Platform: GCP ✅

pnvijay · November 8, 2018, 11:08am

Hi,

When I do this step, I get to see my kernel. But when I try fastai.version it gives me the latest version number. the steps are as follows:

conda env create -f environment.yml (‘environment.yml’ is the file in the old fastai for creation of fastai GPU environment)
This creates an environment called fastai
Then I do “python -m ipykernel install --user --name fastai”
I can see an environment called fastai. I open a notebook in this environment and type “import fastai; print(fastai.version)”
I can see version as 1.0.18

noskill · November 8, 2018, 11:32am

Hey,
Why don’t you try
pip install fastai==0.7??
This will give 0.7 version.

xtra_xtra_medium · November 8, 2018, 1:46pm

I just upgraded my account, but am getting the same error.

Error: The zone ‘projects/divine-outlet-218301/zones/us-west1-b’ does not have enough resources available to fulfill the request. Try a different zone, or try again later.

@adpostma thanks for the suggestion, though. It’s going on three days now that I’ve been unable to spin up my GCP instance. I was hoping that this would sort of resolve itself. It sounds like others are having similar problems (@NandoBr), even after switching zones. I’m still relatively new to all of this, so any advice would be much appreciated.

pnvijay · November 8, 2018, 1:52pm

How to use that in a different conda environment? Also how can I ensure that it does not corrupt my existing V1 in the base and install two versions of pytorch.

Mwai_Newton · November 8, 2018, 2:57pm

Hi, please help out. I can no longer start my instance, when try it says:

Starting VM instance “my-fastai-instance” failed. Error: Quota ‘GPUS_ALL_REGIONS’ exceeded. Limit: 0.0 globally.

arunoda · November 8, 2018, 2:59pm

search this thread. You can see the solution.

Mwai_Newton · November 8, 2018, 3:05pm

Thanks!

Rares · November 8, 2018, 5:13pm

@Czechcheck thank you for the credit!

Madhur · November 8, 2018, 5:55pm

Thanks a lot for the credit

adpostma · November 8, 2018, 6:41pm

Thanx Czechneck. Much appreciated!

kkthai · November 9, 2018, 12:17am

Hi Allan, I’ve tried your instruction but it did not work on my end. I was frustrated and decided to delete my instance and create another one. I am able to open my notebook now. Thanks for your help!

eosoba · November 9, 2018, 10:16pm

Is anyone else having issues with the kaggle api? I am getting 404 - Not Found errors.

miwojc · November 9, 2018, 11:44pm

Did you copy kaggle.json file to .kaggle folder and accept competition rules at kaggle website?

shub.chat · November 10, 2018, 12:26am

Anyone recently observed an increase in time to train an epoch on a GCP instance?By training time per epoch almost shot up by 100%.Have been facing this since past 12 hrs.

gamo · November 10, 2018, 12:32am

If you are new to kaggle you need to verify your account and then sign up for the competition to be able to access the data.

paul · November 10, 2018, 3:13am

Unfortunately, I’m getting the error that several others in this thread have also encountered: “The zone ‘…’ does not have enough resources available to fulfill the request. Try a different zone, or try again later.” I’ve tried to restart the instance at many different times over the last 3 hours without success. I read on the “Introducing fastai-shell” thread by @arunoda that @cwerner also had the same problem a couple hours ago, so maybe it’s because the west coast trying to finish up work before the weekend

I tried to use Jeremy’s tip to move the instance, but I get the following error:

gcloud compute instances move my-fastai-instance
–zone us-west2-b --destination-zone us-east1-b

Moving gce instance my-fastai-instance…failed.
ERROR: (gcloud.compute.instances.move) Instance cannot be moved while in state: TERMINATED

This message is technically correct: I had left the instance on yesterday and it got terminated after 24hours. Is there any other way to access my disk from another location?

paul · November 10, 2018, 5:17am

FYI, I was able to start this instance again a couple of minutes ago. So maybe it was Friday evening load.

kgshrivastava · November 10, 2018, 8:37am

Thank you Petr Simecek! I received my GCP credit yesterday. This will enable a lot of learning experience for me on Deep Learning through experimenting and I deeply appreciate your nice gesture to help fellow learners. My best wishes!
Regards,
KrishnaGopal Shrivastava

Tejaswani · November 10, 2018, 1:36pm

I am also getting the same error. Tried creating a new instance in a different zone(us-west1-b), but that is also giving the same error.
Any help would be greatly appreciated

dreambeats · November 10, 2018, 3:50pm

This happens every now and then, perhaps give us-west2 a try? It usually works for me when west1 is down.