Finegrain/refiners
1
0
Fork 0
mirror of https://github.com/finegrain-ai/refiners.git synced 2024-11-21 21:58:47 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Update src/refiners/foundationals/latent_diffusion/lora.py

Browse source 
Co-authored-by: Laureηt <laurent@lagon.tech>
This commit is contained in:
Pierre Chapuis 2024-02-07 16:00:56 +01:00
parent b1c200c63a
commit 4d85918336
1 changed files with 15 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
src/refiners/foundationals/latent_diffusion/lora.py
View file

@ -235,6 +235,21 @@ class SDLoraManager:
@staticmethod @staticmethod
def sort_keys(key: str, /) -> tuple[str, int]: def sort_keys(key: str, /) -> tuple[str, int]:
"""Compute the score of a key, relatively to its suffix.
When used by [`sorted`][sorted], the keys will only be sorted "at the suffix level".
The idea is that sometimes closely related keys in the state dict are not in the
same order as the one we expect, for instance `q -> k -> v` or `in -> out`. This
attempts to fix that issue, not cases where distant layers are called in a different
order.
Args:
key: The key to sort.
Returns:
The padded prefix of the key.
A score depending on the key's suffix.
"""
# this dict might not be exhaustive # this dict might not be exhaustive
suffix_scores = {"q": 1, "k": 2, "v": 3, "in": 3, "out": 4, "out0": 4, "out_0": 4} suffix_scores = {"q": 1, "k": 2, "v": 3, "in": 3, "out": 4, "out0": 4, "out_0": 4}
patterns = ["_{}", "_{}_lora"] patterns = ["_{}", "_{}_lora"]