CLIPPyX Install error: MLX

PyTorch install works fine, and the clippyx install seems to work until hitting the below error. Tried with both Python 3.13 and 3.12, but no difference

INFO: pip is looking at multiple versions of mlx-embeddings to determine which version is compatible with other requirements. This could take a while. ERROR: Could not find a version that satisfies the requirement mlx>=0.16.3 (from mlx-embeddings) (from versions: none) ERROR: No matching distribution found for mlx>=0.16.3

May 12 '25 23:05 Relysis

hi, I'm finalizing the MLX version soon. I will fix these issues and provide a pytorch free version for macs! Just give me some time

May 16 '25 00:05 0ssamaak0

Running into this issue on windows - which seems strange.

Jul 22 '25 15:07 Prettyprettygood

My issue is on windows as well. Thanks 0ssamaak0 for the work

Aug 09 '25 04:08 Relysis

You’re hitting this because mlx only exists for macOS + Apple Silicon. On Windows/Linux (or Intel Mac) pip will always fail with No matching distribution.

👉 Fix = skip MLX and use Hugging Face models instead:

Fresh venv (Windows example)

python -m venv C:\v\clx
C:\v\clx\Scripts\activate
python -m pip install -U pip

Install without MLX deps

pip install -e . --no-deps
findstr /V /I "mlx" requirements.txt > requirements.win.txt
pip install -r requirements.win.txt
pip install -U transformers sentencepiece pillow pyyaml

Then in config.yaml set:

clip:
  provider: HF_transformers
  HF_transformers_clip: openai/clip-vit-base-patch32

text_embed:
  provider: HF_transformers
  HF_transformers_embeddings: sentence-transformers/all-MiniLM-L6-v2

Run with:

CLIPPyX

✅ This avoids MLX and works fully offline.

Sep 06 '25 17:09 oceanondawave

I've hit the same issue on a Nvidia DGX Spark. I tried @oceanondawave 's solution without success. Several install were missing and after installing most of them successfully, I finally got back to pretty much the first issue:

CLIPPyX$ pip install -r requirements.txt
[... succesfull installs ...]
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
clippyx 1.0.1 requires mlx-embeddings==0.0.3, which is not installed.
clippyx 1.0.1 requires ocrmac==1.0.0, which is not installed.
clippyx 1.0.1 requires chromadb==1.0.0, but you have chromadb 0.5.0 which is incompatible.
Successfully installed Flask-3.1.2 PyYAML-6.0.1 aiohappyeyeballs-2.6.1 aiohttp-3.13.2 aiosignal-1.4.0 anyascii-0.3.3 async-timeout-5.0.1 blinker-1.9.0 braceexpand-0.1.7 chroma-hnswlib-0.7.3 chromadb-0.5.0 clip-benchmark-1.6.2 datasets-4.3.0 defusedxml-0.7.1 dill-0.4.0 einops-0.8.0 flask-cors-6.0.1 frozenlist-1.8.0 fsspec-2025.9.0 ftfy-6.3.1 h5py-3.15.1 itsdangerous-2.2.0 jinja2-3.1.6 joblib-1.5.2 langdetect-1.0.9 markupsafe-3.0.3 ml_dtypes-0.5.3 multidict-6.7.0 multiprocess-0.70.16 networkx-3.4.2 numpy-1.26.4 onnx-1.19.1 open-clip-torch-3.2.0 opencv-python-4.11.0.86 pandas-2.3.3 pillow-11.3.0 propcache-0.4.1 pyarrow-22.0.0 pyclipper-1.3.0.post6 pycocoevalcap-1.2 pycocotools-2.0.10 pypdfium2-4.30.0 python-doctr-1.0.0 pytz-2025.2 rapidfuzz-3.14.3 scikit-learn-1.7.2 scipy-1.15.3 shapely-2.1.2 threadpoolctl-3.6.0 timm-1.0.21 torch-2.9.0 torchvision-0.24.0 tqdm-4.66.4 ttkbootstrap-1.18.0 tzdata-2025.2 validators-0.35.0 wcwidth-0.2.14 webdataset-1.0.2 werkzeug-3.1.3 xxhash-3.6.0 yarl-1.22.0

Has anyone succeeded in installing on a Nvidia DGX Spark system?

Nov 03 '25 14:11 miles-3m