git lfs install && git lfs pull
Add this command before Setup section of main project page
Download the repo:
git clone https://github.com/mit-han-lab/hart
cd hart
conda create -n hart python=3.10
conda activate hart
conda install -c nvidia cuda-toolkit -y
pip install -e .
cd hart/kernels && python setup.py install
Without that many people will try to run the above commands and expect things to work out of the box BUT it wont work and this error will be shown in their terminal :
OSError: You seem to have cloned a repository without having git-lfs installed. Please install git-lfs and run git lfs install followed by git lfs pull in the folder you cloned.
and their pulled repos will have size of 15-20 Mb instead of 4GB and 6GB size and some additonal errors will be thrown on top of this error so this one command will fix that issue for all :-))
Thank you Very Much for this great project team
( Wish i could test this project on my 8GB VRAM Laptop :-)), but unfortunately requirements are quite high for us - 8.24 GiB Qwen2-VL-1.5B-Instruct+13.16 GiB hart-0.7b-1024px is ALOTTT to test on any laptop so we will just enjoy visual development and hope some efficient solutions develop in future for us. Till then we will enjoy quantized flux models :-))
git lfs install && git lfs pullAdd this command before Setup section of main project page
Without that many people will try to run the above commands and expect things to work out of the box BUT it wont work and this error will be shown in their terminal :
OSError: You seem to have cloned a repository without having git-lfs installed. Please install git-lfs and run git lfs install followed by git lfs pull in the folder you cloned.and their pulled repos will have size of 15-20 Mb instead of 4GB and 6GB size and some additonal errors will be thrown on top of this error so this one command will fix that issue for all :-))
Thank you Very Much for this great project team
( Wish i could test this project on my 8GB VRAM Laptop :-)), but unfortunately requirements are quite high for us - 8.24 GiB Qwen2-VL-1.5B-Instruct+13.16 GiB hart-0.7b-1024px is ALOTTT to test on any laptop so we will just enjoy visual development and hope some efficient solutions develop in future for us. Till then we will enjoy quantized flux models :-))