iamthemulti

ProTip: ENABLE_EMAIL_AUTOCONFIRM being set to false in the .env example is also an issue for any self hosters out there who don't have an email sending setup.

@irexyc  Any image will do, I pulled one off of google, as long as it is saved and loaded in a file path. Here is as much of the...

Here I was able to capture it while running --log-level DEBUG if that helps [log.txt](https://github.com/user-attachments/files/17302811/log.txt)

Here is the config for the unquantized model: [orig_config.json](https://github.com/user-attachments/files/17309929/orig_config.json) Here is the config for the lmdeploy AWQ model: [config.json](https://github.com/user-attachments/files/17309952/config.json) I changed architectures from `ChatGLMForConditionalGeneration` to be `ChatGLMModel`, and ran serve....

Thanks for taking a look. This being a VRAM issue seems unusual considering I can run inference on this image successfully using the sample code in the GLM-4V repo, and...