nodes/city96-ComfyUI_ExtraModels
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Convert some missing layers (#72)

Browse Source
This commit is contained in:
frutiemax92
2024-07-02 01:23:52 +02:00
committed by GitHub
parent faf6979f34
commit e866b5b6cb
1 changed files with 22 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
PixArt/diffusers_convert.py
+22 -1
View File
@@ -19,6 +19,9 @@ conversion_map_ms = [ # for multi_scale_train (MS)
def get_depth(state_dict):
return sum(key.endswith('.attn1.to_k.bias') for key in state_dict.keys())
def get_lora_depth(state_dict):
return sum(key.endswith('.attn1.to_k.lora_A.weight') for key in state_dict.keys())
def get_conversion_map(state_dict):
conversion_map = [ # main SD conversion map (PixArt reference, HF Diffusers)
# Patch embeddings
@@ -150,8 +153,9 @@ def convert_lora_state_dict(state_dict, peft=True):
new_state_dict = {k: state_dict[v] for k,v in cmap if k not in missing}
matched = list(v for k,v in cmap if v in state_dict.keys())
lora_depth = get_lora_depth(state_dict)
for fp, fk in ((rep_ap, rep_ak),(rep_bp, rep_bk)):
for depth in range(get_depth(state_dict)):
for depth in range(lora_depth):
# Self Attention
key = lambda a: fp(f"transformer_blocks.{depth}.attn1.to_{a}.weight")
new_state_dict[fk(f"blocks.{depth}.attn.qkv.weight")] = torch.cat((
@@ -164,6 +168,11 @@ def convert_lora_state_dict(state_dict, peft=True):
new_state_dict[rep_pk((f"blocks.{depth}.attn.qkv.weight"))] = state_dict[akey("q")]
matched += [akey('q'), akey('k'), akey('v')]
# Self Attention projection?
key = lambda a: fp(f"transformer_blocks.{depth}.attn1.to_{a}.weight")
new_state_dict[fk(f"blocks.{depth}.attn.proj.weight")] = state_dict[key('out.0')]
matched += [key('out.0')]
# Cross-attention (linear)
key = lambda a: fp(f"transformer_blocks.{depth}.attn2.to_{a}.weight")
new_state_dict[fk(f"blocks.{depth}.cross_attn.q_linear.weight")] = state_dict[key('q')]
@@ -177,6 +186,18 @@ def convert_lora_state_dict(state_dict, peft=True):
new_state_dict[rep_pk((f"blocks.{depth}.cross_attn.kv_linear.weight"))] = state_dict[akey("k")]
matched += [akey('q'), akey('k'), akey('v')]
# Cross Attention projection?
key = lambda a: fp(f"transformer_blocks.{depth}.attn2.to_{a}.weight")
new_state_dict[fk(f"blocks.{depth}.cross_attn.proj.weight")] = state_dict[key('out.0')]
matched += [key('out.0')]
key = fp(f"transformer_blocks.{depth}.ff.net.0.proj.weight")
new_state_dict[fk(f"blocks.{depth}.mlp.fc1.weight")] = state_dict[key]
matched += [key]
key = fp(f"transformer_blocks.{depth}.ff.net.2.weight")
new_state_dict[fk(f"blocks.{depth}.mlp.fc2.weight")] = state_dict[key]
matched += [key]
if len(matched) < len(state_dict):
print(f"PixArt: LoRA conversion has leftover keys! ({len(matched)} vs {len(state_dict)})")