Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

Update new AMI Ids #3826

Merged
merged 10 commits into from
Nov 14, 2016
Merged

Update new AMI Ids #3826

merged 10 commits into from
Nov 14, 2016

Conversation

nswamy
Copy link
Member

@nswamy nswamy commented Nov 14, 2016

No description provided.

@piiswrong
Copy link
Contributor

we need a better solution than updating code everytime ...

@mli
Copy link
Contributor

mli commented Nov 14, 2016

do we need to change the ids for every update? i think we have secret weapons to update ami while keeping the id

@piiswrong
Copy link
Contributor

@mli Make it use module instead of model while you are at it?

@piiswrong
Copy link
Contributor

@mli Also I think we should put some of the logic into functions and move them into the main repo.
Currently I frequently have to copy large chunk of code like argparser, log config, kv store create, model load/save

Here is an idea:


def log_config(log_dir=None, log_file=None, prefix=None, rank=0):
    reload(logging)
    head = '%(asctime)-15s Node[' + str(rank) + '] %(message)s'
    if log_dir:
        logging.basicConfig(level=logging.DEBUG, format=head)
        if not os.path.exists(log_dir):
            os.makedirs(log_dir)
        if not log_file:
            log_file = (prefix if prefix else '') + datetime.now().strftime('_%Y_%m_%d-%H_%M.log')
            log_file = log_file.replace('/', '-')
        else:
            log_file = log_file
        log_file_full_name = os.path.join(log_dir, log_file)
        handler = logging.FileHandler(log_file_full_name, mode='w')
        formatter = logging.Formatter(head)
        handler.setFormatter(formatter)
        logging.getLogger().addHandler(handler)
        logging.info('start with arguments %s', args)
    else:
        logging.basicConfig(level=logging.DEBUG, format=head)
        logging.info('start with arguments %s', args)

@piiswrong piiswrong merged commit 599b8c1 into apache:master Nov 14, 2016
@nswamy
Copy link
Member Author

nswamy commented Nov 15, 2016

I agree, it is not feasible to keep updating everytime a new AMI is made
available. As far as I know there is no other straight forward way other
than to make it a user defined parameter and ask to the user to pickup the
AMI-Ids from S3 file which will always contain the latest AMI ids.
I think we can also make a AWS Lambda Function return the AMI-Id given a
region, I will explore that further.

Thanks for your patience.

-Naveen

On Mon, Nov 14, 2016 at 10:32 AM, Eric Junyuan Xie <notifications@github.com

wrote:

we need a better solution than updating code everytime ...


You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#3826 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/ABJxQkA9_T2kHLDoBWWf8M0mpYM2L734ks5q-KlGgaJpZM4KxpSU
.

arank pushed a commit to arank/mxnet that referenced this pull request Nov 16, 2016
* Rename deeplearning.template to deeplearning.json so that github can highlight syntax
Update to the latest AMIs, tie the Master and Workers to the same AvailabilityZone to reduce the network latency
Add detailed Readme to setup a Deep Learning CloudFormation stack

* Update Authors in Readme.md

* Update Authors in Readme.md

* rename deeplearning.json to deeplearning.template

* rename deeplearning.json to deeplearning.template

* update template with latest AMI

* update new public AMI Ids
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants