-
make a python venv and activate it (i named mine sec-env but like just name it whatever you want and add it to .gitignore)
-
clone moondream repo,
cd moondream
and dopip install transformers einops && pip install -r requirements.txt
(should probably add this as a submodule or something later) -
follow instructions below for putting the weights in the
checkpoints/grid
folder -
feel free to run gradio_demo.py or webcam_gradio_demo.py
bounding box demo (source)
bounding box demo but for webcams (source)
weights in the notion doc in google drive (don't post)
save the google drive files in osint-id folder inside checkpoints/grid
with the names
- text_model.final.pt (make sure to rename this one)
- vision_encoder.final.pt
- vision_projection.final.pt
should look something like this
ignore folder sec-env, that's the python venv