blender插件API

[CZZ000]

直接生成白模就没问题，勾选Generate Texture就出错

[nauseum]

same erorr here and also cannot generate from text prompt
both texture and text work fine in Gradio

[YanWenKun]

确实，看了眼 api_server.py ，它没像 gradio_app.py 一样自动检测材质功能是否编译安装完毕，需要手动启用。
我刚给添加了个新启动脚本，运行 更新.bat 后可以看到。

[YanWenKun]

@nauseum Run UPDATE.bat and then run the new 6-start-api-server-with-texture.bat

[nauseum]

@nauseum Run UPDATE.bat and then run the new 6-start-api-server-with-texture.bat

Thank you , I no longer get an error but it never completes. I get as far as FlashVDM then it revs up my CPU to 100% and 15GB of RAM and hangs.. I let it run for nearly 10 minutes then killed it.. in gradio this take completes in roughly 40 seconds
Repeated attempts 5 times, same result
(if it helps have a 5090 macine so i have had to run blackwell update batch file to get this working in the Gradio version)

Output from the server version never gets past that last line, it passes flash VDM then dies.

Diffusion Sampling:: 100%|█████████████████████████████████████████████████████████████| 20/20 [00:01<00:00, 15.21it/s]
2025-05-03 07:32:51 | ERROR | stderr |
2025-05-03 07:32:51,924 - hy3dgen.shapgen - INFO - FlashVDMVolumeDecoding Resolution: [63, 126, 252]
2025-05-03 07:32:51 | INFO | hy3dgen.shapgen | FlashVDMVolumeDecoding Resolution: [63, 126, 252]
FlashVDM Volume Decoding: 0%| | 0/64 [00:00<?, ?it/s]
FlashVDM Volume Decoding: 100%|███████████████████████████████████████████████████████| 64/64 [00:00<00:00, 995.60it/s]
2025-05-03 07:32:52 | ERROR | stderr |
2025-05-03 07:32:52 | INFO | controller | --- 3.2032601833343506 seconds ---
Server Never gets past here ^^^^

Here is typical output from gradio app on same image, gradio my CPU never goes past 6% usage. Might be the clue.

Diffusion Sampling:: 100%|███████████████████████████████████████████████████████████████| 5/5 [00:00<00:00, 5.63it/s]
Volume Decoding: 100%|████████████████████████████████████████████████████████████| 2122/2122 [00:04<00:00, 430.14it/s]
2025-05-03 07:37:36,906 - hy3dgen.shapgen - INFO - ---Shape generation takes 7.382655620574951 seconds ---
2025-05-03 07:37:42,295 - hy3dgen.shapgen - INFO - ---Face Reduction takes 5.252882242202759 seconds ---
2025-05-03 07:37:57,830 - hy3dgen.shapgen - INFO - ---Texture Generation takes 15.5351243019104 seconds ---
Find html file gradio_cache\8c7b8ed0-258a-490b-b20b-6638fc4e86a2\textured_mesh.html, True, relative HTML path is /static/8c7b8ed0-258a-490b-b20b-6638fc4e86a2\textured_mesh.html

[YanWenKun]

@nauseum Since you're using 5090, you can use DMC instead of MC for FlashVDM.

Run UPDATE.bat and then run the new 6-start-api-server-with-texture-dmc.bat.

Note: API server mode is not using mmgp for VRAM optimization, it sticks to upstream's Python code, which runs faster but uses more VRAM.

[YanWenKun]

By the way, a new release using CUDA 12.8 may help for Blackwell GPUs:

https://github.com/YanWenKun/Hunyuan3D-2-WinPortable/releases/tag/v3.1

[nauseum]

@nauseum Since you're using 5090, you can use DMC instead of MC for FlashVDM.

Run UPDATE.bat and then run the new 6-start-api-server-with-texture-dmc.bat.

Note: API server mode is not using mmgp for VRAM optimization, it sticks to upstream's Python code, which runs faster but uses more VRAM.
By the way, a new release using CUDA 12.8 may help for Blackwell GPUs:

https://github.com/YanWenKun/Hunyuan3D-2-WinPortable/releases/tag/v3.1

Reinstalled the new 3.1 version,
installed Cuda 12.9 and ran that batch file and DMC version.
Same issue remains, output gets as far as completing FlashVDM then it spins up CPU to 90%++ and ~15GB of RAM then remains there for ~ 10 minuites then crashes, ill post output of crash below, the remeshing and texturing are failing as far as i can tell.

Gradio still runs fine.

Output Errors I spotted, these errors exist in both the DMC and regular versions. Aslo in your previous version Cuda 12.8 and new version for 12.9

An error occurred while trying to fetch C:\Hunyuan3D2_WinPortable\\HuggingFaceHub\models--tencent--Hunyuan3D-2\snapshots\34e28261f71c32975727be8db0eace439a280f82\hunyuan3d-paint-v2-0-turbo\vae: Error no file named diffusion_pytorch_model.safetensors found in directory C:\Hunyuan3D2_WinPortable\\HuggingFaceHub\models--tencent--Hunyuan3D-2\snapshots\34e28261f71c32975727be8db0eace439a280f82\hunyuan3d-paint-v2-0-turbo\vae.

A matching Triton is not available, some optimizations will not be enabled
Traceback (most recent call last):
  File "C:\Hunyuan3D2_WinPortable\python_standalone\Lib\site-packages\xformers\__init__.py", line 57, in _is_triton_available
    import triton  # noqa
    ^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'triton'