How to debug torchserve pip install?

deployment
1

We’re trying to use torchserve (via KFServing).

We prepared .mar with our model and requirements.txt (most notable dependency is transformers)

When we try to run in torchserve logs I see a line “Custom pip package installation failed for ”

When connect to container and execute “pip install -r requirements.txt” all dependencies install just fine.

Question: how do I get more detailed error about pip install from torchserve?

2

I need to document this better but the answer is you need to add this line to config.properties install_py_dep_per_model=true