Originally created by @tilllt on GitHub (Mar 5, 2024).
Hey there,
i have several AI Services running (Whisper-API, Auto111/StableDiffusion, TTS-Web-GUI), which all compete for the little 12GB VRAM of our GPU. Open-Webui / Ollama are the most frequently used and it happens quite regularly, that - even if no one is actively using Ollama- there is no VRAM available for the other servies.
It seems that open-webui's detection of when a session is active (keep the model loaded) or not (free the vram) is not working reliably. How is this handled at the moment?
Cheers
Originally created by @tilllt on GitHub (Mar 5, 2024).
Hey there,
i have several AI Services running (Whisper-API, Auto111/StableDiffusion, TTS-Web-GUI), which all compete for the little 12GB VRAM of our GPU. Open-Webui / Ollama are the most frequently used and it happens quite regularly, that - even if no one is actively using Ollama- there is no VRAM available for the other servies.
It seems that open-webui's detection of when a session is active (keep the model loaded) or not (free the vram) is not working reliably. How is this handled at the moment?
Cheers
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @tilllt on GitHub (Mar 5, 2024).
Hey there,
i have several AI Services running (Whisper-API, Auto111/StableDiffusion, TTS-Web-GUI), which all compete for the little 12GB VRAM of our GPU. Open-Webui / Ollama are the most frequently used and it happens quite regularly, that - even if no one is actively using Ollama- there is no VRAM available for the other servies.
It seems that open-webui's detection of when a session is active (keep the model loaded) or not (free the vram) is not working reliably. How is this handled at the moment?
Cheers