ModuleNotFoundError: No module named 'qlinear'

[AshimaBisla]

Hello,

While executing run_awq.py to quantize Llama 2 model, I am facing the error in line 14.

I am not able to figure out how to get qlinear module in my system.
It will be really helpful if someone can help me out with this.

Thanks,
Ashima

[BoruiXu]

Maybe you should run Setup Transformers first in /example/transformers/README.md. The Step 3 will install the qlinear package.

[AshimaBisla]

Hi @BoruiXu, I have run all the prerequisites including the Setup Transformers step, but I am still facing the issue.

[BoruiXu]

Maybe the command terminal causes the error? I tried to run it in the git bash terminal but failed to run. Then I switched to the Anaconda Prompt terminal, it works well.

[fengzhaoxin]

Hi @BoruiXu, I have run all the prerequisites including the Setup Transformers step, but I am still facing the issue.

Do not use windows terminal, you must use cmd.

[kesku]

The Step 3 will install the qlinear package.

Where is this stated? As far as I can tell, Step 3 is only to install RyzenAI-0.0.1

[BoruiXu]

The Step 3 will install the qlinear package.

Where is this stated? As far as I can tell, Step 3 is only to install RyzenAI-0.0.1

You can check the source code in ops\cpp.Or you can try import qlinear in Python script, I think it should work well.

[easygoal]

It seems in RyzenAI-SW 1.2.0, only qlinear_2 is provided. "qlinear" directory doesnot exist in ops\cpp

[easygoal]

It should work with RyzenAI-SW 1.2.0. In my case, I have to clean the build directory and re-compile the opt/cpp directory to make it work.