DiffusionToolkit
DiffusionToolkit copied to clipboard
[BUG] Model name "not found"
Model hash and name is shown in the generation parameters, but below on the model name line I get "not found". I did a model data rebuild, but problem is still there.
Version: 1.2.1
Yep. This also happens with v1.3. I made sure the cache json file had the hash saved. I'm using SD.Next.
Not found (94f45bf623)
"checkpoint/realcartoonRealistic_v8.safetensors": {
"mtime": 1694898626.5027108,
"sha256": "94f45bf6232fc2bb44b106daae23a911c452b8613dae3c3e6c51458cae0a5486"
},
This may be the case if the ModelRoot Path is not set. I was primarily using the models to get the (original) hash, then supplementing with cache.json for the SHA256 hash.
I've changed it so that one is not dependent on the other.
But i do have the model root set (the cache.json you see is the actual file and is symlinked in the SD folder).
I think the fix in 1.4 is not quite right. Now sometimes i get the model repeated:
Two of those copies are probably because the model is repeated in my cache.json file.
yeah, there are more sources I'm pulling from, and sometimes they could be different files so I haven't really bothered to deal with them just yet. There still maybe some other bugs lurking around.
Do you also use the CivitAI models info feature?
I don't think i have used the "Download Civitai models" option.
Just wondering if this is still an issue with the latest build
In v1.5 i don't see any model name, even after downloading civitai models. The model is in models.json with the correct hash, so the only thing i can think of is that it is doing a case sensitive check on the hash (json has it in uppercase but hash is shown in lowercase in the interface).
Sorry, yes 1.5 had a bug where the model name was missing. It's fixed now.
I'm using a caseless match on model hashes when loading the model names. Can you give me an image and a copy of your cache json? It would help identifying the issue.
v1.5.1 fixed the model name for me.