search

ashawkey / stable-dreamfusion

Text-to-3D & Image-to-3D & Mesh Exportation with NeRF + Diffusion.
Apache License 2.0
8.33k stars 733 forks source link

Process killed and locked when creating mesh #202

Open Excitablecell opened 1 year ago

Excitablecell commented 1 year ago

Description

Sorry to bother you guys, but I seem to be having some issues

I can sucessfully generate example 'hamberger' . python3 main.py --text "a hamburger" --workspace trial -O --vram_O

Train,output video and test it successfully.

But generating mesh failed when python3 main.py --workspace trial -O --test --save_mesh

the error is : ==> Start Test, save results to trial/results 100% 100/100 [00:29<00:00, 3.79it/s]==> Finished Test. 100% 100/100 [00:31<00:00, 3.19it/s] ==> Saving mesh to trial/mesh [INFO] marching cubes thresh: 3.75 (0.0 ~ 234.477783203125) [INFO] mesh cleaning: (131020, 3) --> (81740, 3), (254360, 3) --> (164967, 3) [INFO] mesh decimation: (81740, 3) --> (24471, 3), (164967, 3) --> (49999, 3) [INFO] running xatlas to unwrap UVs for mesh: v=(24471, 3) f=(49999, 3) Killed

when I try again, there is a new error: ==> Start Test, save results to trial/results 100% 100/100 [00:29<00:00, 3.67it/s]==> Finished Test. 100% 100/100 [00:31<00:00, 3.17it/s] ==> Saving mesh to trial/mesh [INFO] marching cubes thresh: 2.5 (0.0 ~ 234.477783203125) [INFO] mesh cleaning: (129096, 3) --> (82692, 3), (250544, 3) --> (166709, 3) [INFO] mesh decimation: (82692, 3) --> (24495, 3), (166709, 3) --> (50000, 3) [INFO] running xatlas to unwrap UVs for mesh: v=(24495, 3) f=(50000, 3) Lock file exists in build directory: '/home/excai/.cache/torch_extensions/py310_cu117/nvdiffrast_plugin/lock'

Steps to Reproduce

after testing the model run python3 main.py --workspace trial -O --test --save_mesh

Expected Behavior

Output mesh normmaly

Environment

WSL, Ubuntu22.04 ,pytorch '2.0.0+cu117'

MelonGuo commented 1 year ago

I met the same error

es87m commented 1 year ago

I have the same lock error message, and the process seemed hung, so I halted it with ctrl-c. My train and output video commands worked. [Windows 10 with Unbuntu 22.04; pytorch 2.0.1] There are sites in general about 'Lock file exists in build directory' which suggest that there might be a /.git/lock.index file to delete in some cases; but not in my directory.

There was a 'lock' file in /home/user/.cache/torch_extensions/py310_cu117/nvdiffrast_plugin/ -rwxr-xr-x 1 user user 0 Aug 2 12:05 lock

Deleting that file gave me the 'Killed' message when running the command, as above. /trial/mesh directory is empty.

Pashtetickus commented 1 year ago

Hi. Did you resolve the problem? I met this now and can't train dmtet in threestudio.

es87m commented 1 year ago

Alas, no, I hope future updates might fix it since I’m not a programmer to know how to get around it yet! Best of luck to you,

Erick Singley NIH/NIAAA [Logo Description automatically generated]

From: Pavel @.> Date: Friday, August 4, 2023 at 8:51 AM To: ashawkey/stable-dreamfusion @.> Cc: Singley, Erick (NIH/NIAAA) [E] @.>, Comment @.> Subject: [EXTERNAL] Re: [ashawkey/stable-dreamfusion] Process killed and locked when creating mesh (Issue #202)

Hi. Did you resolve the problem? I met this now and can't train dmtet in threestudio.

— Reply to this email directly, view it on GitHubhttps://github.com/ashawkey/stable-dreamfusion/issues/202#issuecomment-1665558817, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AGDIFKWI2HMFGUVF24Y7MXTXTTV37ANCNFSM6AAAAAAXAG674E. You are receiving this because you commented.Message ID: @.***> CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and are confident the content is safe.