[
https://issues.apache.org/jira/browse/MXNET-123?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Olivier updated MXNET-123:
--------------------------------
Issue Type: Sub-task (was: Bug)
Parent: MXNET-125
> Fix race condition in sparse batch loader
> -----------------------------------------
>
> Key: MXNET-123
> URL: https://issues.apache.org/jira/browse/MXNET-123
> Project: Apache MXNet
> Issue Type: Sub-task
> Reporter: Haibin Lin
> Priority: Major
> Time Spent: 10m
> Remaining Estimate: 0h
>
> The previous implementation of sparse batch loader waits for the number of
> data instances reaches {{batch_size}} before allocating the buffer for copy
> (unlike dense batch loader, we don't know the actual buffer size to allocate
> since the number of non-zeros varies). However, such delayed allocate is
> buggy when batch size is extremely large, because the {{DataInst}} returned
> by {{base_}} is only a reference to the data. By the time copy happens, the
> referenced data might already be updated by the parser.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
