Change in method used to determine Checkpoint Hash?

[mistermcluvin]

Question.
Was there a change on how Forge determines the Checkpoint hash?
I noticed when I post locally generated pictures on Civitai with the current version it can no longer associate it to the checkpoint? When I compared my current generation data to my older version of Forge I noticed the checkpoint hashes do not match?

Example, Cyberrealistic V50 (SD1.5), same model file (shared folder).

Current Version of Forge:
cyberrealistic_v50
Description
Filename: 0.000_01\cyberrealistic_v50.safetensors
File size: 4GB
Hash: 321b1198d8
Modified: 2024-05-29 10:04

Older Version of Forge (version: [f0.0.17v1.8.0rc-latest-276-g29be1da7]).
cyberrealistic_v50
Description
Filename: 0.000_01\cyberrealistic_v50.safetensors
File size: 4GB
Hash: 0b20a3f2cd
Modified: 2024-05-29 10:04

It's really not a huge deal but I was just curious.
Thanks!

[sunny-sko]

i replaced folder \cache from AUTOMATIC1111, and it help'd only in part,
for forge rehash some loras/chpnts anyway

[sunny-sko]

lil' correction: forge writes 3 lists of hashes in file:

Steps: 24, Sampler: IPNDM, Schedule type: Beta, CFG scale: 1, Distilled CFG Scale: 3.5, Seed: 3129685724, Size: 896x1152,

Model hash: 4610115bb0, Model: flux1-dev, RNG: NV, Style Selector Enabled: True,
Lora hashes: " - Flux1 - vanta_black_V1.0: ea6991ec31ba, ImageUpgraderV2: e5598e0ef11d, Dever_Flux_Enhancer: 6be64a302443, FluxDFaeTasticDetails: 5be8e41a3a18, Simon_Stålenhag_Flux: 5497055a3e2d, ink_monsters: e3b0c44298fc",

Beta schedule alpha: 0.6, Beta schedule beta: 0.6, Version: f2.0.1v1.10.1-previous-442-g230e3911, Module 1: clip_l, Module 2: ae, Module 3: t5xxl_fp16,

Hashes: {"lora:Dever_Flux_Enhancer": "471ee7052a", "lora:FluxDFaeTasticDetails": "4259dc3287", "lora:ink_monsters": "a8079603fe", "model": "4610115bb0"}

and only third one is "correct" but civitai seems to reads from first and second, and not all of 'em are