ostris/ai-toolkit
1
0
Fork 0
mirror of https://github.com/ostris/ai-toolkit.git synced 2026-01-26 16:39:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed some mismatched weights by adjusting tolerance. The mismatch ironically made the models better lol

Browse Source
This commit is contained in:
Jaret Burkett
2023-08-29 15:20:03 -06:00
parent 14ff51ceb4
commit 836fee47a6
8 changed files with 265 additions and 163 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
jobs/process/BaseSDTrainProcess.py
View File

@@ -481,8 +481,33 @@ class BaseSDTrainProcess(BaseTrainProcess):
params = self.embedding.get_trainable_params()
else:
# set them to train or not
if self.train_config.train_unet:
self.sd.unet.requires_grad_(True)
self.sd.unet.train()
else:
self.sd.unet.requires_grad_(False)
self.sd.unet.eval()
if self.train_config.train_text_encoder:
if isinstance(self.sd.text_encoder, list):
for te in self.sd.text_encoder:
te.requires_grad_(True)
te.train()
else:
self.sd.text_encoder.requires_grad_(True)
self.sd.text_encoder.train()
else:
if isinstance(self.sd.text_encoder, list):
for te in self.sd.text_encoder:
te.requires_grad_(False)
te.eval()
else:
self.sd.text_encoder.requires_grad_(False)
self.sd.text_encoder.eval()
# will only return savable weights and ones with grad
params = self.sd.prepare_optimizer_params(
vae=False,
unet=self.train_config.train_unet,
text_encoder=self.train_config.train_text_encoder,
text_encoder_lr=self.train_config.lr,

2
testing/generate_weight_mappings.py
View File

@@ -93,7 +93,7 @@ total_keys = len(ldm_dict_keys)
matched_ldm_keys = []
matched_diffusers_keys = []
error_margin = 1e-4
error_margin = 1e-6
tmp_merge_key = "TMP___MERGE"

29
testing/test_model_load_save.py
View File

@@ -3,6 +3,8 @@ import os
# add project root to sys path
import sys
from tqdm import tqdm
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
import torch
@@ -20,6 +22,8 @@ from toolkit.stable_diffusion_model import StableDiffusion
# you probably wont need this. Unless they change them.... again... again
# on second thought, you probably will
project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
device = torch.device('cpu')
dtype = torch.float32
@@ -109,7 +113,26 @@ keys_in_both.sort()
if len(keys_not_in_state_dict_2) == 0 and len(keys_not_in_state_dict_1) == 0:
print("All keys match!")
print("Checking values...")
mismatch_keys = []
loss = torch.nn.MSELoss()
tolerance = 1e-6
for key in tqdm(keys_in_both):
if loss(state_dict_file_1[key], state_dict_file_2[key]) > tolerance:
print(f"Values for key {key} don't match!")
print(f"Loss: {loss(state_dict_file_1[key], state_dict_file_2[key])}")
mismatch_keys.append(key)
if len(mismatch_keys) == 0:
print("All values match!")
else:
print("Some valued font match!")
print(mismatch_keys)
mismatched_path = os.path.join(project_root, 'config', 'mismatch.json')
with open(mismatched_path, 'w') as f:
f.write(json.dumps(mismatch_keys, indent=4))
exit(0)
else:
print("Keys don't match!, generating info...")
@@ -132,17 +155,17 @@ for key in keys_not_in_state_dict_2:
# print(json_data)
project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
json_save_path = os.path.join(project_root, 'config', 'keys.json')
json_matched_save_path = os.path.join(project_root, 'config', 'matched.json')
json_duped_save_path = os.path.join(project_root, 'config', 'duped.json')
state_dict_1_filename = os.path.basename(args.file_1[0])
state_dict_2_filename = os.path.basename(args.file_2[0])
# state_dict_2_filename = os.path.basename(args.file_2[0])
# save key names for each in own file
with open(os.path.join(project_root, 'config', f'{state_dict_1_filename}.json'), 'w') as f:
f.write(json.dumps(state_dict_1_keys, indent=4))
with open(os.path.join(project_root, 'config', f'{state_dict_2_filename}.json'), 'w') as f:
with open(os.path.join(project_root, 'config', f'{state_dict_1_filename}_loop.json'), 'w') as f:
f.write(json.dumps(state_dict_2_keys, indent=4))
with open(json_save_path, 'w') as f:

2
toolkit/config_modules.py
View File

@@ -99,6 +99,8 @@ class ModelConfig:
self.use_text_encoder_1: bool = kwargs.get('use_text_encoder_1', True)
self.use_text_encoder_2: bool = kwargs.get('use_text_encoder_2', True)
self.experimental_xl: bool = kwargs.get('experimental_xl', False)
if self.name_or_path is None:
raise ValueError('name_or_path must be specified')

