Skip to content
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

Error on TPU Pod #596

Closed
rd16395p opened this issue Nov 7, 2019 · 15 comments
Closed

Error on TPU Pod #596

rd16395p opened this issue Nov 7, 2019 · 15 comments

Comments

@rd16395p
Copy link

rd16395p commented Nov 7, 2019

I am trying to connect to my TPU Pod using the documentation found here and the example in the TPU github for MNIST.

https://cloud.google.com/tpu/docs/training-on-tpu-pods

However, when I try to connect, I receive the following error;
[I1107 14:24:47.276069 140521844594496 transport.py:157] Attempting refresh to obtain initial access_token
[W1107 14:24:47.301813 140521844594496 http.py:118] Invalid JSON content from response: b'{\n "error": {\n "code": 403,\n "message": "Request had insufficient authentication scopes.",\n "status": "PERMISSION_DENIED"\n }\n}\n'

Does anyone know how to fix this issue? I have ran gcloud init to ensure I am using my account.

@CyFeng16
Copy link

CyFeng16 commented Nov 8, 2019

It is that your GCE instance does not have the right to access TPU domain.
Just Check the GCE instance's configuration and give it all necessary access right needed.

@CyFeng16
Copy link

CyFeng16 commented Nov 8, 2019

@rd16395p You may want to read this.

@rd16395p
Copy link
Author

rd16395p commented Nov 8, 2019

Yes that was the issue, I changed it via editing the VM. Cheers!

@rd16395p rd16395p closed this as completed Nov 8, 2019
@CyFeng16
Copy link

CyFeng16 commented Nov 9, 2019

@rd16395p Congrats and enjoy the TPUs.

@manhlab
Copy link

manhlab commented Jun 14, 2020

please tell me how you fixed this. I tried it but not success !

@rd16395p
Copy link
Author

rd16395p commented Jun 14, 2020

please tell me how you fixed this. I tried it but not success !

@manhlab

You need to edit the VM to have permissions to do the task.

See "Changing the service account and access scopes for an instance" here:
https://cloud.google.com/compute/docs/access/create-enable-service-accounts-for-instances

@manhlab
Copy link

manhlab commented Jun 14, 2020

sorry but I already have this
Screenshot_20200614_142531

@manhlab
Copy link

manhlab commented Jun 14, 2020

I try to create a new key. But I think all think is already but don't know Why not running TPU?

@rd16395p
Copy link
Author

@manhlab Did you try fixing the VM?

Do you have a TPU instance running?

@manhlab
Copy link

manhlab commented Jun 14, 2020

what i did: run VM, run TPU
export TPU_NAME=****

@rd16395p
Copy link
Author

@manhlab

  1. Create a VM with those settings I link to
  2. Create a TPU and get the IP to connect to. You can find this in the details of it.
  3. Use the following code ; and follow along to " Get TPU Address";
    https://blog.tensorflow.org/2019/01/keras-on-tpus-in-colab.html

@manhlab
Copy link

manhlab commented Jun 14, 2020

thanks for your advince. but i am running code on colab. all working with my code. but don't know how to link TPU GC and VM instance. This is first time of me

@manhlab
Copy link

manhlab commented Jun 14, 2020

i still cant fix this. please help me

@rd16395p
Copy link
Author

@manhlab Show me your connection string and I can help you edit it.

@rd16395p
Copy link
Author

@manhlab https://github.com/rd16395p/TPU-Pod-Examples/tree/master/iris%20example
This code may help you. I have instructions on how to use it as well.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants