[bug]: Lora settings are not preserved on page refresh

[adoermenen]

Is there an existing issue for this problem?

• I have searched the existing issues

Install method

Invoke's Launcher

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

RTX 4090

GPU VRAM

24GB

Version number

v6.6.0

Browser

Edge 139.0.3405.125

System Information

No response

What happened

Lora's added to the concept list and their weights are not preserved on page refresh

What you expected to happen

I expected Lora's in the concept - list to be preserved upon page refresh like most other settings

How to reproduce the problem

1. Select a given Checkpoint/Model in the "Generate"-tab's model-list
2. Add a couple Lora's to the concept list with one of them having a weight less than -1 (due to v6.6.0's -1/+2 weight-limits I need to recall a Lora with weight -2 from a previously generated image for this)
3. Refresh the Invoke-page
4. Note that the previously selected Checkpoint/Model gets restored as expected but that all selected Lora's are gone from the concept list

Additional context

No response

Discord username

No response

[psychedelicious]

It's working as it should for me, and there is no special handling for LoRAs in the state persistence setup.

I think you may just need to wait a moment before refreshing, for the persistence to do its thing. If you refresh too soon, it won't have had time to save the state.

[adoermenen]

@psychedelicious Apologies, it looks like my "steps to reproduce" were inaccurate. I've tested it further here and it looks like this can be reliably reproduced (on my end at least) if the Lora weights are below -1. I think I've been experiencing this issue since before v6.6.0 so it might be unrelated to the -1/+2 Lora weight limits that were introduced with that version

Possibly related console error:
warn [system]: Error rehydrating slice "loras", falling back to default initial state Objecterror: message: "[\n {\n \"origin\": \"number\",\n \"code\": \"too_small\",\n \"minimum\": -1,\n \"inclusive\": true,\n \"path\": [\n \"loras\",\n 3,\n \"weight\"\n ],\n \"message\": \"Too small: expected number to be >=-1\"\n }\n]"name: "ZodError"stack: "ZodError\n at Object.migrate (http://192.168.4.88:9090/assets/index-DdvW16Jl.js:464:40944)\n at sPt (http://192.168.4.88:9090/assets/index-DdvW16Jl.js:530:8442)\n at http://192.168.4.88:9090/assets/index-DdvW16Jl.js:526:2025\n at Array.reduce (<anonymous>)\n at http://192.168.4.88:9090/assets/index-DdvW16Jl.js:526:1961\n at Generator.next (<anonymous>)\n at O7 (http://192.168.4.88:9090/assets/index-DdvW16Jl.js:526:1281)\n at s (http://192.168.4.88:9090/assets/index-DdvW16Jl.js:526:1484)"

[psychedelicious]

Ah, that does help. I see the issue. This bug has been present for a couple months now. Will be fixed in v6.7.0.