Unable to index MLX models on external drive of Mac M3 - Same drive and models work on mlx-engine separately

[nppyro]

Which version of LM Studio?
Example: LM Studio 0.3.10 (Build 6) - For ARM / Apple Silicon

Which operating system?
macOS Sequoia - 15.3.1 (24D70)
MacBookPro
Mac15,6
M3 Pro 18GB

What is the bug?
When pointing my model directory to an external drive, I receive errors with indexing MLX files only GGUF are ok. Same drive and models work on mlx-engine separately.

Failed to index the following folders:
- /Volumes/Lexar 1/LMStudio/Models/mlx-community/DeepSeek-R1-Distill-Llama-8B-4bit

Safetensors header is unexpectedly large: bytes=2199142139136

• Drive is formatted ExFat. File size is not too large like FAT32.
• Checked first 8 bytes and is ok.
  • Only 84357
  • Not matching error of 2199142139136

npitcher@MacBook-Pro-2 DeepSeek-R1-Distill-Llama-8B-4bit % od -An -t u8 -N 8 -v /Volumes/Lexar\ 1/LMStudio/Models/mlx-community/DeepSeek-R1-Distill-Llama-8B-4bit/model.safetensors
                                        84357  

• Was able to do a local test using your mlx-engine and was successful
  • git clone https://github.com/lmstudio-ai/mlx-engine.git
  • Loaded model from external drive without error and received reply

(.venv) npitcher@MacBook-Pro-2 mlx-engine % python demo.py --model /Volumes/Lexar\ 1/LMStudio/Models/mlx-community/DeepSeek-R1-Distill-Llama-8B-4bit 
This is a beta version of the video understanding. It may not work as expected.
[ModelKit][INFO] Loading model from /Volumes/Lexar 1/LMStudio/Models/mlx-community/DeepSeek-R1-Distill-Llama-8B-4bit...
[ModelKit][INFO] Model loaded successfully
Okay, so I need to explain the rules of chess in one sentence. Hmm, where do I start? I know chess is a strategic game involving two players and a board with 64 squares. But how do I condense all the rules into one sentence?

First, I should mention that there are two players: one uses white pieces and the other black. The objective is to checkmate the opponent's king. So, the goal is to capture the king. Got it.

Pieces move in specific ways: pawns move forward one square at a time, rooks move any number of squares in a straight line, bishops move diagonally, knights move in an L-shape, and so on. But I can't list all the piece movements in one sentence. Maybe I can say that each piece has unique movement rules.

Then there's check and checkmate. Check happens when a player's king is under threat of being captured. Checkmate is when the king is in check with no legal moves to escape. So, the goal is to force the opponent into checkmate.

Wait, I need to make sure all these elements are included without being too wordy. Let me try putting it together.

"Chess is a game where two players, using white and black pieces, play to checkmate the opponent's king, with each piece moving according to its unique rules, capturing or checkmating when possible."

Does that cover all the basics? White and black, checkmate, movement. Maybe it's a bit long, but it's concise. I think it works.
</think>

Chess is a game where two players, using white and black pieces, play to checkmate the opponent's king, with each piece moving according to its unique rules, capturing or checkmating when possible.

Generation stats:
 - Tokens per second: 28.88
 - Number of accepted draft tokens: 0
 - Time to first token: 1.25s
 - Total tokens generated: 362
 - Total time: 13.79s

Stopped generation due to: eos_token
Stop string: <｜end▁of▁sentence｜>

• Results of pip install of mlx-engine

Successfully installed aiofiles-23.2.1 aiohappyeyeballs-2.4.4 aiohttp-3.11.11 aiosignal-1.3.2 airportsdata-20241001 annotated-types-0.7.0 anyio-4.8.0 attrs-24.3.0 certifi-2025.1.31 charset-normalizer-3.4.1 click-8.1.8 cloudpickle-3.1.1 datasets-3.3.1 dill-0.3.8 diskcache-5.6.3 fastapi-0.115.8 ffmpy-0.5.0 filelock-3.16.1 frozenlist-1.5.0 fsspec-2024.9.0 genson-1.3.0 gradio-5.16.0 gradio-client-1.7.0 h11-0.14.0 httpcore-1.0.7 httpx-0.28.1 huggingface-hub-0.28.1 idna-3.10 interegular-0.3.3 jinja2-3.1.4 jsonschema-4.23.0 jsonschema-specifications-2024.10.1 lark-1.2.2 markdown-it-py-3.0.0 markupsafe-2.1.5 mdurl-0.1.2 mlx-0.23.1 mlx-lm-0.21.0 mlx-vlm-0.1.13 mpmath-1.3.0 multidict-6.1.0 multiprocess-0.70.16 nest-asyncio-1.6.0 networkx-3.4.2 numpy-2.0.2 opencv-python-4.10.0.84 orjson-3.10.15 outlines-0.1.14 outlines-core-0.1.26 packaging-24.2 pandas-2.2.3 pillow-11.1.0 propcache-0.2.1 protobuf-5.29.3 pyarrow-19.0.0 pycountry-24.6.1 pydantic-2.10.6 pydantic-core-2.27.2 pydub-0.25.1 pygments-2.19.1 python-dateutil-2.9.0.post0 python-multipart-0.0.20 pytz-2025.1 pyyaml-6.0.2 referencing-0.36.2 regex-2024.11.6 requests-2.32.3 rich-13.9.4 rpds-py-0.22.3 ruff-0.9.6 safehttpx-0.1.6 safetensors-0.5.2 scipy-1.13.1 semantic-version-2.10.0 sentencepiece-0.2.0 setuptools-75.6.0 shellingham-1.5.4 six-1.17.0 sniffio-1.3.1 starlette-0.45.3 sympy-1.13.1 tokenizers-0.21.0 tomlkit-0.13.2 torch-2.5.1 tqdm-4.67.1 transformers-4.49.0.dev0 typer-0.15.1 typing-extensions-4.12.2 tzdata-2025.1 urllib3-2.3.0 uvicorn-0.34.0 websockets-14.2 xxhash-3.5.0 yarl-1.18.3

[nppyro]

Similar to issue 79 on mlx-engine, but filed here since it works on mlx-engine alone.

lmstudio-ai/mlx-engine#79

They resolved by formatting as APFS, but I need cross compatibility with these same files on Windows and Linux.