116
toolkit/keymaps/stable_diffusion_sd1.json
View File

@@ -58,7 +58,7 @@
"cond_stage_model.transformer.text_model.encoder.layers.11.mlp.fc1.weight": "te_text_model.encoder.layers.11.mlp.fc1.weight",
"cond_stage_model.transformer.text_model.encoder.layers.11.mlp.fc2.bias": "te_text_model.encoder.layers.11.mlp.fc2.bias",
"cond_stage_model.transformer.text_model.encoder.layers.11.mlp.fc2.weight": "te_text_model.encoder.layers.11.mlp.fc2.weight",
"cond_stage_model.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "te_text_model.encoder.layers.2.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "te_text_model.encoder.layers.11.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.11.self_attn.k_proj.weight": "te_text_model.encoder.layers.11.self_attn.k_proj.weight",
"cond_stage_model.transformer.text_model.encoder.layers.11.self_attn.out_proj.bias": "te_text_model.encoder.layers.11.self_attn.out_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.11.self_attn.out_proj.weight": "te_text_model.encoder.layers.11.self_attn.out_proj.weight",
@@ -74,7 +74,7 @@
"cond_stage_model.transformer.text_model.encoder.layers.2.mlp.fc1.weight": "te_text_model.encoder.layers.2.mlp.fc1.weight",
"cond_stage_model.transformer.text_model.encoder.layers.2.mlp.fc2.bias": "te_text_model.encoder.layers.2.mlp.fc2.bias",
"cond_stage_model.transformer.text_model.encoder.layers.2.mlp.fc2.weight": "te_text_model.encoder.layers.2.mlp.fc2.weight",
"cond_stage_model.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "te_text_model.encoder.layers.3.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "te_text_model.encoder.layers.2.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.2.self_attn.k_proj.weight": "te_text_model.encoder.layers.2.self_attn.k_proj.weight",
"cond_stage_model.transformer.text_model.encoder.layers.2.self_attn.out_proj.bias": "te_text_model.encoder.layers.2.self_attn.out_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.2.self_attn.out_proj.weight": "te_text_model.encoder.layers.2.self_attn.out_proj.weight",
@@ -90,7 +90,7 @@
"cond_stage_model.transformer.text_model.encoder.layers.3.mlp.fc1.weight": "te_text_model.encoder.layers.3.mlp.fc1.weight",
"cond_stage_model.transformer.text_model.encoder.layers.3.mlp.fc2.bias": "te_text_model.encoder.layers.3.mlp.fc2.bias",
"cond_stage_model.transformer.text_model.encoder.layers.3.mlp.fc2.weight": "te_text_model.encoder.layers.3.mlp.fc2.weight",
"cond_stage_model.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "te_text_model.encoder.layers.11.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "te_text_model.encoder.layers.3.self_attn.k_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.3.self_attn.k_proj.weight": "te_text_model.encoder.layers.3.self_attn.k_proj.weight",
"cond_stage_model.transformer.text_model.encoder.layers.3.self_attn.out_proj.bias": "te_text_model.encoder.layers.3.self_attn.out_proj.bias",
"cond_stage_model.transformer.text_model.encoder.layers.3.self_attn.out_proj.weight": "te_text_model.encoder.layers.3.self_attn.out_proj.weight",
@@ -446,11 +446,11 @@
"first_stage_model.quant_conv.weight": "vae_quant_conv.weight",
"model.diffusion_model.input_blocks.0.0.bias": "unet_conv_in.bias",
"model.diffusion_model.input_blocks.0.0.weight": "unet_conv_in.weight",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.weight": "unet_down_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.1.0.in_layers.0.bias": "unet_down_blocks.0.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.0.weight": "unet_down_blocks.0.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "unet_down_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "unet_down_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.2.weight": "unet_down_blocks.0.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.1.0.out_layers.0.bias": "unet_down_blocks.0.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.1.0.out_layers.0.weight": "unet_down_blocks.0.resnets.0.norm2.weight",
@@ -482,31 +482,31 @@
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "unet_time_embedding.linear_2.weight",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.10.0.in_layers.0.bias": "unet_down_blocks.3.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.0.weight": "unet_down_blocks.3.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "unet_down_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "unet_down_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.2.weight": "unet_down_blocks.3.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.10.0.out_layers.0.bias": "unet_down_blocks.3.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.10.0.out_layers.0.weight": "unet_down_blocks.3.resnets.0.norm2.weight",
"model.diffusion_model.input_blocks.10.0.out_layers.3.bias": "unet_down_blocks.3.resnets.0.conv2.bias",
"model.diffusion_model.input_blocks.10.0.out_layers.3.weight": "unet_down_blocks.3.resnets.0.conv2.weight",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.11.0.in_layers.0.bias": "unet_down_blocks.3.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.0.weight": "unet_down_blocks.3.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "unet_down_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "unet_down_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.2.weight": "unet_down_blocks.3.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.11.0.out_layers.0.bias": "unet_down_blocks.3.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.11.0.out_layers.0.weight": "unet_down_blocks.3.resnets.1.norm2.weight",
"model.diffusion_model.input_blocks.11.0.out_layers.3.bias": "unet_down_blocks.3.resnets.1.conv2.bias",
"model.diffusion_model.input_blocks.11.0.out_layers.3.weight": "unet_down_blocks.3.resnets.1.conv2.weight",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.weight": "unet_down_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.2.0.in_layers.0.bias": "unet_down_blocks.0.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.0.weight": "unet_down_blocks.0.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "unet_down_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "unet_down_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.2.weight": "unet_down_blocks.0.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.2.0.out_layers.0.bias": "unet_down_blocks.0.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.2.0.out_layers.0.weight": "unet_down_blocks.0.resnets.1.norm2.weight",
@@ -540,11 +540,11 @@
"model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.0.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.3.0.op.bias": "unet_down_blocks.0.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.3.0.op.weight": "unet_down_blocks.0.downsamplers.0.conv.weight",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.weight": "unet_down_blocks.1.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.4.0.in_layers.0.bias": "unet_down_blocks.1.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.0.weight": "unet_down_blocks.1.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "unet_down_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "unet_down_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.2.weight": "unet_down_blocks.1.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.4.0.out_layers.0.bias": "unet_down_blocks.1.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.4.0.out_layers.0.weight": "unet_down_blocks.1.resnets.0.norm2.weight",
@@ -578,11 +578,11 @@
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.weight": "unet_down_blocks.1.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.5.0.in_layers.0.bias": "unet_down_blocks.1.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.0.weight": "unet_down_blocks.1.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "unet_down_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "unet_down_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.2.weight": "unet_down_blocks.1.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.5.0.out_layers.0.bias": "unet_down_blocks.1.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.5.0.out_layers.0.weight": "unet_down_blocks.1.resnets.1.norm2.weight",
@@ -616,11 +616,11 @@
"model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.1.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.6.0.op.bias": "unet_down_blocks.1.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.6.0.op.weight": "unet_down_blocks.1.downsamplers.0.conv.weight",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.weight": "unet_down_blocks.2.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.7.0.in_layers.0.bias": "unet_down_blocks.2.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.0.weight": "unet_down_blocks.2.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "unet_down_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "unet_down_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.2.weight": "unet_down_blocks.2.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.7.0.out_layers.0.bias": "unet_down_blocks.2.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.7.0.out_layers.0.weight": "unet_down_blocks.2.resnets.0.norm2.weight",
@@ -654,11 +654,11 @@
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.weight": "unet_down_blocks.2.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.8.0.in_layers.0.bias": "unet_down_blocks.2.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.0.weight": "unet_down_blocks.2.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "unet_down_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "unet_down_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.2.weight": "unet_down_blocks.2.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.8.0.out_layers.0.bias": "unet_down_blocks.2.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.8.0.out_layers.0.weight": "unet_down_blocks.2.resnets.1.norm2.weight",
@@ -692,11 +692,11 @@
"model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.2.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.9.0.op.bias": "unet_down_blocks.2.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.9.0.op.weight": "unet_down_blocks.2.downsamplers.0.conv.weight",
"model.diffusion_model.middle_block.0.emb_layers.1.bias": "unet_mid_block.resnets.0.conv1.bias",
"model.diffusion_model.middle_block.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.middle_block.0.emb_layers.1.bias": "unet_mid_block.resnets.0.time_emb_proj.bias",
"model.diffusion_model.middle_block.0.emb_layers.1.weight": "unet_mid_block.resnets.0.time_emb_proj.weight",
"model.diffusion_model.middle_block.0.in_layers.0.bias": "unet_mid_block.resnets.0.norm1.bias",
"model.diffusion_model.middle_block.0.in_layers.0.weight": "unet_mid_block.resnets.0.norm1.weight",
"model.diffusion_model.middle_block.0.in_layers.2.bias": "unet_mid_block.resnets.0.time_emb_proj.bias",
"model.diffusion_model.middle_block.0.in_layers.2.bias": "unet_mid_block.resnets.0.conv1.bias",
"model.diffusion_model.middle_block.0.in_layers.2.weight": "unet_mid_block.resnets.0.conv1.weight",
"model.diffusion_model.middle_block.0.out_layers.0.bias": "unet_mid_block.resnets.0.norm2.bias",
"model.diffusion_model.middle_block.0.out_layers.0.weight": "unet_mid_block.resnets.0.norm2.weight",
@@ -728,11 +728,11 @@
