Install in mainland China
The first launch downloads the AI models and the software they run in. In mainland China those downloads fail or stall, because the servers they come from are blocked or very slow. The fix is to use mirrors: copies of the same files on Chinese servers.
You need AddaxAI 7.0.13 or newer.
Point the app to the mirrors
Set these three, see environment variables for how to do that on your system:
| Variable | Value |
|---|---|
ADDAXAI_HF_ENDPOINT | https://hf-mirror.com |
ADDAXAI_HF_HUB_DISABLE_XET | 1 |
ADDAXAI_PYTORCH_INDEX_URL | https://mirror.nju.edu.cn/pytorch/whl |
Point the installer to the mirrors
- Windows
- macOS
- Linux
-
Open Notepad and paste this:
[global]
index-url = https://mirrors.aliyun.com/pypi/simple/ -
Save it as
pip.iniin the folderC:\Users\<your user name>\AppData\Roaming\pip. Create thepipfolder if it is not there.
Run this in Terminal:
mkdir -p ~/.config/pip
cat > ~/.config/pip/pip.conf << 'EOF'
[global]
index-url = https://mirrors.aliyun.com/pypi/simple/
EOF
Run this in a terminal:
mkdir -p ~/.config/pip
cat > ~/.config/pip/pip.conf << 'EOF'
[global]
index-url = https://mirrors.aliyun.com/pypi/simple/
EOF
Restart
Restart your computer, open AddaxAI, and run the setup again. It continues where it stopped.
The mirrors are free services from universities and volunteers. They can move or go offline. If one stops working, let me know.
If it still does not work
The steps above are enough for most people. Only if setup still fails, open the part below that matches your problem.
Setup sits at the same percentage for a long time#
On Windows one file is about 3.4 GB. The percentage stays the same while it arrives, but the megabytes below it keep counting up. As long as that number grows, setup is working. On a slow line this one file can take an hour or more.
Do not close and restart the setup. A part-finished file is not kept, so every restart begins that 3.4 GB again.
A model download keeps failing#
You can fetch the model files yourself. Open the model's page on the mirror, download every file, and put them in the model's folder under C:\Users\<your user name>\AddaxAI\models (on macOS and Linux: ~/AddaxAI/models). Keep the file names as they are.
| Model | Files | Folder |
|---|---|---|
| MegaDetector v5A | hf-mirror.com/Addax-Data-Science/MD5A-0-0 | models\det\MD5A-0-0 |
| DINOv2 ViT-S/14 | hf-mirror.com/Addax-Data-Science/DINOV2-VITS14 | models\emb\DINOV2-VITS14 |
| SpeciesNet 4.0.2a | hf-mirror.com/Addax-Data-Science/SPECIESNET-v4-0-2-A | models\cls\SPECIESNET-v4-0-2-A |
Other species models work the same way. Their folder is models\cls\<model id>, with the id as shown in the model zoo.
Still stuck
Email peter@addaxdatascience.com with a diagnostic report, and mention which country or network you are on. I will help you through it.