist_wheel for pathlib ... done
  Stored in directory: /home/kasianenko/.cache/pip/wheels/f9/b2/4a/68efdfe5093638a9918bd1bb734af625526e849487200aa171
  Running setup.py bdist_wheel for regex ... done
  Stored in directory: /home/kasianenko/.cache/pip/wheels/75/07/38/3c16b529d50cb4e0cd3dbc7b75cece8a09c132692c74450b01
Successfully built pathlib regex
distributed 1.22.0 requires msgpack, which is not installed.
twisted 18.7.0 has requirement Automat>=0.3.0, but you'll have automat 0.0.0 which is incompatible.
Installing collected packages: pathlib, regex
Could not install packages due to an EnvironmentError: [Errno 13] Permission denied: '/opt/conda/envs/fastai/lib/python3.6/site-packages/pathlib.py'
Consider using the `--user` option or check the permissions.

You are using pip version 10.0.1, however version 18.0 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.

EnvironmentError: [Errno 13] Permission denied: ‘/opt/conda/envs/fastai/lib/python3.6/site-packages/pathlib.py’

Taking in consideration that spacy is package that used in dl1 and dl2 courses, it might be good to have it pre-installed.

I went to dl2 imdb notebook and found out that the only problem is access rights

    ---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
<ipython-input-3-584ff827d0cb> in <module>()
      1 CLAS_PATH=Path('data/imdb_clas/')
----> 2 CLAS_PATH.mkdir(exist_ok=True)
      3 
      4 LM_PATH=Path('data/imdb_lm/')
      5 LM_PATH.mkdir(exist_ok=True)

/opt/conda/envs/fastai/lib/python3.6/pathlib.py in mkdir(self, mode, parents, exist_ok)
   1244             self._raise_closed()
   1245         try:
-> 1246             self._accessor.mkdir(self, mode)
   1247         except FileNotFoundError:
   1248             if not parents or self.parent == self:

/opt/conda/envs/fastai/lib/python3.6/pathlib.py in wrapped(pathobj, *args)
    385         @functools.wraps(strfunc)
    386         def wrapped(pathobj, *args):
--> 387             return strfunc(str(pathobj), *args)
    388         return staticmethod(wrapped)
    389 

FileNotFoundError: [Errno 2] No such file or directory: 'data/imdb_clas'

Do you have to manually activate every new user?


(Igor Kasianenko) #70

I have a question: is it possible to select what GPU will be used? I’d prefer to run on P106 since half of my time I read code and manipulate with data.

And a suggestion for improvement: command snark stop requires pod ID. I assume that most of fastai students will run only 1 instance and it is not best user experience every time to run snark ls copy paste podID before stopping. I suggest adding keyword like snark stop all to terminate all running instances.


#71

Hi @Kasianenko you have sudo access, so running the pip install with sudo should resolve the permission denied problem. We do not manually activate new users. It’s supposed to be all automatic. For dl2 notebook, the error message says there’s no data folder — we haven’t pre-download datasets for dl2 but we’re working on it. In the meantime, you can download the dataset yourself as a remedy. Sorry for the inconvenience!


#72

@Kasianenko Thanks for the suggestion! It totally makes sense. We’ll implement these in the next package release. It’s P106 by default and You can select GPU by $snark start -s P106 or $snark start -s 1080


(Phil Weslow) #73

Hello @snarkai,

I just went through the steps you mentioned at the top of the thread. I created my account, then from my Mac terminal:

pip3 install snark
snark login

Which results in output which contains the following:

Welcome to Snark AI pod!
Pod Type: 1x NVIDIA P106
You have full SUDO access.

So, I was able to login without issues. However, when I try to launch the Jupyter NB, I get the following message:

snark start --pod_type fast.ai --jupyter
-bash: snark: command not found

When I login to my account via the browser, I notice that the GPU time I have available continues to run down. Since I cannot stop run snark commands after logging in, I cannot stop the pod via the command line. In addition to whatever command line issues are present, it would be great if the browser-based GUI offered a way to shut down a running pod.

I am guessing that I probably have to update my $PATH variable, but am not sure how to do so to get Snark running.


(Kalyan D) #74

Is the command option to specify GPU “-s” or “-g”? -g says GPU counter.


(Kalyan D) #75

I am able to launch pod with P106 GPU but when I try and specify 1080 I get the error message:

snark start --pod_type fast.ai --jupyter -s '1080'
Setting up the pod...
Error: Couldn't successfully schedule pod execution. Please try again

Is this because of infrastructure issue? Online it shows I have 42 hrs left on 1080.


#76

Unfortunately we run out of GTX 1080 supply at the moment. Can you try with P106?


#77

Hi @P_Wes can you please try running snark start --pod_type fast.ai --jupyter outside of pod, on your computer? I guess you were trying to run that after logging into the pod? And yes we will implement better web-based interface. Thanks for the suggestion!


#78

Right sorry about the typo. Corrected.


(Phil Weslow) #79

@snarkai, your suggestion of running snark start --pod_type fast.ai --jupyter from another terminal window gets me one step further:

Setting up the pod...
Error: Free users are limited to run only one pod at the same time. 
 - To attach to your previous pod please run: snark attach pod_*****
 - To stop the pod please run: snark stop pod_*****
bash-3.2$ snark attach pod_*****
Connecting to the pod...

However, from there, when I try to launch jupyter, I am unable to from either inside the pod or from an alternate terminal window.

From inside the pod:

snark start --jupyter
-bash: snark: command not found

From outside:

snark start --jupyter
Setting up the pod...
Error: Free users are limited to run only one pod at the same time. 
 - To attach to your previous pod please run: snark attach pod_*****
 - To stop the pod please run: snark stop pod_*****

Update:

When I tried shutting down all running pods and logging in again:

$snark login
$snark start --pod_type fast.ai --jupyter

I was able to get Jupyter NB up and running.


#80

Thanks @snarkai, I applied the promo code and pulled some Tensorflow tutorials (specifically this CNN notebook) to try. While running I kept hitting the error - “The kernel appears to have died. It will restart automatically.” and I couldn’t complete the training.

A visit to nvidia-smi shows that the RAM is near maximum capacity.

11%20PM

May I know how I can get this tutorial to run without kernel restarting? Thank you.


#81

Thank you for the free credits! :slightly_smiling_face:

When I start a new pod and do $snark ls, I noticed that my GPU spec says P106 but when I check the usage on the Snark AI dashboard it appears to be reducing the time for the 1080 GPU. Could you please look into it?