[MXNET-1086] Increased Operator Coverage for TensorRT Integration

• Type: Improvement
• Status: To Do
• Priority: Major
• Resolution: Unresolved
• Assignee: Unassigned
• Reporter: Kellen Sunderland
• Created: 10/Oct/18 15:34
• Updated: 26/Jun/19 08:22

Description

The current operator coverage for TensorRT integration into MXNet is fairly limited.  As a user I would like to enable as many MXNet models as possible for acceleration with TensorRT.  In order to achieve this I would like all operators that are available in TensorRT to be properly mapped from their MXNet equivalents.

Issue Links

• links to: GitHub Pull Request #15344
• mentioned in: Page Loading...