[Bug] mlc-llm not working, tvm check returns none

[CallMeTkt]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

1. i dont even know bro

Expected behavior

Environment

• Platform (e.g. WebGPU/Vulkan/IOS/Android/CUDA):
• Operating system (e.g. Ubuntu/Windows/MacOS/...):
• Device (e.g. iPhone 12 Pro, PC+RTX 3090, ...)
• How you installed MLC-LLM (conda, source):
• How you installed TVM-Unity (pip, source):
• Python version (e.g. 3.10):
• GPU driver version (if applicable):
• CUDA/cuDNN version (if applicable):
• TVM Unity Hash Tag (python -c "import tvm; print('\n'.join(f'{k}: {v}' for k, v in tvm.support.libinfo().items()))", applicable if you compile models):
• Any other relevant information:

Additional context

[tqchen]

make sure you installed mlc_llm. on windows we recommend running through a conda env

[CallMeTkt]

Could we potentially get on a discord call? I believe I already installed it but I'm not sure what i did wrong exactly

[CallMeTkt]

here is what happens when i try to install it again

[CallMeTkt]

this one doesnt work either, am i missing something obvious?

[CallMeTkt]

its here in my android project as well

[tqchen]

seems you are not installing into the conda env, and python3 you invoked was not from conda env. try to run which python3 to ensure the python comes from conda, make sure you installed python in conda.

Consider change to python -m if that python comes from conda

[CallMeTkt]

what should i do to install it to conda environment?

[tqchen]

usually conda install python would suffice, then you run python -m pip install instead of python3 -m pip which goes into your system python

[CallMeTkt]

got to the second part but got this error

[CallMeTkt]

im still not sure how i would fix the mlc_llm install, i already have it installed but its not picking it up

[tqchen]

what i meant was run python -m pip install --pre -U -f https://mlc.ai/wheels mlc-llm-nightly mlc-ai-nightly

Make sure you also run conda install python pip

[CallMeTkt]

Alright, I'll try that when I get off school today, should be in about 8 hours. Also thanks for all the help

[CallMeTkt]

did what you said, now its saying invalid config file

[CallMeTkt]

should i make a file and put all these in it in models?

[CallMeTkt]

ok so it ran, but now it says numpy is not available, is this a problem?

[CallMeTkt]

whenever i run prepare model lib it just doesnt output any files in build, is there a fix?

[tqchen]

you do need to install the necessary dependencies such as numpy, usually they should come with pip install, but given it is missing, you need to manually install it

[CallMeTkt]

Would I run a command or just Google it and run a installer

[CallMeTkt]

it says its already installed

[tqchen]

unfortunately seems your python did not pick it up, i dn't know why, seems to be issue with your loal env

[CallMeTkt]

is there a environment variable i should setup?

[tqchen]

I am sorry i don't really know, normally if the conda environment contains the python and you dpn't use your system python, and that python contains all necessary dependency, we should be fine

[CallMeTkt]

what version of python would it use.

i came across this right here, maybe i need a different version of python and numpy

[CallMeTkt]

i managed to get it working by some downgrading of numpy, now but i cant seem to find that output after running ./prepare_libs.sh