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.weight": "unet_mid_block.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.bias": "unet_mid_block.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.weight": "unet_mid_block.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.middle_block.2.emb_layers.1.bias": "unet_mid_block.resnets.1.conv1.bias",
"model.diffusion_model.middle_block.2.emb_layers.1.weight": "unet_up_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.middle_block.2.emb_layers.1.bias": "unet_mid_block.resnets.1.time_emb_proj.bias",
"model.diffusion_model.middle_block.2.emb_layers.1.weight": "unet_mid_block.resnets.1.time_emb_proj.weight",
"model.diffusion_model.middle_block.2.in_layers.0.bias": "unet_mid_block.resnets.1.norm1.bias",
"model.diffusion_model.middle_block.2.in_layers.0.weight": "unet_mid_block.resnets.1.norm1.weight",
"model.diffusion_model.middle_block.2.in_layers.2.bias": "unet_mid_block.resnets.1.time_emb_proj.bias",
"model.diffusion_model.middle_block.2.in_layers.2.bias": "unet_mid_block.resnets.1.conv1.bias",
"model.diffusion_model.middle_block.2.in_layers.2.weight": "unet_mid_block.resnets.1.conv1.weight",
"model.diffusion_model.middle_block.2.out_layers.0.bias": "unet_mid_block.resnets.1.norm2.bias",
"model.diffusion_model.middle_block.2.out_layers.0.weight": "unet_mid_block.resnets.1.norm2.weight",
@@ -742,11 +742,11 @@
"model.diffusion_model.out.0.weight": "unet_conv_norm_out.weight",
"model.diffusion_model.out.2.bias": "unet_conv_out.bias",
"model.diffusion_model.out.2.weight": "unet_conv_out.weight",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.0.0.in_layers.0.bias": "unet_up_blocks.0.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.0.weight": "unet_up_blocks.0.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "unet_up_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "unet_up_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.2.weight": "unet_up_blocks.0.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.0.0.out_layers.0.bias": "unet_up_blocks.0.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.0.0.out_layers.0.weight": "unet_up_blocks.0.resnets.0.norm2.weight",
@@ -754,11 +754,11 @@
"model.diffusion_model.output_blocks.0.0.out_layers.3.weight": "unet_up_blocks.0.resnets.0.conv2.weight",
"model.diffusion_model.output_blocks.0.0.skip_connection.bias": "unet_up_blocks.0.resnets.0.conv_shortcut.bias",
"model.diffusion_model.output_blocks.0.0.skip_connection.weight": "unet_up_blocks.0.resnets.0.conv_shortcut.weight",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.1.0.in_layers.0.bias": "unet_up_blocks.0.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.0.weight": "unet_up_blocks.0.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.1.0.in_layers.2.bias": "unet_up_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.2.bias": "unet_up_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.2.weight": "unet_up_blocks.0.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.1.0.out_layers.0.bias": "unet_up_blocks.0.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.1.0.out_layers.0.weight": "unet_up_blocks.0.resnets.1.norm2.weight",
@@ -766,11 +766,11 @@
"model.diffusion_model.output_blocks.1.0.out_layers.3.weight": "unet_up_blocks.0.resnets.1.conv2.weight",
"model.diffusion_model.output_blocks.1.0.skip_connection.bias": "unet_up_blocks.0.resnets.1.conv_shortcut.bias",
"model.diffusion_model.output_blocks.1.0.skip_connection.weight": "unet_up_blocks.0.resnets.1.conv_shortcut.weight",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.10.0.in_layers.0.bias": "unet_up_blocks.3.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.0.weight": "unet_up_blocks.3.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.10.0.in_layers.2.bias": "unet_up_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.2.bias": "unet_up_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.2.weight": "unet_up_blocks.3.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.10.0.out_layers.0.bias": "unet_up_blocks.3.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.10.0.out_layers.0.weight": "unet_up_blocks.3.resnets.1.norm2.weight",
@@ -804,11 +804,11 @@
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.11.0.in_layers.0.bias": "unet_up_blocks.3.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.0.weight": "unet_up_blocks.3.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.11.0.in_layers.2.bias": "unet_up_blocks.3.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.2.bias": "unet_up_blocks.3.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.2.weight": "unet_up_blocks.3.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.11.0.out_layers.0.bias": "unet_up_blocks.3.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.11.0.out_layers.0.weight": "unet_up_blocks.3.resnets.2.norm2.weight",
@@ -842,11 +842,11 @@
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.weight": "unet_mid_block.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.2.0.in_layers.0.bias": "unet_up_blocks.0.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.0.weight": "unet_up_blocks.0.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.2.0.in_layers.2.bias": "unet_up_blocks.0.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.2.bias": "unet_up_blocks.0.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.2.weight": "unet_up_blocks.0.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.2.0.out_layers.0.bias": "unet_up_blocks.0.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.2.0.out_layers.0.weight": "unet_up_blocks.0.resnets.2.norm2.weight",
@@ -856,11 +856,11 @@
"model.diffusion_model.output_blocks.2.0.skip_connection.weight": "unet_up_blocks.0.resnets.2.conv_shortcut.weight",
"model.diffusion_model.output_blocks.2.1.conv.bias": "unet_up_blocks.0.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.2.1.conv.weight": "unet_up_blocks.0.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.3.0.in_layers.0.bias": "unet_up_blocks.1.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.0.weight": "unet_up_blocks.1.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.3.0.in_layers.2.bias": "unet_up_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.2.bias": "unet_up_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.2.weight": "unet_up_blocks.1.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.3.0.out_layers.0.bias": "unet_up_blocks.1.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.3.0.out_layers.0.weight": "unet_up_blocks.1.resnets.0.norm2.weight",
@@ -894,11 +894,11 @@
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.4.0.in_layers.0.bias": "unet_up_blocks.1.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.0.weight": "unet_up_blocks.1.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.4.0.in_layers.2.bias": "unet_up_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.2.bias": "unet_up_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.2.weight": "unet_up_blocks.1.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.4.0.out_layers.0.bias": "unet_up_blocks.1.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.4.0.out_layers.0.weight": "unet_up_blocks.1.resnets.1.norm2.weight",
@@ -932,11 +932,11 @@
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.5.0.in_layers.0.bias": "unet_up_blocks.1.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.0.weight": "unet_up_blocks.1.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.5.0.in_layers.2.bias": "unet_up_blocks.1.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.2.bias": "unet_up_blocks.1.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.2.weight": "unet_up_blocks.1.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.5.0.out_layers.0.bias": "unet_up_blocks.1.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.5.0.out_layers.0.weight": "unet_up_blocks.1.resnets.2.norm2.weight",
@@ -972,11 +972,11 @@
"model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.5.2.conv.bias": "unet_up_blocks.1.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.5.2.conv.weight": "unet_up_blocks.1.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.6.0.in_layers.0.bias": "unet_up_blocks.2.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.0.weight": "unet_up_blocks.2.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.6.0.in_layers.2.bias": "unet_up_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.2.bias": "unet_up_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.2.weight": "unet_up_blocks.2.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.6.0.out_layers.0.bias": "unet_up_blocks.2.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.6.0.out_layers.0.weight": "unet_up_blocks.2.resnets.0.norm2.weight",
@@ -1010,11 +1010,11 @@
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.7.0.in_layers.0.bias": "unet_up_blocks.2.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.0.weight": "unet_up_blocks.2.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.7.0.in_layers.2.bias": "unet_up_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.2.bias": "unet_up_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.2.weight": "unet_up_blocks.2.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.7.0.out_layers.0.bias": "unet_up_blocks.2.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.7.0.out_layers.0.weight": "unet_up_blocks.2.resnets.1.norm2.weight",
@@ -1048,11 +1048,11 @@
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.8.0.in_layers.0.bias": "unet_up_blocks.2.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.0.weight": "unet_up_blocks.2.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.8.0.in_layers.2.bias": "unet_up_blocks.2.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.2.bias": "unet_up_blocks.2.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.2.weight": "unet_up_blocks.2.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.8.0.out_layers.0.bias": "unet_up_blocks.2.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.8.0.out_layers.0.weight": "unet_up_blocks.2.resnets.2.norm2.weight",
@@ -1088,11 +1088,11 @@
"model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.8.2.conv.bias": "unet_up_blocks.2.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.8.2.conv.weight": "unet_up_blocks.2.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.9.0.in_layers.0.bias": "unet_up_blocks.3.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.0.weight": "unet_up_blocks.3.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.9.0.in_layers.2.bias": "unet_up_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.2.bias": "unet_up_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.2.weight": "unet_up_blocks.3.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.9.0.out_layers.0.bias": "unet_up_blocks.3.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.9.0.out_layers.0.weight": "unet_up_blocks.3.resnets.0.norm2.weight",
@@ -1129,7 +1129,7 @@
"model.diffusion_model.time_embed.0.bias": "unet_time_embedding.linear_1.bias",
"model.diffusion_model.time_embed.0.weight": "unet_time_embedding.linear_1.weight",
"model.diffusion_model.time_embed.2.bias": "unet_time_embedding.linear_2.bias",
"model.diffusion_model.time_embed.2.weight": "unet_mid_block.resnets.1.time_emb_proj.weight"
"model.diffusion_model.time_embed.2.weight": "unet_time_embedding.linear_2.weight"
},
"ldm_diffusers_shape_map": {
"first_stage_model.decoder.mid.attn_1.k.weight": [

