Any-to-Any
Transformers
Safetensors
multilingual
minicpmo
feature-extraction
minicpm-o
omni
vision
ocr
multi-image
video
custom_code
audio
speech
voice cloning
live Streaming
realtime speech conversation
asr
tts
Instructions to use openbmb/MiniCPM-o-2_6 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use openbmb/MiniCPM-o-2_6 with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("openbmb/MiniCPM-o-2_6", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
Supports MacOS in MiniCPM-o 2.6
#19
pinned
by wanderor - opened
Decouples from cuda.
Note: verified in torch 2.5.1. It does not work in torch 2.3.1 on MacOS (specified in requirements).
bokesyo pinned discussion
Thank you for your contribution! Because we only have Mac with 16GB unified memory, so we can't run it on MacOS. However, I ran your modified modeling_minicpmo.py on Linux+CUDA platforms, and there is no problem, and I am happy to merge this PR into main branch, thank you again!
And could you please also add some usage details in README.md, for example, Mac users won't run the code model.cuda() and they need to do model.to('mps')? So that other Mac users can replicate your code, thank you!
bokesyo changed pull request status to merged