Hi all,
I just installed tvm and was trying out https://tvm.apache.org/docs/tutorials/get_started/tvmc_command_line_driver.html#sphx-glr-tutorials-get-started-tvmc-command-line-driver-py
I used method 2 (in https://tvm.apache.org/docs/install/from_source.html) to install the tvm package.
I’m using pipenv to manage my virtual envs. The setup.py was run by the virtual environment.
I installed all the other dependencies in the virtual environment with pipenv as well.
If i try running tvmc
i get this traceback:
Traceback (most recent call last):
File "/home/josse/.local/share/virtualenvs/tryout-tvm-MK_O0g7I/bin/tvmc", line 33, in <module>
sys.exit(load_entry_point('tvm==0.8.dev122+g1b6ef5c80', 'console_scripts', 'tvmc')())
File "/home/josse/.local/share/virtualenvs/tryout-tvm-MK_O0g7I/bin/tvmc", line 25, in importlib_load_entry_point
return next(matches).load()
StopIteration
It seems that I can run python -m tvm.driver.tvmc
So it’s not really bugging me too much, but I thought I’d mention it.
Was not sure if I could make github issue for this small of an issue. 
I’m on Arch Linux
Thanks in advance!
Hi @JosseVanDelm,
The way that tvmc
command line appears when you install TVM is by using a setup.py entry_points
clause, here tvm/python/setup.py at main · apache/tvm · GitHub.
For some reason it looks like it doesn’t work as expected with pipenv
.
I don’t use pipenv
, can you please expand a bit on how are you using it to install TVM, meaning, which commands you are using?
Note: I did a quick search and it looks like pipenv
, apparently, doesn’t play well with setup.py
? Please correct me if I’m something that doesn’t make sense.
Anyway, we can surely have a look to see what is wrong
. Thanks for reporting the issue.
LN
Hi @leandron,
Thanks for the quick reply!
I’m experimenting with pipenv
myself a bit (right now it actually seems rather slow, i wouldn’t recommend it actually
) , so i’m not sure what could have gone wrong.
I’m not really sure what the entry_points clause does that you referenced, and i’m also not sure what the deal is with setup.py and pipenv. But I’ll just try to explain what I did and why i did.
Also apparently I just fixed it myself. 
So the way I see pipenv right now is that it’s just a replacement for pip but it will manage package dependencies in a better way (apparently
) and install them in a virtualenv automatically.
So I followed the tutorial but every time I had to use python
(for setup.py
) i made sure that I first activated the virtual environment with pipenv shell
(similar to conda activate
and source .../activate
I guess). And all packages i installed with pipenv install
instead of pip
.
I think I mistakingly installed the tvm package as well (not apache-tvm
) which might have caused the issue. After uninstalling it with pipenv uninstall tvm
I’m now getting the expected behavior when calling tvmc
. Thanks!
1 Like
No worries, any issue, let me know 
1 Like