[
https://issues.apache.org/jira/browse/SINGA-406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16691303#comment-16691303
]
wangwei commented on SINGA-406:
-------------------------------
2. In the documentation page,
{noformat}
If the model is to be run on GPU, Rafiki will make a best-effort attempt to map
dependencies to their GPU-enabled versions, if required. For example, {
'tensorflow': '1.12.0' } will be installed as { 'tensorflow-gpu': '1.12.0'
}{noformat}
it means we (rafiki developer) need to manually do some adjustment for some
libs. The gpu case is one example. Another example: if the base worker image is
extended from nvidia/cuda:9.0-runtime-ubuntu16.04 and the default version of
the lib (e.g., tf) from PIP requires cuda 8.0 or cuda10.0, we need to find the
proper lib to install it (using pip or conda). We can handle singa in the same
way.
> [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)
