Downloading from Hugging Face

2026-02-11 08:08 UTC gpt-5-2-thinking Open in ChatGPT ↗

That repo has a 6.76 GB model.safetensors and it’s stored as xet on the Hub, so the “just click download in the browser / git clone” paths are often where things get annoying. citeturn0view0

Here are the sane ways to download it.

Option A (recommended): Hugging Face CLI

Install the CLI:

pip install -U "huggingface_hub[cli]"

(Optional but recommended) login so downloads are smoother (and required for gated models):

huggingface-cli login

Download the whole repo into a folder:

huggingface-cli download depth-anything/DA3NESTED-GIANT-LARGE-1.1 \
  --local-dir DA3NESTED-GIANT-LARGE-1.1 \
  --revision main

Or download just the big weights:

huggingface-cli download depth-anything/DA3NESTED-GIANT-LARGE-1.1 model.safetensors \
  --local-dir DA3NESTED-GIANT-LARGE-1.1 \
  --revision main

You should end up with:

config.json
README.md
model.safetensors (the big one) citeturn0view0

Option B: Python (best if you’re already in code)

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="depth-anything/DA3NESTED-GIANT-LARGE-1.1",
    revision="main",
    local_dir="DA3NESTED-GIANT-LARGE-1.1",
    local_dir_use_symlinks=False,  # makes a “real” copy in that folder
)

If you only want certain files:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="depth-anything/DA3NESTED-GIANT-LARGE-1.1",
    revision="main",
    local_dir="DA3NESTED-GIANT-LARGE-1.1",
    local_dir_use_symlinks=False,
    allow_patterns=["model.safetensors", "config.json", "README.md"],
)

Option C: Browser (works, but flaky for huge files)

Click model.safetensors → look for the download button. For multi-GB files this can be slow/unstable; CLI/Python is usually much more reliable. citeturn0view0

Why “git clone” often disappoints here

git clone is great when a repo uses git-lfs, but this repo’s big file is labeled xet on the Hub, so the Hub download tools (huggingface-cli download / snapshot_download) are the path of least pain. citeturn0view0

If you tell me your OS (Linux/macOS/Windows) and whether you want it for Transformers loading or just “get the file onto disk”, I’ll give you the exact one-liner + where it ends up on your machine.

You’re on Windows PowerShell, and the CLI entrypoint isn’t on your PATH (or the package isn’t installed in the Python you’re using).

Do this in order:

1) Install the HF Hub CLI into the Python you’re actually using

In PowerShell:

py -m pip install -U "huggingface_hub[cli]"

If py isn’t available, try:

python -m pip install -U "huggingface_hub[cli]"

2) Run the CLI via Python (bypasses PATH issues)

Instead of huggingface-cli ..., run:

py -m huggingface_hub.cli download depth-anything/DA3NESTED-GIANT-LARGE-1.1 `
  --local-dir DA3NESTED-GIANT-LARGE-1.1 `
  --revision main

(That backtick ` is PowerShell line-continuation. You can also put it all on one line.)

Or download only the big weights:

py -m huggingface_hub.cli download depth-anything/DA3NESTED-GIANT-LARGE-1.1 model.safetensors `
  --local-dir DA3NESTED-GIANT-LARGE-1.1 `
  --revision main

py -m huggingface_hub.cli login

If you want `huggingface-cli` to work as a command

Add Python’s Scripts folder to PATH (or use the “App execution aliases” properly), but honestly the py -m ... approach above is the least-baffling and works everywhere.

Quick sanity checks (PowerShell)

See which Python you’re using:

py -0p

Verify the package is installed:

py -m pip show huggingface_hub

If you paste the output of:

py -0p

I can tell you exactly where your Python “Scripts” folder is (the one that would contain huggingface-cli.exe)—but you don’t need it if you use py -m huggingface_hub.cli ....

Yep — that error is because huggingface_hub isn’t meant to be run as python -m huggingface_hub.cli (there’s no __main__ there). On current Hugging Face docs, the CLI command is hf (not huggingface-cli). citeturn1view0turn1view1

Here’s the clean Windows/PowerShell fix.

1) Install / upgrade the package (in your Python 3.14)

py -m pip install -U "huggingface_hub[cli]"

(If this errors, paste the error — but usually it’s fine.)

2) Make sure the `hf` command exists

Run:

Get-Command hf

If it prints a path (something like ...\Python314\Scripts\hf.exe) you’re good.
If it says it can’t find hf, it means your Python Scripts folder isn’t on PATH.

Quick workaround (no PATH changes): run `hf.exe` by full path

Since your Python is: C:\Users\GeorgeAdamon\AppData\Local\Programs\Python\Python314\python.exe

Your Scripts folder is usually: C:\Users\GeorgeAdamon\AppData\Local\Programs\Python\Python314\Scripts\

So try:

& "C:\Users\GeorgeAdamon\AppData\Local\Programs\Python\Python314\Scripts\hf.exe" --help

3) Download that model repo

Whole repo:

hf download depth-anything/DA3NESTED-GIANT-LARGE-1.1 --local-dir .\DA3NESTED-GIANT-LARGE-1.1 --revision main

Just the big weights:

hf download depth-anything/DA3NESTED-GIANT-LARGE-1.1 model.safetensors --local-dir .\DA3NESTED-GIANT-LARGE-1.1 --revision main

(Those commands are exactly what Hugging Face documents for hf download.) citeturn1view0turn1view1

hf auth login

citeturn1view0

If you want, paste the output of:

py -m pip show huggingface_hub
Get-Command hf

…and I’ll tell you exactly what’s missing (install vs PATH) and the shortest fix.