Predict outliers and predict influential individuals

[SaraSALHI2]

Hello everyone,
I can't use predict outliers or influential individuals in PharmR as I always receive the same error "No module named 'tflite_runtime"

Could you please advise

[rikardn]

Unfortunately this is a bit complicated at the moment. Installing tflite ranges from a bit involved to almost impossible depending on which operating system and Python version you are using.

It is currently supported on linux with Python 3.8, 3.9 or 3.10, MacOS with Python 3.8 and 3.9 and Windows with Python 3.8 and 3.9. Since you are using R the package needs to be manually installed in the virtual environment that reticulate is using (by default r-reticulate). reticulate::py_config can be used to check which environment and Python version is being used. If you are on one of the supported platforms you can activate the virtual environment and install using one of the pip descriptions below:

tflite-runtime==2.11.0; sys_platform == 'linux' and python_version == '3.8'
tflite-runtime==2.11.0; sys_platform == 'linux' and python_version == '3.9'
tflite-runtime==2.5.0; sys_platform == 'darwin' and python_version == '3.8'
tflite_runtime @ https://github.com/hjonnala/snippets/raw/main/wheels/python3.10/tflite_runtime-2.5.0.post1-cp310-cp310-linux_x86_64.whl ; sys_platform == 'linux' and python_version == '3.10'
--extra-index-url https://google-coral.github.io/py-repo/
tflite_runtime==2.5.0.post1; sys_platform == 'darwin' and python_version == '3.9'
tflite_runtime==2.5.0.post1; sys_platform == 'win32' and python_version == '3.8'
tflite_runtime==2.5.0.post1; sys_platform == 'win32' and python_version == '3.9'

Sorry for this being difficult at the moment. In the future we will try to support more platforms and make it easier to use.

[SaraSALHI2]

Hi,

Thank you Rikard, it worked very well.
PharmPy was installed in a virtual folder however now that I installed PsN library in Rstudio I realized that I have two PharmPy one in the virtual folder and another in miniconda3. In Rstudio it mutated to miniconda3 instead of .entvirtual path therefore I can't fit the models anymore and don't receive modelfitresults even if I added NONMEM path to both config.py file
Do you have an idea why it happened