Error installing with --no-torch option

[titan581]

Hi
I tried to install but i got some errors, any help is appreciated:
PS D:\unslothStudio\unsloth> .\install.ps1 --local --no-torch

🦥 Unsloth Studio Installer (Windows)
────────────────────────────────────────────────────

python Python 3.13 already installed
preserving existing environment for rollback...
previous environment preserved for rollback
venv creating Python 3.13 virtual environment
C:\Users\Admin.unsloth\studio\unsloth_studio
gpu NVIDIA GPU detected
skipping PyTorch (--no-torch flag set).
installing unsloth (this may take a few minutes)...
overlaying local repo (editable)...
overlaying unsloth-zoo from git main...
setup running unsloth studio setup...
Traceback (most recent call last):
File "", line 198, in run_module_as_main
File "", line 88, in run_code
File "C:\Users\Admin.unsloth\studio\unsloth_studio\Scripts\unsloth.exe_main.py", line 4, in
from unsloth_cli import app
File "D:\unslothStudio\unsloth\unsloth_cli_init.py", line 4, in
import typer
File "C:\Users\Admin.unsloth\studio\unsloth_studio\Lib\site-packages\typer_init_.py", line 7, in
from click.exceptions import Abort as Abort
ModuleNotFoundError: No module named 'click'
[ERROR] unsloth studio setup failed (exit code 1)
restoring previous environment after failed install...
restored previous environment

[rolandtannous]

Hello,

This should have been fixed on 07-05-2026 in the main repo (check the commits list at the bottom). The no-torch path was missing the full typer dependency chain (click, shellingham and some other packages) which is what you're hitting. You might have pulled the repository before the fix was merged.

Do you mind pulling the latest main branch and trying again:

cd D:\unslothStudio\unsloth
Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
.\install.ps1 --local --no-torch

Let us know if you still see the error after pulling.

Commits:

• 13f79e4
• 8bc52ff1
• c84ff191
• f813403c

[titan581]

Hello,
Thank you.
I run following commands:
PS D:\unslothStudio\unsloth> git pull origin main
remote: Enumerating objects: 156, done.
remote: Counting objects: 100% (112/112), done.
remote: Compressing objects: 100% (72/72), done.
remote: Total 156 (delta 63), reused 40 (delta 40), pack-reused 44 (from 1)
Receiving objects: 100% (156/156), 449.36 KiB | 1.04 MiB/s, done.
Resolving deltas: 100% (70/70), completed with 9 local objects.
From https://github.com/unslothai/unsloth

• branch main -> FETCH_HEAD
  b364080..a21e2d8 main -> origin/main
  Updating b364080..a21e2d8
  Fast-forward
  .github/dependabot.yml | 48 +
  .
  .
  .
  unsloth_cli/commands/studio.py | 55 +-
  84 files changed, 26854 insertions(+), 269 deletions(-)
  create mode 100644 .github/workflows/consolidated-tests-ci.yml
  .
  .
  .
  create mode 100644 tests/vllm_compat/test_vllm_pinned_symbols.py
  PS D:\unslothStudio\unsloth> Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass
  PS D:\unslothStudio\unsloth> .\install.ps1 --local --no-torch

🦥 Unsloth Studio Installer (Windows)
────────────────────────────────────────────────────

python Python 3.13 already installed
preserving existing environment for rollback...
previous environment preserved for rollback
venv creating Python 3.13 virtual environment
C:\Users\Admin.unsloth\studio\unsloth_studio
gpu NVIDIA GPU detected
skipping PyTorch (--no-torch flag set).
installing unsloth (this may take a few minutes)...
overlaying local repo (editable)...
overlaying unsloth-zoo from git main...
setup running unsloth studio setup...

🦥 Unsloth Studio Setup
────────────────────────────────────────────────────
gpu NVIDIA GPU detected
long paths enabled
git git version 2.54.0.windows.1
cmake cmake version 4.3.2
vs Visual Studio 17 2022 (filesystem (BuildTools))
cl.exe: C:\Program Files (x86)\Microsoft Visual
Studio\2022\BuildTools\VC\Tools\MSVC\14.44.35207\bin\Hostx64\x64\cl.exe
driver supports up to CUDA 13.2
GPU Compute Capability = 12.0 (sm_120)
using existing CUDA Toolkit at CUDA_PATH (nvcc: C:\Program Files\NVIDIA GPU Computing
Toolkit\CUDA\v13.0\bin\nvcc.exe)
Persisted CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0 to user environment
Set CUDA_PATH_V13_0 (cleared other CUDA_PATH_V* vars)
cuda C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0\bin\nvcc.exe
CUDA_PATH = C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0
CudaToolkitDir = C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v13.0
Node v24.11.1 and npm 11.6.2 already meet requirements.
node v24.11.1 | npm 11.6.2
bun already installed (1.3.13)
Python 3.13.0

system prerequisites ready

             Frontend source changed since last build -- rebuilding...

             building frontend...

Using bun for package install (faster)
frontend built
installing OXC validator runtime...
oxc runtime installed

             setting up Python environment...
             Using C:\Program Files\Python313\python.exe (Python 3.13.0)
             reusing existing virtual environment at C:\Users\Admin\.unsloth\studio\unsloth_studio
             TORCHINDUCTOR_CACHE_DIR set to C:\tc (avoids MAX_PATH issues)
             installing PyTorch with CUDA support (cu130)...
             (This download is ~2.8 GB -- may take a few minutes)

But it wants now to download PyTorch, even it wrote above " skipping PyTorch (--no-torch flag set)."