leeexyz opened a new pull request #9889:
URL: https://github.com/apache/tvm/pull/9889


   Hi All,
   
   In the GraphExecutor use case, we always should get the input info of a 
compiled graph module, including shape and dtype. For `TVMC`, it provides 
[get_input_info](https://github.com/apache/tvm/blob/8c40dfd8493aff01f055ef13f2d44caedf5cf74b/python/tvm/driver/tvmc/runner.py#L583-L584)
 to get and then to generate the input data. But it is inconvenient if we only 
deploy the CPP runtime because there is no graph info to be exposed outside of 
GraphExecutor, and it is very cumbersome to load and parse the Graph JSON 
again. So here I move the `get_input_info` function into GraphExecutor, then we 
can easily write a 
[make_inputs_dict](https://github.com/apache/tvm/blob/8c40dfd8493aff01f055ef13f2d44caedf5cf74b/python/tvm/driver/tvmc/runner.py#L372)
 function on the CPP side.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to