Dillon Laird

I found it easier to use the DINOv2 model provided in mmpretrain [here](https://github.com/open-mmlab/mmpretrain/tree/main/configs/dinov2). You can load it with something like the following config: ``` pretrained = 'https://download.openmmlab.com/mmpretrain/v1.0/dinov2/vit-small-p14_dinov2-pre_3rdparty_20230426-5641ca5a.pth' model = dict(...

Python 3.8 is the first version of Python that implements SharedMemory, so you should be using the official SharedMemory class instead of the one built in this repo https://docs.python.org/3/library/multiprocessing.shared_memory.html I...

ViT-Adapter wraps around the DINOv2 model with injector and extractor modules, see the paper [here](https://arxiv.org/abs/2205.08534) so all you need to do is build the ViTAdapter model from [here](https://github.com/facebookresearch/dinov2/blob/main/dinov2/eval/segmentation_m2f/models/backbones/vit_adapter.py) and pass...

It's run using MMLab, specifically [MMSegmentation](https://github.com/open-mmlab/mmsegmentation/tree/main). You can follow the notebook [here](https://github.com/facebookresearch/dinov2/blob/main/notebooks/semantic_segmentation.ipynb) to load the mmsegmentation config file used to run the model. You may have to modify some of...

Got it, thank you! Do you also know if there's a straight forward way to load these using MMSeg/MMDet instead of detectron2?

You can download the tokenizer.model from one of the provided links, for example this is the one used for LLaMA-7B https://huggingface.co/nyanko7/LLaMA-7B/tree/main That file tree is not correct, tokenzier.model is a...

Hey @Eric-Canas, glad you are finding it useful! Good point, I don't have much time to look into it now but I'll add it to my list. I believe most...

You're right, this is unfortunate. The models are always prompted to output code using XML tags, like here https://github.com/landing-ai/vision-agent/blob/main/vision_agent/agent/vision_agent_planner_prompts_v2.py#L557-L559 but now it seems GPT-4o ignores that instruction and use Markdown...

I heard they updated the same endpoint without versioning it, https://community.openai.com/t/was-anyone-elses-experience-with-gpt4o-completely-ruined-after-recent-update/1107600 I wonder if this is related

It seems like it's more variable. I ran this test 3 times: ```python In [4]: lmm([{"role": "user", "content": "Write some code to add two numbers. Respond with your code in...