Fix installation guide

[samsara-ku]

I believe the provided installation guide was worked well with someone, but it is not my case.

Here is installation instruction for me (maybe this is not worked with someone 😅)

conda env create -f env_sigma.yml

pip install -U openmim
mim install mmengine
mim install "mmcv>=2.0.1"
mim install "mmdet>=3.1.0"
mim install "mmpose>=1.1.0"

conda install -c conda-forge cudatoolkit-dev -y

pip install gradio pillow-avif-plugin torchaudio==2.0.0 huggingface-hub==0.25.2

git submodule update --init detectron2/

pip install 'git+https://github.com/facebookresearch/[email protected]#egg=detectron2'
pip install 'git+https://github.com/facebookresearch/[email protected]#subdirectory=projects/DensePose'

pip install refile audio_augmentations

git clone https://github.com/facebookresearch/WavAugment.git && cd WavAugment && python setup.py develop

pip install openai-clip

[samsara-ku]

I tried to do my best to run demo/run_gradio.py, and I don't know why the repository maintainer just leave with those half instruction. There are too many omitted steps to run this codes.

Here is my additional instruction started from the above steps:

1. You should change the line 28 of animate/megactor-sigma/animate.py like this:

AS-IS
from refile import smart_open, smart_load_from

TO-BE
from megfile import smart_open, smart_load_from

and run pip install megfile

I don't know this is right answer, but I can't find appropriate version of refile that contain that function. The only alternative library that has smart_open and smart_load_from is the megfile, the another library of this maintainer.

2. Run pip install ffmpeg-python

3. Following the installation guide of XPOSE. Otherwise, probably you'll see the error message of MSDeformAttnFunction.

c.f.) It takes so much time (~600s), just starting the demo/run_gradio.py code.. It will be appreciated someone help this time-consuming process.