chenguolin / InstructScene

[ICLR 2024 spotlight] Official implementation of "InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior".
https://chenguolin.github.io/projects/InstructScene
MIT License
55 stars 7 forks source link

FileNotFoundError: [Errno 2] No such file or directory: 'out/threedfront_objfeat_vqvae/objfeat_bounds.pkl' #1

Closed foojayyy closed 3 months ago

foojayyy commented 4 months ago

I am encountering a FileNotFoundError when trying to train the Sg2Sc objfeat model. The error message is: FileNotFoundError: [Errno 2] No such file or directory: 'out/threedfront_objfeat_vqvae/objfeat_bounds.pkl' The error occurs at line 179 of train_sg2sc.py when the script tries to open the objfeat_bounds.pkl file with the following command: with open(f"{args.output_dir}/{args.fvqvae_tag}/objfeat_bounds.pkl", "rb") as f: I have confirmed that the out directory and the threedfront_objfeat_vqvae subdirectory exist, but the objfeat_bounds.pkl file is missing. I would appreciate any guidance on how to resolve this issue. I have tried to debug it myself, but I am stuck and need further assistance. Thank you for your help.

chenguolin commented 4 months ago

Hi @foojayyy,

The objfeat_bounds.pkl file is produced in the beginning of training fVQ-VAE: https://github.com/chenguolin/InstructScene/blob/1086b15d7e783f7347ca356e43df58e6dbb092e9/src/train_objfeatvqvae.py#L202

I have uploaded this file to Huggingface and updated README.md, you can download it by:

import os
from huggingface_hub import hf_hub_url
# os.system("mkdir -p out/threedfront_objfeat_vqvae/checkpoints")
# url = hf_hub_url(repo_id="chenguolin/InstructScene_dataset", filename="threedfront_objfeat_vqvae_epoch_01999.pth", repo_type="dataset")
# os.system(f"wget {url} -O out/threedfront_objfeat_vqvae/checkpoints/epoch_01999.pth")
url = hf_hub_url(repo_id="chenguolin/InstructScene_dataset", filename="objfeat_bounds.pkl", repo_type="dataset")
os.system(f"wget {url} -O out/threedfront_objfeat_vqvae/objfeat_bounds.pkl")

You can also prepare it by running train_objfeatvqvae.py.

Thank you for pointing it out!

foojayyy commented 4 months ago

Thank you for your guidance and sharing, I have downloaded the required file according to your instructions!