104
toolkit/keymaps/stable_diffusion_sd2.json
View File

@@ -484,11 +484,11 @@
"first_stage_model.quant_conv.weight": "vae_quant_conv.weight",
"model.diffusion_model.input_blocks.0.0.bias": "unet_conv_in.bias",
"model.diffusion_model.input_blocks.0.0.weight": "unet_conv_in.weight",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.1.0.emb_layers.1.weight": "unet_down_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.1.0.in_layers.0.bias": "unet_down_blocks.0.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.0.weight": "unet_down_blocks.0.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "unet_down_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "unet_down_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.1.0.in_layers.2.weight": "unet_down_blocks.0.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.1.0.out_layers.0.bias": "unet_down_blocks.0.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.1.0.out_layers.0.weight": "unet_down_blocks.0.resnets.0.norm2.weight",
@@ -520,31 +520,31 @@
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.0.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "unet_time_embedding.linear_2.weight",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.10.0.in_layers.0.bias": "unet_down_blocks.3.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.0.weight": "unet_down_blocks.3.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "unet_down_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "unet_down_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.10.0.in_layers.2.weight": "unet_down_blocks.3.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.10.0.out_layers.0.bias": "unet_down_blocks.3.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.10.0.out_layers.0.weight": "unet_down_blocks.3.resnets.0.norm2.weight",
"model.diffusion_model.input_blocks.10.0.out_layers.3.bias": "unet_down_blocks.3.resnets.0.conv2.bias",
"model.diffusion_model.input_blocks.10.0.out_layers.3.weight": "unet_down_blocks.3.resnets.0.conv2.weight",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "unet_down_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.11.0.in_layers.0.bias": "unet_down_blocks.3.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.0.weight": "unet_down_blocks.3.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "unet_down_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "unet_down_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.11.0.in_layers.2.weight": "unet_down_blocks.3.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.11.0.out_layers.0.bias": "unet_down_blocks.3.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.11.0.out_layers.0.weight": "unet_down_blocks.3.resnets.1.norm2.weight",
"model.diffusion_model.input_blocks.11.0.out_layers.3.bias": "unet_down_blocks.3.resnets.1.conv2.bias",
"model.diffusion_model.input_blocks.11.0.out_layers.3.weight": "unet_down_blocks.3.resnets.1.conv2.weight",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "unet_down_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.2.0.emb_layers.1.weight": "unet_down_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.2.0.in_layers.0.bias": "unet_down_blocks.0.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.0.weight": "unet_down_blocks.0.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "unet_down_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "unet_down_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.2.0.in_layers.2.weight": "unet_down_blocks.0.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.2.0.out_layers.0.bias": "unet_down_blocks.0.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.2.0.out_layers.0.weight": "unet_down_blocks.0.resnets.1.norm2.weight",
@@ -578,11 +578,11 @@
"model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.0.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.3.0.op.bias": "unet_down_blocks.0.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.3.0.op.weight": "unet_down_blocks.0.downsamplers.0.conv.weight",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.4.0.emb_layers.1.weight": "unet_down_blocks.1.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.4.0.in_layers.0.bias": "unet_down_blocks.1.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.0.weight": "unet_down_blocks.1.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "unet_down_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "unet_down_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.4.0.in_layers.2.weight": "unet_down_blocks.1.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.4.0.out_layers.0.bias": "unet_down_blocks.1.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.4.0.out_layers.0.weight": "unet_down_blocks.1.resnets.0.norm2.weight",
@@ -616,11 +616,11 @@
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.1.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "unet_down_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.5.0.emb_layers.1.weight": "unet_down_blocks.1.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.5.0.in_layers.0.bias": "unet_down_blocks.1.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.0.weight": "unet_down_blocks.1.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "unet_down_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "unet_down_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.5.0.in_layers.2.weight": "unet_down_blocks.1.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.5.0.out_layers.0.bias": "unet_down_blocks.1.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.5.0.out_layers.0.weight": "unet_down_blocks.1.resnets.1.norm2.weight",
@@ -654,11 +654,11 @@
"model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.1.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.6.0.op.bias": "unet_down_blocks.1.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.6.0.op.weight": "unet_down_blocks.1.downsamplers.0.conv.weight",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.7.0.emb_layers.1.weight": "unet_down_blocks.2.resnets.0.time_emb_proj.weight",
"model.diffusion_model.input_blocks.7.0.in_layers.0.bias": "unet_down_blocks.2.resnets.0.norm1.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.0.weight": "unet_down_blocks.2.resnets.0.norm1.weight",
"model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "unet_down_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "unet_down_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.input_blocks.7.0.in_layers.2.weight": "unet_down_blocks.2.resnets.0.conv1.weight",
"model.diffusion_model.input_blocks.7.0.out_layers.0.bias": "unet_down_blocks.2.resnets.0.norm2.bias",
"model.diffusion_model.input_blocks.7.0.out_layers.0.weight": "unet_down_blocks.2.resnets.0.norm2.weight",
@@ -692,11 +692,11 @@
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "unet_down_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.8.0.emb_layers.1.weight": "unet_down_blocks.2.resnets.1.time_emb_proj.weight",
"model.diffusion_model.input_blocks.8.0.in_layers.0.bias": "unet_down_blocks.2.resnets.1.norm1.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.0.weight": "unet_down_blocks.2.resnets.1.norm1.weight",
"model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "unet_down_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "unet_down_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.input_blocks.8.0.in_layers.2.weight": "unet_down_blocks.2.resnets.1.conv1.weight",
"model.diffusion_model.input_blocks.8.0.out_layers.0.bias": "unet_down_blocks.2.resnets.1.norm2.bias",
"model.diffusion_model.input_blocks.8.0.out_layers.0.weight": "unet_down_blocks.2.resnets.1.norm2.weight",
@@ -730,11 +730,11 @@
"model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.weight": "unet_down_blocks.2.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.input_blocks.9.0.op.bias": "unet_down_blocks.2.downsamplers.0.conv.bias",
"model.diffusion_model.input_blocks.9.0.op.weight": "unet_down_blocks.2.downsamplers.0.conv.weight",
"model.diffusion_model.middle_block.0.emb_layers.1.bias": "unet_mid_block.resnets.0.conv1.bias",
"model.diffusion_model.middle_block.0.emb_layers.1.weight": "unet_down_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.middle_block.0.emb_layers.1.bias": "unet_mid_block.resnets.0.time_emb_proj.bias",
"model.diffusion_model.middle_block.0.emb_layers.1.weight": "unet_mid_block.resnets.0.time_emb_proj.weight",
"model.diffusion_model.middle_block.0.in_layers.0.bias": "unet_mid_block.resnets.0.norm1.bias",
"model.diffusion_model.middle_block.0.in_layers.0.weight": "unet_mid_block.resnets.0.norm1.weight",
"model.diffusion_model.middle_block.0.in_layers.2.bias": "unet_mid_block.resnets.0.time_emb_proj.bias",
"model.diffusion_model.middle_block.0.in_layers.2.bias": "unet_mid_block.resnets.0.conv1.bias",
"model.diffusion_model.middle_block.0.in_layers.2.weight": "unet_mid_block.resnets.0.conv1.weight",
"model.diffusion_model.middle_block.0.out_layers.0.bias": "unet_mid_block.resnets.0.norm2.bias",
"model.diffusion_model.middle_block.0.out_layers.0.weight": "unet_mid_block.resnets.0.norm2.weight",
@@ -766,11 +766,11 @@
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.weight": "unet_mid_block.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.bias": "unet_mid_block.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.weight": "unet_mid_block.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.middle_block.2.emb_layers.1.bias": "unet_mid_block.resnets.1.conv1.bias",
"model.diffusion_model.middle_block.2.emb_layers.1.weight": "unet_up_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.middle_block.2.emb_layers.1.bias": "unet_mid_block.resnets.1.time_emb_proj.bias",
"model.diffusion_model.middle_block.2.emb_layers.1.weight": "unet_mid_block.resnets.1.time_emb_proj.weight",
"model.diffusion_model.middle_block.2.in_layers.0.bias": "unet_mid_block.resnets.1.norm1.bias",
"model.diffusion_model.middle_block.2.in_layers.0.weight": "unet_mid_block.resnets.1.norm1.weight",
"model.diffusion_model.middle_block.2.in_layers.2.bias": "unet_mid_block.resnets.1.time_emb_proj.bias",
"model.diffusion_model.middle_block.2.in_layers.2.bias": "unet_mid_block.resnets.1.conv1.bias",
"model.diffusion_model.middle_block.2.in_layers.2.weight": "unet_mid_block.resnets.1.conv1.weight",
"model.diffusion_model.middle_block.2.out_layers.0.bias": "unet_mid_block.resnets.1.norm2.bias",
"model.diffusion_model.middle_block.2.out_layers.0.weight": "unet_mid_block.resnets.1.norm2.weight",
@@ -780,11 +780,11 @@
"model.diffusion_model.out.0.weight": "unet_conv_norm_out.weight",
"model.diffusion_model.out.2.bias": "unet_conv_out.bias",
"model.diffusion_model.out.2.weight": "unet_conv_out.weight",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.0.0.in_layers.0.bias": "unet_up_blocks.0.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.0.weight": "unet_up_blocks.0.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "unet_up_blocks.0.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "unet_up_blocks.0.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.0.0.in_layers.2.weight": "unet_up_blocks.0.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.0.0.out_layers.0.bias": "unet_up_blocks.0.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.0.0.out_layers.0.weight": "unet_up_blocks.0.resnets.0.norm2.weight",
@@ -792,11 +792,11 @@
"model.diffusion_model.output_blocks.0.0.out_layers.3.weight": "unet_up_blocks.0.resnets.0.conv2.weight",
