search

lhnguyen102 / cuTAGI

CUDA implementation of Tractable Approximate Gaussian Inference
MIT License
30 stars 9 forks source link

Release 0.1.7 #62

Closed lhnguyen102 closed 5 months ago

lhnguyen102 commented 5 months ago

Description

This PRs added the draft release for pytagi that connects to the latest C++/CUDA backend

Changes Made

Note for Reviewers

  1. Create and Activate a New Conda Environment Create a new Conda environment named pytagi_env with Python 3.10, and activate it: 
    conda create --name pytagi_env python=3.10
conda activate pytagi_env
  2. Install pytagi in the cuTAGI Folder 
    cd path/to/cuTAGI
pip install .
pip install -r requirements.txt
  3. Run examples. Copy examples folder to new directory path 
    cd /new/path/examples
python -m examples.regression
jamesgoulet commented 5 months ago

@lhnguyen102 1 - When I copy the example directory in a new folder following the 3- from the readme, I am having a missing import error:

(pytagi_test) jg@MLCIVS1:~/cuTAGI/new/examples$ python -m examples.regression
Traceback (most recent call last):
  File "/home/jg/miniconda3/envs/pytagi_test/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/jg/miniconda3/envs/pytagi_test/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/home/jg/miniconda3/envs/pytagi_test/lib/python3.10/site-packages/examples/regression.py", line 3, in <module>
    import fire
ModuleNotFoundError: No module named 'fire'

I do not get what is the point of copying this directory?

2- Should you explicitly add #net.to_device('cuda') & #net.set_threads(20) in the regression.py & classification.py example files?

3- Since the LayerNorm( ) layer do not work on CUDA... should we remove it from the classification.py example?

4- For time_series_forecasting the number of epochs should be reduced to 5 in order to avoid nans -> num_epochs: int = 5