MotionLCM 部署笔记

目录

依赖项

渲染黑白图问题解决：

humanml3d：

sentence-t5-large

下载数据：

报错：No module named 'sentence_transformers'

继续报错：from transformers.integrations import CodeCarbonCallback

解决方法：

推理相关

转mesh：

module ‘matplotlib.cm‘ has no attribute ‘get_cmap‘

GitHub - Dai-Wenxun/MotionLCM: [ ECCV 2024 ] MotionLCM: This repo is the official implementation of "MotionLCM: Real-time Controllable Motion Generation via Latent Consistency Model"

依赖项

matplotlib 必须指定版本：pip install matplotlib==3.3.4

否则会渲染失败

渲染黑白图问题解决：

matplotlib 必须指定版本：pip install matplotlib==3.3.4

否则会渲染失败

humanml3d：

GitHub - EricGuo5513/HumanML3D: HumanML3D: A large and diverse 3d human motion-language dataset.

sentence-t5-large

git clone https://huggingface.co/sentence-transformers/sentence-t5-large

下载数据：

https://drive.google.com/drive/folders/1D3bf2G2o4Hv-Ale26YW18r1Wrh7oIAwK

解压到路径：

MotionLCM-main/MotionLCM-main/datasets/humanml3d

报错：No module named 'sentence_transformers'

解决方法：

pip install sentence-transformers==2.2.2

继续报错：from transformers.integrations import CodeCarbonCallback

解决方法：

pip install transformers==4.38.0
pip install sentence-transformers==2.2.2

推理相关

batch是输入length 和text，输出是length x 22 x 3

batch = {"length": length, "text": text}
length：196 
输出维度：196 22 3

转mesh：

chumpy==0.70

python fit.py

module ‘matplotlib.cm‘ has no attribute ‘get_cmap‘

module ‘matplotlib.cm‘ has no attribute ‘get_cmap‘-CSDN博客

bpy 4.0.0 4.3.0都可以渲染。