-
Hello! When I was using your code for training, I found that the metrics for refcoco are always very low, even when reproducing your DINOv2 + SigLIP 384px (Naive Resize) configuration. During finetuni…
-
### What is the issue?
I don't have a problem running codestral so the problem isn't with the model size, right?
```bash
ollama run deepseek-coder-v2
pulling manifest
pulling 5ff0abeeac1d...…
-
Currently we are not support `llava-hf/llava-v1.6-mistral-7b-hf`
```
ValueError: Model type mistral not supported. Currently only 'llama' is supported
```
It would be nice if we support this one…
-
The llava_shared.dll in LLamaSharp.Backend.Cuda12 is only 850KB, the file size is much smaller than llava_shared.dll with cuda. It will take about 126000+ ms to embding an image. Take it to llava_sha…
-
Hello, thank you for your great work!
We are currently exploring the utilization of radio as a vision encoder for vision language models. In our specific setup, we employ [SigClip](https://huggingfac…
-
I found some VLMs are too sensitive to prompt. For example, when I use **mlx-community/llava-1.5-7b-4bit**:
the image is:
![image](https://github.com/Blaizzy/mlx-vlm/assets/72635723/1ab52f9b-085a-47…
cmgzy updated
3 months ago
-
Full repro code here: https://github.com/LLaVA-VL/LLaVA-NeXT/issues/37
```
INFO: 172.16.0.42:27134 - "POST /generate HTTP/1.1" 200 OK
ERROR: Exception in ASGI application
Traceback (m…
-
I don't know if this is a bug or not, just checking. The original llava LLM-image model allows you to send images. llava with llama3 for example, does not allow to send images, while I do think it's t…
-
# Prerequisites
Please answer the following questions for yourself before submitting an issue.
- [x] I am running the latest code. Development is very rapid so there are no tagged versions as of…
-
**LocalAI version:**
2.11.0
**Environment, CPU architecture, OS, and Version:**
Windows 11 latest, Xeon(R) w5-3435X, 256GB, 2x 20GB RTX 4000 NVIDIA-SMI 550.65 Driver Version: 551.86 CUDA Vers…