"model.diffusion_model.output_blocks.0.0.skip_connection.bias": "unet_up_blocks.0.resnets.0.conv_shortcut.bias",
"model.diffusion_model.output_blocks.0.0.skip_connection.weight": "unet_up_blocks.0.resnets.0.conv_shortcut.weight",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.1.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.1.0.in_layers.0.bias": "unet_up_blocks.0.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.0.weight": "unet_up_blocks.0.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.1.0.in_layers.2.bias": "unet_up_blocks.0.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.2.bias": "unet_up_blocks.0.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.1.0.in_layers.2.weight": "unet_up_blocks.0.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.1.0.out_layers.0.bias": "unet_up_blocks.0.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.1.0.out_layers.0.weight": "unet_up_blocks.0.resnets.1.norm2.weight",
@@ -804,11 +804,11 @@
"model.diffusion_model.output_blocks.1.0.out_layers.3.weight": "unet_up_blocks.0.resnets.1.conv2.weight",
"model.diffusion_model.output_blocks.1.0.skip_connection.bias": "unet_up_blocks.0.resnets.1.conv_shortcut.bias",
"model.diffusion_model.output_blocks.1.0.skip_connection.weight": "unet_up_blocks.0.resnets.1.conv_shortcut.weight",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.10.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.10.0.in_layers.0.bias": "unet_up_blocks.3.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.0.weight": "unet_up_blocks.3.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.10.0.in_layers.2.bias": "unet_up_blocks.3.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.2.bias": "unet_up_blocks.3.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.10.0.in_layers.2.weight": "unet_up_blocks.3.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.10.0.out_layers.0.bias": "unet_up_blocks.3.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.10.0.out_layers.0.weight": "unet_up_blocks.3.resnets.1.norm2.weight",
@@ -842,11 +842,11 @@
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.3.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.11.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.11.0.in_layers.0.bias": "unet_up_blocks.3.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.0.weight": "unet_up_blocks.3.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.11.0.in_layers.2.bias": "unet_up_blocks.3.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.2.bias": "unet_up_blocks.3.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.11.0.in_layers.2.weight": "unet_up_blocks.3.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.11.0.out_layers.0.bias": "unet_up_blocks.3.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.11.0.out_layers.0.weight": "unet_up_blocks.3.resnets.2.norm2.weight",
@@ -880,11 +880,11 @@
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.3.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.weight": "unet_mid_block.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.bias": "unet_up_blocks.0.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.2.0.emb_layers.1.weight": "unet_up_blocks.0.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.2.0.in_layers.0.bias": "unet_up_blocks.0.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.0.weight": "unet_up_blocks.0.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.2.0.in_layers.2.bias": "unet_up_blocks.0.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.2.bias": "unet_up_blocks.0.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.2.0.in_layers.2.weight": "unet_up_blocks.0.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.2.0.out_layers.0.bias": "unet_up_blocks.0.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.2.0.out_layers.0.weight": "unet_up_blocks.0.resnets.2.norm2.weight",
@@ -894,11 +894,11 @@
"model.diffusion_model.output_blocks.2.0.skip_connection.weight": "unet_up_blocks.0.resnets.2.conv_shortcut.weight",
"model.diffusion_model.output_blocks.2.1.conv.bias": "unet_up_blocks.0.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.2.1.conv.weight": "unet_up_blocks.0.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.3.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.3.0.in_layers.0.bias": "unet_up_blocks.1.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.0.weight": "unet_up_blocks.1.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.3.0.in_layers.2.bias": "unet_up_blocks.1.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.2.bias": "unet_up_blocks.1.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.3.0.in_layers.2.weight": "unet_up_blocks.1.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.3.0.out_layers.0.bias": "unet_up_blocks.1.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.3.0.out_layers.0.weight": "unet_up_blocks.1.resnets.0.norm2.weight",
@@ -932,11 +932,11 @@
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.4.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.4.0.in_layers.0.bias": "unet_up_blocks.1.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.0.weight": "unet_up_blocks.1.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.4.0.in_layers.2.bias": "unet_up_blocks.1.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.2.bias": "unet_up_blocks.1.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.4.0.in_layers.2.weight": "unet_up_blocks.1.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.4.0.out_layers.0.bias": "unet_up_blocks.1.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.4.0.out_layers.0.weight": "unet_up_blocks.1.resnets.1.norm2.weight",
@@ -970,11 +970,11 @@
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.bias": "unet_up_blocks.1.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.5.0.emb_layers.1.weight": "unet_up_blocks.1.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.5.0.in_layers.0.bias": "unet_up_blocks.1.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.0.weight": "unet_up_blocks.1.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.5.0.in_layers.2.bias": "unet_up_blocks.1.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.2.bias": "unet_up_blocks.1.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.5.0.in_layers.2.weight": "unet_up_blocks.1.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.5.0.out_layers.0.bias": "unet_up_blocks.1.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.5.0.out_layers.0.weight": "unet_up_blocks.1.resnets.2.norm2.weight",
@@ -1010,11 +1010,11 @@
"model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.1.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.5.2.conv.bias": "unet_up_blocks.1.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.5.2.conv.weight": "unet_up_blocks.1.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.6.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.6.0.in_layers.0.bias": "unet_up_blocks.2.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.0.weight": "unet_up_blocks.2.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.6.0.in_layers.2.bias": "unet_up_blocks.2.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.2.bias": "unet_up_blocks.2.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.6.0.in_layers.2.weight": "unet_up_blocks.2.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.6.0.out_layers.0.bias": "unet_up_blocks.2.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.6.0.out_layers.0.weight": "unet_up_blocks.2.resnets.0.norm2.weight",
@@ -1048,11 +1048,11 @@
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.0.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.7.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.1.time_emb_proj.weight",
"model.diffusion_model.output_blocks.7.0.in_layers.0.bias": "unet_up_blocks.2.resnets.1.norm1.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.0.weight": "unet_up_blocks.2.resnets.1.norm1.weight",
"model.diffusion_model.output_blocks.7.0.in_layers.2.bias": "unet_up_blocks.2.resnets.1.time_emb_proj.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.2.bias": "unet_up_blocks.2.resnets.1.conv1.bias",
"model.diffusion_model.output_blocks.7.0.in_layers.2.weight": "unet_up_blocks.2.resnets.1.conv1.weight",
"model.diffusion_model.output_blocks.7.0.out_layers.0.bias": "unet_up_blocks.2.resnets.1.norm2.bias",
"model.diffusion_model.output_blocks.7.0.out_layers.0.weight": "unet_up_blocks.2.resnets.1.norm2.weight",
@@ -1086,11 +1086,11 @@
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm2.weight": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm2.weight",
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.bias": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm3.bias",
"model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.1.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.bias": "unet_up_blocks.2.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.8.0.emb_layers.1.weight": "unet_up_blocks.2.resnets.2.time_emb_proj.weight",
"model.diffusion_model.output_blocks.8.0.in_layers.0.bias": "unet_up_blocks.2.resnets.2.norm1.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.0.weight": "unet_up_blocks.2.resnets.2.norm1.weight",
"model.diffusion_model.output_blocks.8.0.in_layers.2.bias": "unet_up_blocks.2.resnets.2.time_emb_proj.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.2.bias": "unet_up_blocks.2.resnets.2.conv1.bias",
"model.diffusion_model.output_blocks.8.0.in_layers.2.weight": "unet_up_blocks.2.resnets.2.conv1.weight",
"model.diffusion_model.output_blocks.8.0.out_layers.0.bias": "unet_up_blocks.2.resnets.2.norm2.bias",
"model.diffusion_model.output_blocks.8.0.out_layers.0.weight": "unet_up_blocks.2.resnets.2.norm2.weight",
@@ -1126,11 +1126,11 @@
"model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm3.weight": "unet_up_blocks.2.attentions.2.transformer_blocks.0.norm3.weight",
"model.diffusion_model.output_blocks.8.2.conv.bias": "unet_up_blocks.2.upsamplers.0.conv.bias",
"model.diffusion_model.output_blocks.8.2.conv.weight": "unet_up_blocks.2.upsamplers.0.conv.weight",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.bias": "unet_up_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.9.0.emb_layers.1.weight": "unet_up_blocks.3.resnets.0.time_emb_proj.weight",
"model.diffusion_model.output_blocks.9.0.in_layers.0.bias": "unet_up_blocks.3.resnets.0.norm1.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.0.weight": "unet_up_blocks.3.resnets.0.norm1.weight",
"model.diffusion_model.output_blocks.9.0.in_layers.2.bias": "unet_up_blocks.3.resnets.0.time_emb_proj.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.2.bias": "unet_up_blocks.3.resnets.0.conv1.bias",
"model.diffusion_model.output_blocks.9.0.in_layers.2.weight": "unet_up_blocks.3.resnets.0.conv1.weight",
"model.diffusion_model.output_blocks.9.0.out_layers.0.bias": "unet_up_blocks.3.resnets.0.norm2.bias",
"model.diffusion_model.output_blocks.9.0.out_layers.0.weight": "unet_up_blocks.3.resnets.0.norm2.weight",
@@ -1167,7 +1167,7 @@
"model.diffusion_model.time_embed.0.bias": "unet_time_embedding.linear_1.bias",
"model.diffusion_model.time_embed.0.weight": "unet_time_embedding.linear_1.weight",
"model.diffusion_model.time_embed.2.bias": "unet_time_embedding.linear_2.bias",
"model.diffusion_model.time_embed.2.weight": "unet_mid_block.resnets.1.time_emb_proj.weight"
"model.diffusion_model.time_embed.2.weight": "unet_time_embedding.linear_2.weight"
},
"ldm_diffusers_shape_map": {
"first_stage_model.decoder.mid.attn_1.k.weight": [

