[
https://issues.apache.org/jira/browse/SINGA-406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16691231#comment-16691231
]
wangwei commented on SINGA-406:
-------------------------------
# if nvidia docker images can run on cpu-only machines, then there is no
problem. we just need a single base worker image.
# explanation of 'runtime': shall we configure the base worker image to only
install small/necessary libs? and let the worker container install the other
model specific libs (configured by the model contributor) when the job is
started, i.e., at runtime. For example, we configure the base worker image to
install rafiki dependent libs and common model dependent libs like numpy and
scikit-learn. The model contributor can specify other dependent libs like
singa, pytorch etc. when he/she adds a new model. These libs are installed when
the model is used in a job. The model contributor can also provide the docker
image directly (the name/tag) that extends the base worker image.
# we should not let the app developer to configure the container. they only
need to prepare the data and start the job.
> [Rafiki] Add POS tagging task & add GPU support (0.0.7)
> -------------------------------------------------------
>
> Key: SINGA-406
> URL: https://issues.apache.org/jira/browse/SINGA-406
> Project: Singa
> Issue Type: New Feature
> Reporter: Ngin Yun Chuan
> Priority: Major
>
> Refer to https://github.com/nginyc/rafiki/pull/71 for details
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
