python APIs like tvm.contrib.graph_executor.GraphModule and tvm.device/tvm.runtime.device has not exist in tvm 0.8, and your tutorial docs are still showing them. It brings trouble to freshmen like me.
Thanks for sharing this information. @tqchen How do we avoid this condition in the future? Doc tutorial doesn’t exist in CI but it is very important, but developers forget to update it very easily.
Thank you for bringing this up and sorry for the confusion. Right now we are still in the 0.8 development cycle and the 0.8.dev0 docs is synced to the latest in main where these changes are being bought during this development cycle (so they are not stale docs). The tutorials are being executed on the CI to make sure they are up to date.
Starting v0.8 we will work publish the docs that corresponds the release tag so that developers can look into these versions.
Is it possible that we keep a backup doc for each release?
For example a tag option in the website that users can choose to see the doc of 0.7-release, 0.8-release, nightly…
Oh, just ignore my words above. @tqchen
I didn’t see your last reply. 
thanks for all you guys’ reply! Wish tvm a better future!
I completely agree. I think one of the biggest confusions is that some users (i.e. me 
) are still compiling models with an explicit json and params.
@tqchen, I’m actually trying to build the graph runtime in C++, but graph_runtime is deprecated, and there is no documentation for the use of graph_executor.
I would love to get this working and try to update some examples