70
toolkit/keymaps/stable_diffusion_sdxl.json
View File

@@ -58,7 +58,7 @@
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.mlp.fc1.weight": "te0_text_model.encoder.layers.11.mlp.fc1.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.mlp.fc2.bias": "te0_text_model.encoder.layers.11.mlp.fc2.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.mlp.fc2.weight": "te0_text_model.encoder.layers.11.mlp.fc2.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "te0_text_model.encoder.layers.2.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "te0_text_model.encoder.layers.11.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.self_attn.k_proj.weight": "te0_text_model.encoder.layers.11.self_attn.k_proj.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.self_attn.out_proj.bias": "te0_text_model.encoder.layers.11.self_attn.out_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.11.self_attn.out_proj.weight": "te0_text_model.encoder.layers.11.self_attn.out_proj.weight",
@@ -74,7 +74,7 @@
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.mlp.fc1.weight": "te0_text_model.encoder.layers.2.mlp.fc1.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.mlp.fc2.bias": "te0_text_model.encoder.layers.2.mlp.fc2.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.mlp.fc2.weight": "te0_text_model.encoder.layers.2.mlp.fc2.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "te0_text_model.encoder.layers.3.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "te0_text_model.encoder.layers.2.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.self_attn.k_proj.weight": "te0_text_model.encoder.layers.2.self_attn.k_proj.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.self_attn.out_proj.bias": "te0_text_model.encoder.layers.2.self_attn.out_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.2.self_attn.out_proj.weight": "te0_text_model.encoder.layers.2.self_attn.out_proj.weight",
@@ -90,7 +90,7 @@
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.mlp.fc1.weight": "te0_text_model.encoder.layers.3.mlp.fc1.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.mlp.fc2.bias": "te0_text_model.encoder.layers.3.mlp.fc2.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.mlp.fc2.weight": "te0_text_model.encoder.layers.3.mlp.fc2.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "te0_text_model.encoder.layers.11.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "te0_text_model.encoder.layers.3.self_attn.k_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.self_attn.k_proj.weight": "te0_text_model.encoder.layers.3.self_attn.k_proj.weight",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.self_attn.out_proj.bias": "te0_text_model.encoder.layers.3.self_attn.out_proj.bias",
"conditioner.embedders.0.transformer.text_model.encoder.layers.3.self_attn.out_proj.weight": "te0_text_model.encoder.layers.3.self_attn.out_proj.weight",
@@ -221,7 +221,7 @@
"conditioner.embedders.1.model.transformer.resblocks.1.mlp.c_proj.bias": "te1_text_model.encoder.layers.1.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.1.mlp.c_proj.weight": "te1_text_model.encoder.layers.1.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.10.attn.out_proj.bias": "te1_text_model.encoder.layers.10.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.10.attn.out_proj.weight": "te1_text_model.encoder.layers.2.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.10.attn.out_proj.weight": "te1_text_model.encoder.layers.10.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.10.ln_1.bias": "te1_text_model.encoder.layers.10.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.10.ln_1.weight": "te1_text_model.encoder.layers.10.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.10.ln_2.bias": "te1_text_model.encoder.layers.10.layer_norm2.bias",
@@ -231,7 +231,7 @@
"conditioner.embedders.1.model.transformer.resblocks.10.mlp.c_proj.bias": "te1_text_model.encoder.layers.10.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.10.mlp.c_proj.weight": "te1_text_model.encoder.layers.10.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.11.attn.out_proj.bias": "te1_text_model.encoder.layers.11.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.11.attn.out_proj.weight": "te1_text_model.encoder.layers.3.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.11.attn.out_proj.weight": "te1_text_model.encoder.layers.11.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.11.ln_1.bias": "te1_text_model.encoder.layers.11.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.11.ln_1.weight": "te1_text_model.encoder.layers.11.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.11.ln_2.bias": "te1_text_model.encoder.layers.11.layer_norm2.bias",
@@ -241,7 +241,7 @@
"conditioner.embedders.1.model.transformer.resblocks.11.mlp.c_proj.bias": "te1_text_model.encoder.layers.11.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.11.mlp.c_proj.weight": "te1_text_model.encoder.layers.11.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.12.attn.out_proj.bias": "te1_text_model.encoder.layers.12.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.12.attn.out_proj.weight": "te1_text_model.encoder.layers.4.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.12.attn.out_proj.weight": "te1_text_model.encoder.layers.12.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.12.ln_1.bias": "te1_text_model.encoder.layers.12.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.12.ln_1.weight": "te1_text_model.encoder.layers.12.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.12.ln_2.bias": "te1_text_model.encoder.layers.12.layer_norm2.bias",
@@ -251,7 +251,7 @@
"conditioner.embedders.1.model.transformer.resblocks.12.mlp.c_proj.bias": "te1_text_model.encoder.layers.12.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.12.mlp.c_proj.weight": "te1_text_model.encoder.layers.12.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.13.attn.out_proj.bias": "te1_text_model.encoder.layers.13.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.13.attn.out_proj.weight": "te1_text_model.encoder.layers.5.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.13.attn.out_proj.weight": "te1_text_model.encoder.layers.13.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.13.ln_1.bias": "te1_text_model.encoder.layers.13.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.13.ln_1.weight": "te1_text_model.encoder.layers.13.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.13.ln_2.bias": "te1_text_model.encoder.layers.13.layer_norm2.bias",
@@ -271,7 +271,7 @@
"conditioner.embedders.1.model.transformer.resblocks.14.mlp.c_proj.bias": "te1_text_model.encoder.layers.14.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.14.mlp.c_proj.weight": "te1_text_model.encoder.layers.14.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.15.attn.out_proj.bias": "te1_text_model.encoder.layers.15.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.15.attn.out_proj.weight": "te1_text_model.encoder.layers.6.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.15.attn.out_proj.weight": "te1_text_model.encoder.layers.15.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.15.ln_1.bias": "te1_text_model.encoder.layers.15.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.15.ln_1.weight": "te1_text_model.encoder.layers.15.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.15.ln_2.bias": "te1_text_model.encoder.layers.15.layer_norm2.bias",
@@ -321,7 +321,7 @@
"conditioner.embedders.1.model.transformer.resblocks.19.mlp.c_proj.bias": "te1_text_model.encoder.layers.19.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.19.mlp.c_proj.weight": "te1_text_model.encoder.layers.19.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.2.attn.out_proj.bias": "te1_text_model.encoder.layers.2.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.2.attn.out_proj.weight": "te1_text_model.encoder.layers.7.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.2.attn.out_proj.weight": "te1_text_model.encoder.layers.2.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.2.ln_1.bias": "te1_text_model.encoder.layers.2.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.2.ln_1.weight": "te1_text_model.encoder.layers.2.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.2.ln_2.bias": "te1_text_model.encoder.layers.2.layer_norm2.bias",
@@ -431,7 +431,7 @@
"conditioner.embedders.1.model.transformer.resblocks.29.mlp.c_proj.bias": "te1_text_model.encoder.layers.29.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.29.mlp.c_proj.weight": "te1_text_model.encoder.layers.29.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.3.attn.out_proj.bias": "te1_text_model.encoder.layers.3.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.3.attn.out_proj.weight": "te1_text_model.encoder.layers.8.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.3.attn.out_proj.weight": "te1_text_model.encoder.layers.3.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.3.ln_1.bias": "te1_text_model.encoder.layers.3.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.3.ln_1.weight": "te1_text_model.encoder.layers.3.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.3.ln_2.bias": "te1_text_model.encoder.layers.3.layer_norm2.bias",
@@ -461,7 +461,7 @@
"conditioner.embedders.1.model.transformer.resblocks.31.mlp.c_proj.bias": "te1_text_model.encoder.layers.31.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.31.mlp.c_proj.weight": "te1_text_model.encoder.layers.31.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.4.attn.out_proj.bias": "te1_text_model.encoder.layers.4.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.4.attn.out_proj.weight": "te1_text_model.encoder.layers.9.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.4.attn.out_proj.weight": "te1_text_model.encoder.layers.4.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.4.ln_1.bias": "te1_text_model.encoder.layers.4.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.4.ln_1.weight": "te1_text_model.encoder.layers.4.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.4.ln_2.bias": "te1_text_model.encoder.layers.4.layer_norm2.bias",
@@ -471,7 +471,7 @@
"conditioner.embedders.1.model.transformer.resblocks.4.mlp.c_proj.bias": "te1_text_model.encoder.layers.4.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.4.mlp.c_proj.weight": "te1_text_model.encoder.layers.4.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.5.attn.out_proj.bias": "te1_text_model.encoder.layers.5.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.5.attn.out_proj.weight": "te1_text_model.encoder.layers.10.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.5.attn.out_proj.weight": "te1_text_model.encoder.layers.5.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.5.ln_1.bias": "te1_text_model.encoder.layers.5.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.5.ln_1.weight": "te1_text_model.encoder.layers.5.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.5.ln_2.bias": "te1_text_model.encoder.layers.5.layer_norm2.bias",
