[BUG]: Error running an example per https://builds.modular.com/builds/llama3/python

[skoppisetti]

Bug description

I followed all the instructions on the page and in addition after CD ing into the pipelines/python folder I ran the "magic shell" to activate the environment. However I am stuck with the following error:

None of PyTorch, TensorFlow >= 2.0, or Flax have been found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
...
ValueError: cannot reshape array of size 12679776 into shape (4096,11760)

Steps to reproduce

Run Llama3 on your laptop

System information

- Mac OS Sanoma - 14.6.1 (23G93)
- max 24.5.0 (e8aacb95)
- mojo 24.5.0 (e8aacb95)
- Modular version 24.5.0

[ehsanmok]

Thanks for reporting! magic run llama3 --prompt "hi" works however when I magic shell and run python pipeline.py llama3 --prompt "hi" I get a different error. We'll fix the magic shell path.

[ehsanmok]

For my case, it turned out I needed to deactivate any conda env even base by conda deactivate then magic shell path worked as expected. If that's the case for you, could you share the outputs of these cases?

1. magic run env | grep MOD
2. magic shell then env | grep MOD
3. magic shell then which python