How to deploy LTX2.3 as a service in ubuntu system?
I use SGLang to deploy LTX2.3, but do I need the configuration files of LTX2.3 when starting up, such as model_index.json, etc.?
How much memory is needed? I have 64G of memory and 64GB of swap memory. However, the error -9 is still reported during startup. Does this mean there is insufficient memory? Is there an official document? I'm not using comfyui to deploy, I want to deploy it as a service
We don't have official documentation for SGLang support. If there's enough demand we can look at adding that.
Error -9 is an OOM so yes, that's not enough memory. Which version of the model checkpoint are you using? If it's not Lightricks/LTX-2.3-fp8, give that a try.
It's Lightricks/LTX-2.3-fp8. Even if SG-Lang is not applicable, the normal startup error is -9 Not enough memory. I use 64G memory + 100G swap space, 24G video memory, and still get error -9. It cannot be started without SG-Lang. Does LTX2.3 need to use the model_index.json of LTX2.0 and some other basic things, such as audio_vae, connectors, text_encoder, tokenizer, etc. I really can’t find a way to use pyhton3 to run LTX2.3 directly on the Internet. Can you give me a tutorial?
This is what I did. I downloaded the basic version of LTX2.0, excluding the weight file, and then used the weight file of Lightricks/LTX-2.3-fp8 to put it in the root directory. Is this right?