@@ -481,7 +481,7 @@
"conditioner.embedders.1.model.transformer.resblocks.5.mlp.c_proj.bias": "te1_text_model.encoder.layers.5.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.5.mlp.c_proj.weight": "te1_text_model.encoder.layers.5.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.6.attn.out_proj.bias": "te1_text_model.encoder.layers.6.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.6.attn.out_proj.weight": "te1_text_model.encoder.layers.11.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.6.attn.out_proj.weight": "te1_text_model.encoder.layers.6.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.6.ln_1.bias": "te1_text_model.encoder.layers.6.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.6.ln_1.weight": "te1_text_model.encoder.layers.6.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.6.ln_2.bias": "te1_text_model.encoder.layers.6.layer_norm2.bias",
@@ -491,7 +491,7 @@
"conditioner.embedders.1.model.transformer.resblocks.6.mlp.c_proj.bias": "te1_text_model.encoder.layers.6.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.6.mlp.c_proj.weight": "te1_text_model.encoder.layers.6.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.7.attn.out_proj.bias": "te1_text_model.encoder.layers.7.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.7.attn.out_proj.weight": "te1_text_model.encoder.layers.12.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.7.attn.out_proj.weight": "te1_text_model.encoder.layers.7.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.7.ln_1.bias": "te1_text_model.encoder.layers.7.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.7.ln_1.weight": "te1_text_model.encoder.layers.7.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.7.ln_2.bias": "te1_text_model.encoder.layers.7.layer_norm2.bias",
@@ -501,7 +501,7 @@
"conditioner.embedders.1.model.transformer.resblocks.7.mlp.c_proj.bias": "te1_text_model.encoder.layers.7.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.7.mlp.c_proj.weight": "te1_text_model.encoder.layers.7.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.8.attn.out_proj.bias": "te1_text_model.encoder.layers.8.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.8.attn.out_proj.weight": "te1_text_model.encoder.layers.13.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.8.attn.out_proj.weight": "te1_text_model.encoder.layers.8.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.8.ln_1.bias": "te1_text_model.encoder.layers.8.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.8.ln_1.weight": "te1_text_model.encoder.layers.8.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.8.ln_2.bias": "te1_text_model.encoder.layers.8.layer_norm2.bias",
@@ -511,7 +511,7 @@
"conditioner.embedders.1.model.transformer.resblocks.8.mlp.c_proj.bias": "te1_text_model.encoder.layers.8.mlp.fc2.bias",
"conditioner.embedders.1.model.transformer.resblocks.8.mlp.c_proj.weight": "te1_text_model.encoder.layers.8.mlp.fc2.weight",
"conditioner.embedders.1.model.transformer.resblocks.9.attn.out_proj.bias": "te1_text_model.encoder.layers.9.self_attn.out_proj.bias",
"conditioner.embedders.1.model.transformer.resblocks.9.attn.out_proj.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_out.0.weight",
"conditioner.embedders.1.model.transformer.resblocks.9.attn.out_proj.weight": "te1_text_model.encoder.layers.9.self_attn.out_proj.weight",
"conditioner.embedders.1.model.transformer.resblocks.9.ln_1.bias": "te1_text_model.encoder.layers.9.layer_norm1.bias",
"conditioner.embedders.1.model.transformer.resblocks.9.ln_1.weight": "te1_text_model.encoder.layers.9.layer_norm1.weight",
"conditioner.embedders.1.model.transformer.resblocks.9.ln_2.bias": "te1_text_model.encoder.layers.9.layer_norm2.bias",
@@ -1013,7 +1013,7 @@
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn1.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn1.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_k.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_k.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_out.0.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_out.0.bias",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_out.0.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_q.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_q.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.attn2.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.4.attn2.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.4.ff.net.0.proj.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.4.ff.net.0.proj.bias",
@@ -1033,7 +1033,7 @@
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn1.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn1.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_k.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_k.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_out.0.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_out.0.bias",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_q.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_q.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.attn2.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.5.attn2.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.5.ff.net.0.proj.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.5.ff.net.0.proj.bias",
@@ -1053,7 +1053,7 @@
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn1.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn1.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_k.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_k.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_out.0.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_out.0.bias",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_out.0.weight": "te1_text_model.encoder.layers.15.self_attn.out_proj.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_out.0.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_q.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_q.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.attn2.to_v.weight": "unet_down_blocks.2.attentions.0.transformer_blocks.6.attn2.to_v.weight",
"model.diffusion_model.input_blocks.7.1.transformer_blocks.6.ff.net.0.proj.bias": "unet_down_blocks.2.attentions.0.transformer_blocks.6.ff.net.0.proj.bias",
@@ -1369,7 +1369,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.0.ff.net.0.proj.bias",
@@ -1409,7 +1409,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.2.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.2.ff.net.0.proj.bias",
@@ -1429,7 +1429,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.3.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.3.ff.net.0.proj.bias",
@@ -1449,7 +1449,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.4.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.4.ff.net.0.proj.bias",
@@ -1469,7 +1469,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.5.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.5.ff.net.0.proj.bias",
@@ -1489,7 +1489,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.6.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.6.ff.net.0.proj.bias",
@@ -1509,7 +1509,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.7.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.7.ff.net.0.proj.bias",
@@ -1529,7 +1529,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.8.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.8.ff.net.0.proj.bias",
@@ -1549,7 +1549,7 @@
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn1.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn1.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_k.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_k.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_out.0.bias": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_out.0.bias",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.2.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_out.0.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_q.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_q.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.attn2.to_v.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_v.weight",
"model.diffusion_model.middle_block.1.transformer_blocks.9.ff.net.0.proj.bias": "unet_mid_block.attentions.0.transformer_blocks.9.ff.net.0.proj.bias",
@@ -2037,7 +2037,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.3.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.0.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.0.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.0.ff.net.0.proj.bias",
@@ -2117,7 +2117,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.4.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.4.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.4.ff.net.0.proj.bias",
@@ -2137,7 +2137,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.4.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.5.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.5.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.5.ff.net.0.proj.bias",
@@ -2157,7 +2157,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.6.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.6.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.6.ff.net.0.proj.bias",
@@ -2177,7 +2177,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.7.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.7.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.7.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.7.ff.net.0.proj.bias",
@@ -2197,7 +2197,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.8.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.8.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.8.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.8.ff.net.0.proj.bias",
@@ -2217,7 +2217,7 @@
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn1.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn1.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_k.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_k.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_out.0.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_out.0.bias",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_out.0.weight": "unet_mid_block.attentions.0.transformer_blocks.9.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_out.0.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_out.0.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_q.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_q.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.attn2.to_v.weight": "unet_up_blocks.0.attentions.2.transformer_blocks.9.attn2.to_v.weight",
"model.diffusion_model.output_blocks.2.1.transformer_blocks.9.ff.net.0.proj.bias": "unet_up_blocks.0.attentions.2.transformer_blocks.9.ff.net.0.proj.bias",

