Skip to content

Frequently Asked Questions

If your question is not listed, try using DeepWiki's AI assistant for common issues.

For unresolved problems, join our Discord or WeChat community for support.

What should I do if inference is slow after installing directly on Windows?

What should I do if inference is slow after installing directly on Windows?

If inference is slow after a direct Windows installation, CUDA acceleration dependencies are usually not installed correctly. Choose the solution based on your GPU architecture:

  • Volta / Turing / Ampere / Ada Lovelace GPUs, such as V100, 20 series, T4, 30 series, and 40 series: install CUDA-enabled torch and torchvision directly. Go to the PyTorch official website and select the Windows installation command that matches your CUDA version.
  • Blackwell GPUs, such as the RTX 50xx series: install the lmdeploy 0.11.1 + cu128 Windows wheel. Set PYTHON_VERSION to your current Python version, for example 310 / 311 / 312 / 313 for Python 3.10 / 3.11 / 3.12 / 3.13.
$env:LMDEPLOY_VERSION = "0.11.1"
$env:PYTHON_VERSION = "312"

$wheel = "https://github.com/InternLM/lmdeploy/releases/download/v$($env:LMDEPLOY_VERSION)/lmdeploy-$($env:LMDEPLOY_VERSION)+cu128-cp$($env:PYTHON_VERSION)-cp$($env:PYTHON_VERSION)-win_amd64.whl"
pip install $wheel --extra-index-url https://download.pytorch.org/whl/cu128

If you are using a Blackwell GPU and have already installed the cu128 version of torch, define $wheel first and then run the following command to avoid downloading a lower-version torch again:

pip install $wheel --no-dependencies
Encountered the error ImportError: libGL.so.1: cannot open shared object file: No such file or directory in Ubuntu 22.04 on WSL2

Encountered the error ImportError: libGL.so.1: cannot open shared object file: No such file or directory in Ubuntu 22.04 on WSL2

The libgl library is missing in Ubuntu 22.04 on WSL2. You can install the libgl library with the following command to resolve the issue:

sudo apt-get install libgl1-mesa-glx

Reference: #388

Missing text information in parsing results when installing and using on Linux systems.

Missing text information in parsing results when installing and using on Linux systems.

MinerU uses pypdfium2 instead of pymupdf as the PDF page rendering engine in versions >=2.0 to resolve AGPLv3 license issues. On some Linux distributions, due to missing CJK fonts, some text may be lost during the process of rendering PDFs to images. To solve this problem, you can install the noto font package with the following commands, which are effective on Ubuntu/Debian systems: 

sudo apt update
sudo apt install fonts-noto-core
sudo apt install fonts-noto-cjk
fc-cache -fv
You can also directly use our Docker deployment method to build the image, which includes the above font packages by default.

Reference: #2915