Max Lund

``` └─ $ ▶ pytest tests/test_pymediainfo.py::MediaInfoUnicodeFileNameTest::test_parse_unicode_file =============================================================================================== test session starts ================================================================================================ platform darwin -- Python 3.10.9, pytest-8.4.0, pluggy-1.6.0 -- /usr/local/bin/python3 cachedir: .pytest_cache rootdir: /Users/maxlund/pymediainfo configfile: pyproject.toml plugins: anyio-4.4.0 collected 1...

@sbraz Sorry for the late reply, I missed your message somehow - yes, this fixes things: ``` lib, handle = MediaInfo._get_library(MEDIAINFO_LIB_PATH)[0:2] lib.MediaInfo_Option(handle, "setlocale_LC_CTYPE", "") lib.MediaInfo_Option(handle, "CharSet", "UTF-8") ``` I also...

Looking at the code, I would want to avoid this though I suppose? ``` # Reset all options to their defaults so that they aren't # retained when the parse...

> If you don't do call `parse()` several times in multiple threads, you should be fine. Understood, but we could have situations where that happens. But using a runtime hook...

Correct, SigLIP2 would be the one. Any resource(s) you could point me towards for converting models and running them on mlx?

@Blaizzy guessing we could figure it out from here [mlx_vlm/models/paligemma/vision.py ](https://github.com/Blaizzy/mlx-vlm/blob/c70c916ac64b22d9105d68ed9e507bf0452ea145/mlx_vlm/models/paligemma/vision.py), will the conversion scripts in the comments work for SigLIP2? ```` from transformers import AutoModelForCausalLM, AutoProcessor model_id= "" model...

Found the issue where people had issues converting SigLIP models to MLX: https://github.com/ml-explore/mlx-examples/issues/747

Amazing! Will be very interesting to see what kind of speedups we can get compared to MPS/PyTorch. Been playing around with the mexma-siglip2 variant recently which has shown very good...

> [@maxlund](https://github.com/maxlund) looks like it might be possible to add with a model config and no code changes, though not quite sure about the text pooling https://huggingface.co/visheratin/mexma-siglip2/blob/cc7d73c89f452514ec3b996528543d026ad58f72/mexma_siglip.py#L70-L76 > > It's...

``` import os import time from typing import Generator, Optional import cv2 import mlx.core as mx import numpy as np import torch from PIL import Image from mlx_embeddings.utils import load...