Playing with LTX-2

The second issue is of less importance overall. I've just been playing around with the LTX-2 release just because I can. I have an RTX5090 32GB running on an MSI Infinity box 64GBRAM so I'm just experimenting to see what she will do. (Running it in FP8 KJ mode to prevent OOM possibility).

One workflow I am trying to run contains the CUI-core node "Load Latent Upscale Model". The workflow creator, as usual, has the slot populated with his suggested model. I downloaded it, refreshed and tried the drop down window method of selection, nothing dropped. Hitting the arrows resulted in a NULL as if the node wasn't functioning as it should by seeing the folder with the model. I switched the model to different folders but nothing ever drops down. Dead!

As I mentioned this LTX-2 thing is just experimenting so not as important as my first issue. A number of other attempts with CUI workflows for LTX-2 produced OOM and I had given up. This workflow with quantized down models looked promising. Because of the non-loading issue, I disabled the node and it was humming along until it got to that point and stopped so obviously that node is crucial. But it can't be populated with the model. By the way, I dropped this node in a couple of other unrelated workflows and tried to populate it; same issue so it isn't the original workflow. And the creator doesn't understand why since his works and he's been praised for it by others.

Issue here: CUI front end issue mentioned in last post? Just broken? Can it be fixed?

Any insights will be much appreciated.

┆Issue is synchronized with this [Notion page](https://www.notion.so/Issue-7960-Playing-with-LTX-2-2e56d73d365081909febc5bb298435a8) by [Unito](https://www.unito.io)