AiuniAI / Unique3D

[NeurIPS 2024] Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image
https://wukailu.github.io/Unique3D/
MIT License
3.11k stars 246 forks source link

有大佬测试显存占用吗? #18

Open withoutbuff opened 5 months ago

withoutbuff commented 5 months ago

输入179×179分辨率的图片,爆显存了 Tue Jun 11 14:08:51 2024
+---------------------------------------------------------------------------------------+ | NVIDIA-SMI 545.29.06 Driver Version: 545.29.06 CUDA Version: 12.3 | |-----------------------------------------+----------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+======================+======================| | 0 NVIDIA GeForce RTX 4060 Ti Off | 00000000:01:00.0 On | N/A | | 34% 37C P2 25W / 165W | 15895MiB / 16380MiB | 12% Default | | | | N/A | +-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=======================================================================================| | 0 N/A N/A 1133 G /usr/lib/xorg/Xorg 292MiB | | 0 N/A N/A 1361 G /usr/bin/gnome-shell 212MiB | | 0 N/A N/A 3917 G ...09,262144 --variations-seed-version 80MiB | | 0 N/A N/A 359557 G ...erProcess --variations-seed-version 28MiB | | 0 N/A N/A 390188 G ...irefox/4336/usr/lib/firefox/firefox 195MiB | | 0 N/A N/A 1415633 C+G python 15016MiB | +---------------------------------------------------------------------------------------+

withoutbuff commented 5 months ago

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 256.00 MiB. GPU

kexul commented 5 months ago

我在4090能跑,大概20G左右显存占用。

Aiunii commented 5 months ago

Hey!  Welcome to 4D digital art creation universe—— Aiuni.

🤖First, join our community by:  https://discord.gg/aiuni

🦄️Then you will get the invitation code at “getting-started” channel.

Start your creative journey and have fun!🌟

@.> 时间: 2024年6月11日 (周二) 下午2:15 主题: [AiuniAI/Unique3D] 有大佬测试显存占用吗? (Issue #18) @.> @.***> 输入179×179分辨率的图片,爆显存了

Tue Jun 11 14:08:51 2024

+---------------------------------------------------------------------------------------+

| NVIDIA-SMI 545.29.06 Driver Version: 545.29.06 CUDA Version: 12.3 |

|-----------------------------------------+----------------------+----------------------+

| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |

| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |

| | | MIG M. |

|=========================================+======================+======================|

| 0 NVIDIA GeForce RTX 4060 Ti Off | 00000000:01:00.0 On | N/A |

| 34% 37C P2 25W / 165W | 15895MiB / 16380MiB | 12% Default |

| | | N/A |

+-----------------------------------------+----------------------+----------------------+ +---------------------------------------------------------------------------------------+

| Processes: |

| GPU GI CI PID Type Process name GPU Memory |

| ID ID Usage |

|=======================================================================================|

| 0 N/A N/A 1133 G /usr/lib/xorg/Xorg 292MiB |

| 0 N/A N/A 1361 G /usr/bin/gnome-shell 212MiB |

| 0 N/A N/A 3917 G ...09,262144 --variations-seed-version 80MiB |

| 0 N/A N/A 359557 G ...erProcess --variations-seed-version 28MiB |

| 0 N/A N/A 390188 G ...irefox/4336/usr/lib/firefox/firefox 195MiB |

| 0 N/A N/A 1415633 C+G python 15016MiB |

+---------------------------------------------------------------------------------------+ — Reply to this email directly, view it on GitHubhttps://github.com/AiuniAI/Unique3D/issues/18, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BI3NO5U22MOLMUW2GVXBC4DZG2IXXAVCNFSM6AAAAABJDT5ZZSVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM2DKNBYG42TINQ. You are receiving this because you are subscribed to this thread.[image: https://github.com/notifications/beacon/BI3NO5U7UHKIXUVBOE5W6LLZG2IXXA5CNFSM6AAAAABJDT5ZZSWGG33NNVSW45C7OR4XAZNFJFZXG5LFVJRW63LNMVXHIX3JMTHIXTKMXI.gif]Message ID: @.***>

wukailu commented 5 months ago

运行显存需求和输入图像分辨率基本上没有关系(即2048分辨率和240分辨率用的是差不多的) 通过在各种 pipeline 增加 enable_model_cpu_offload 可以降低到15GB 显存需求左右,但是会导致内存泄漏(我也不知道为什么)

ouceduxzk commented 4 months ago

实测 4090 20g左右, 速度很快,几秒钟就出一个模型

userYKK commented 4 months ago

实测 4090 20g左右, 速度很快,几秒钟就出一个模型

你好,你cuda 版本是多少呀, https://github.com/AiuniAI/Unique3D/issues/81 我被nvdiffrast-torch 卡住了