VictorMylle/Kohya-ss-sd-scripts
1
0
Fork 1
mirror of https://github.com/kohya-ss/sd-scripts.git synced 2026-04-10 15:00:23 +00:00
Code Issues Packages Projects Releases Wiki Activity

make LoRA compatible with ComfyUI #1119

Browse Source
This commit is contained in:
Kohya S
2024-02-25 20:01:37 +09:00
parent 40f2c688db
commit 3a2a48c15d
2 changed files with 10 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
library/stable_cascade.py
View File

@@ -199,7 +199,7 @@ class Attention(nn.Module):
self.to_q = Linear(c, c, bias=True) self.to_q = Linear(c, c, bias=True)
self.to_k = Linear(c, c, bias=True) self.to_k = Linear(c, c, bias=True)
self.to_v = Linear(c, c, bias=True) self.to_v = Linear(c, c, bias=True)
self.to_out = Linear(c, c, bias=True) self.out_proj = Linear(c, c, bias=True)
self.nhead = nhead self.nhead = nhead
self.dropout = dropout self.dropout = dropout
self.scale = (c // nhead) ** -0.5 self.scale = (c // nhead) ** -0.5
@@ -237,7 +237,7 @@ class Attention(nn.Module):
del q, k, v del q, k, v
out = rearrange(out, "(b h) n d -> b n (h d)", h=self.nhead) out = rearrange(out, "(b h) n d -> b n (h d)", h=self.nhead)
return self.to_out(out) return self.out_proj(out)
def _attention(self, query, key, value): def _attention(self, query, key, value):
# if self.upcast_attention: # if self.upcast_attention:

16
library/stable_cascade_utils.py
View File

@@ -196,7 +196,7 @@ def load_previewer_model(previewer_checkpoint_path, dtype=None, device="cpu") ->
def convert_state_dict_mha_to_normal_attn(state_dict): def convert_state_dict_mha_to_normal_attn(state_dict):
# convert nn.MultiheadAttention to to_q/k/v and to_out # convert nn.MultiheadAttention to to_q/k/v and out_proj
print("convert_state_dict_mha_to_normal_attn") print("convert_state_dict_mha_to_normal_attn")
for key in list(state_dict.keys()): for key in list(state_dict.keys()):
if "attention.attn." in key: if "attention.attn." in key:
@@ -214,15 +214,15 @@ def convert_state_dict_mha_to_normal_attn(state_dict):
state_dict[key.replace("in_proj_weight", "to_v.weight")] = qkv[2] state_dict[key.replace("in_proj_weight", "to_v.weight")] = qkv[2]
elif "out_proj.bias" in key: elif "out_proj.bias" in key:
value = state_dict.pop(key) value = state_dict.pop(key)
state_dict[key.replace("out_proj.bias", "to_out.bias")] = value state_dict[key.replace("out_proj.bias", "out_proj.bias")] = value
elif "out_proj.weight" in key: elif "out_proj.weight" in key:
value = state_dict.pop(key) value = state_dict.pop(key)
state_dict[key.replace("out_proj.weight", "to_out.weight")] = value state_dict[key.replace("out_proj.weight", "out_proj.weight")] = value
return state_dict return state_dict
def convert_state_dict_normal_attn_to_mha(state_dict): def convert_state_dict_normal_attn_to_mha(state_dict):
# convert to_q/k/v and to_out to nn.MultiheadAttention # convert to_q/k/v and out_proj to nn.MultiheadAttention
for key in list(state_dict.keys()): for key in list(state_dict.keys()):
if "attention.attn." in key: if "attention.attn." in key:
if "to_q.bias" in key: if "to_q.bias" in key:
@@ -235,12 +235,12 @@ def convert_state_dict_normal_attn_to_mha(state_dict):
k = state_dict.pop(key.replace("to_q.weight", "to_k.weight")) k = state_dict.pop(key.replace("to_q.weight", "to_k.weight"))
v = state_dict.pop(key.replace("to_q.weight", "to_v.weight")) v = state_dict.pop(key.replace("to_q.weight", "to_v.weight"))
state_dict[key.replace("to_q.weight", "in_proj_weight")] = torch.cat([q, k, v]) state_dict[key.replace("to_q.weight", "in_proj_weight")] = torch.cat([q, k, v])
elif "to_out.bias" in key: elif "out_proj.bias" in key:
v = state_dict.pop(key) v = state_dict.pop(key)
state_dict[key.replace("to_out.bias", "out_proj.bias")] = v state_dict[key.replace("out_proj.bias", "out_proj.bias")] = v
elif "to_out.weight" in key: elif "out_proj.weight" in key:
v = state_dict.pop(key) v = state_dict.pop(key)
state_dict[key.replace("to_out.weight", "out_proj.weight")] = v state_dict[key.replace("out_proj.weight", "out_proj.weight")] = v
return state_dict return state_dict