Aeontropy on Nostr: >SD3 will be censored yes >it's over not really This is the only model from SAI that ...
>SD3 will be censored
yes
>it's over
not really
This is the only model from SAI that gives me hope
The DiT architecture is really robust and it learns faster and easier, check out PixART which is using the same DiT architecture and they trained it on the same level as SD 1.5 in only 10% of the training time of 1.5
Still good starting points to train lewds on. unless it's purely online in which case who cares, there are a trillion models on civit and plenty of github projects in the works.
Published at
2024-02-24 19:19:23Event JSON
{
"id": "6daa5d92cf510461238886af8374095e34a7ec99e202d017f772954e37093183",
"pubkey": "b75b9a3131f4263add94ba20beb352a11032684f2dac07a7e1af827c6f3c1505",
"created_at": 1708802363,
"kind": 1,
"tags": [],
"content": "https://image.nostr.build/1c4c1fc0206db1df70f73180e4e89be8fb04971a1ef85c64ae2f7b2e2c439df8.jpg\n\nhttps://image.nostr.build/181614064ddb22f7084c2c24c0c933334ed3dc7fe9351f950729c90cf3f88620.jpg\n\n\u003eSD3 will be censored\nyes\n\u003eit's over\nnot really\n\nThis is the only model from SAI that gives me hope\nThe DiT architecture is really robust and it learns faster and easier, check out PixART which is using the same DiT architecture and they trained it on the same level as SD 1.5 in only 10% of the training time of 1.5\n\nStill good starting points to train lewds on. unless it's purely online in which case who cares, there are a trillion models on civit and plenty of github projects in the works.",
"sig": "9acdbcb3edadec5297b33544575747c18e03d09ec39c1d606efbec26c4d4a233c7346e49ea1e00de54fd8ead50b74c19b5178b2f666ef99b3b6084e2479ac106"
}
