Thireus ☠
Thireus ☠
I'd like to know as well.
Would be awesome to have!
Code examples are there: - Images: https://github.com/turboderp-org/exllamav2/blob/master/examples/multimodal.py - Videos: https://github.com/turboderp-org/exllamav2/blob/master/examples/multimodal_video.py Looks like it is a matter of calling `ExLlamaV2VisionTower` in modules/exllamav2_hf.py and modules/exllamav2.py with the appropriate conf. When the images/videos...
Have you been able to solve this issue?
Just curious to know if there's been any progress with supporting V1. Thank you!
Here is how to upgrade to v1: ``` # Upgrade Kokoro to 82M v1.0 uv pip install --upgrade kokoro-onnx uv pip install --upgrade onnxruntime onnxruntime-gpu export KOKORO_REVISION=f3ff3571791e39611d31c381e3a41a3af07b4987 huggingface-cli download hexgrad/Kokoro-82M...
Also need to do `uv pip install --upgrade onnxruntime onnxruntime-gpu` for GPU support.
@Alelau18, could you explain how you've obtained/compiled these? Here's what worked for me: Compile flash attention on Linux/WSL: ``` # IMPORTANT: Execute the first portion of the wsl.sh script up...
@JakeMadsen follow these steps: https://github.com/oobabooga/text-generation-webui/issues/6729#issuecomment-2725225001
By default the client app will convert the image. I propose to add an additional option in the settings, we already have "high" quality, we should have "original" quality as...