Issues with installation: No module named 'dust3r'

[Arborle]

Hi, I’m using a MacBook Pro M2 Max with macOS 15.2 Dev Beta.

After creating a Miniconda environment with Python 3.9, I was able to clone the repo and run the following commands as described in the README:

pip3 install -r requirements.txt --index-url https://download.pytorch.org/whl/cu121
pip3 install git+https://github.com/facebookresearch/pytorch3d@stable
pip3 install git+https://github.com/NVlabs/nvdiffrast

I didn’t encounter any errors and successfully downloaded the pretrained models. However, when I try to run the build with:

python3 main.py
--text_prompt "a lovely rabbit"
--save_folder ./outputs/test/
--max_faces_num 90000
--do_texture_mapping
--do_render

I get an error stating that the module dust3r cannot be found. I did find a file named dust3r_utils.py in the third-party subdirectory.

Has anyone experienced this issue, or does anyone have an idea how to resolve it?

Thank you in advance!

[liftyourgame]

Do you do this from the README?

because of dust3r, we offer a guide:

cd third_party
git clone --recursive https://github.com/naver/dust3r.git

cd ../third_party/weights
wget https://download.europe.naverlabs.com/ComputerVision/DUSt3R/DUSt3R_ViTLarge_BaseDecoder_512_dpt.pth

[Kallamamran]

I get the same error, but after successfully installing, starting the gradio interface and running t23d.
An OOM on my 24GB 3090 got me trying the --save_memory which generated this error.
I'm running on Windows 10

Don't know what to do now 🤨

After doing the things suggested by @liftyourgame I now get this error instead:

(venv) Q:\Hunyuan3D-1>python app.py --save_memory
sys path insert Q:\Hunyuan3D-1\third_party\dust3r
Warning, cannot find cuda-compiled version of RoPE2D, using a slow pytorch version instead
Q:\Hunyuan3D-1\third_party\dust3r\dust3r\cloud_opt\base_opt.py:275: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
  @torch.cuda.amp.autocast(enabled=False)
Baking is unavailable, please download related files in README
Baking is unavailable, please download related files in README
Baking is unavailable, please download related files in README

import baking related fail, run without baking
usage: app.py [-h] [--use_lite] [--mv23d_cfg_path MV23D_CFG_PATH] [--mv23d_ckt_path MV23D_CKT_PATH] [--text2image_path TEXT2IMAGE_PATH] [--save_memory SAVE_MEMORY] [--device DEVICE]
app.py: error: argument --save_memory: expected one argument

[Kallamamran]

Aha... The Readme file is not clear on how to use --save_memory.
It says to start the gradio like this "python3 app.py --use_lite --save_memory" when instead --save_memory expect an argument and should be started like this: "python3 app.py --use_lite --save_memory TRUE"

[Kallamamran]

Still OOM's on text-to-3d. Works ok with image-to-3D except for the GLB Baking which shows nothing 🤨