76
toolkit/stable_diffusion_model.py
View File

@@ -1,13 +1,14 @@
import gc
import json
import typing
from typing import Union, List, Tuple
from typing import Union, List, Tuple, Iterator
import sys
import os
from collections import OrderedDict
from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl import rescale_noise_cfg
from safetensors.torch import save_file
from safetensors.torch import save_file, load_file
from torch.nn import Parameter
from tqdm import tqdm
from torchvision.transforms import Resize
@@ -31,12 +32,12 @@ import diffusers
# tell it to shut up
diffusers.logging.set_verbosity(diffusers.logging.ERROR)
VAE_PREFIX_UNET = "vae"
SD_PREFIX_VAE = "vae"
SD_PREFIX_UNET = "unet"
SD_PREFIX_TEXT_ENCODER = "te"
SD_PREFIX_TEXT_ENCODER1 = "te1"
SD_PREFIX_TEXT_ENCODER2 = "te2"
SD_PREFIX_TEXT_ENCODER1 = "te0"
SD_PREFIX_TEXT_ENCODER2 = "te1"
# prefixed diffusers keys
DO_NOT_TRAIN_WEIGHTS = [
@@ -184,6 +185,21 @@ class StableDiffusion:
text_encoder.requires_grad_(False)
text_encoder.eval()
text_encoder = text_encoders
if self.model_config.experimental_xl:
print("Experimental XL mode enabled")
print("Loading and injecting alt weights")
# load the mismatched weight and force it in
raw_state_dict = load_file(model_path)
replacement_weight = raw_state_dict['conditioner.embedders.1.model.text_projection'].clone()
del raw_state_dict
# get state dict for for 2nd text encoder
te1_state_dict = text_encoders[1].state_dict()
# replace weight with mismatched weight
te1_state_dict['text_projection.weight'] = replacement_weight.to(self.device_torch, dtype=dtype)
flush()
print("Injecting alt weights")
else:
if self.custom_pipeline is not None:
pipln = self.custom_pipeline
@@ -707,7 +723,7 @@ class StableDiffusion:
state_dict = OrderedDict()
if vae:
for k, v in self.vae.state_dict().items():
new_key = k if k.startswith(f"{VAE_PREFIX_UNET}") else f"{VAE_PREFIX_UNET}_{k}"
new_key = k if k.startswith(f"{SD_PREFIX_VAE}") else f"{SD_PREFIX_VAE}_{k}"
state_dict[new_key] = v
if text_encoder:
if isinstance(self.text_encoder, list):
@@ -726,6 +742,35 @@ class StableDiffusion:
state_dict[new_key] = v
return state_dict
def named_parameters(self, vae=True, text_encoder=True, unet=True, state_dict_keys=False) -> OrderedDict[str, Parameter]:
named_params: OrderedDict[str, Parameter] = OrderedDict()
if vae:
for name, param in self.vae.named_parameters(recurse=True, prefix=f"{SD_PREFIX_VAE}"):
named_params[name] = param
if text_encoder:
if isinstance(self.text_encoder, list):
for i, encoder in enumerate(self.text_encoder):
for name, param in encoder.named_parameters(recurse=True, prefix=f"{SD_PREFIX_TEXT_ENCODER}{i}"):
named_params[name] = param
else:
for name, param in self.text_encoder.named_parameters(recurse=True, prefix=f"{SD_PREFIX_TEXT_ENCODER}"):
named_params[name] = param
if unet:
for name, param in self.unet.named_parameters(recurse=True, prefix=f"{SD_PREFIX_UNET}"):
named_params[name] = param
# convert to state dict keys, jsut replace . with _ on keys
if state_dict_keys:
new_named_params = OrderedDict()
for k, v in named_params.items():
# replace only the first . with an _
new_key = k.replace('.', '_', 1)
new_named_params[new_key] = v
named_params = new_named_params
return named_params
def save(self, output_file: str, meta: OrderedDict, save_dtype=get_torch_dtype('fp16'), logit_scale=None):
version_string = '1'
if self.is_v2:
@@ -764,24 +809,31 @@ class StableDiffusion:
trainable_parameters = []
# we use state dict to find params
if unet:
state_dict = self.state_dict(vae=False, unet=unet, text_encoder=False)
named_params = self.named_parameters(vae=False, unet=unet, text_encoder=False, state_dict_keys=True)
unet_lr = unet_lr if unet_lr is not None else default_lr
params = []
for key, diffusers_key in ldm_diffusers_keymap.items():
if diffusers_key in state_dict and diffusers_key not in DO_NOT_TRAIN_WEIGHTS:
params.append(state_dict[diffusers_key])
if diffusers_key in named_params and diffusers_key not in DO_NOT_TRAIN_WEIGHTS:
if named_params[diffusers_key].requires_grad:
params.append(named_params[diffusers_key])
param_data = {"params": params, "lr": unet_lr}
trainable_parameters.append(param_data)
print(f"Found {len(params)} trainable parameter in unet")
if text_encoder:
state_dict = self.state_dict(vae=False, unet=unet, text_encoder=text_encoder)
named_params = self.named_parameters(vae=False, unet=False, text_encoder=text_encoder, state_dict_keys=True)
text_encoder_lr = text_encoder_lr if text_encoder_lr is not None else default_lr
params = []
for key, diffusers_key in ldm_diffusers_keymap.items():
if diffusers_key in state_dict and diffusers_key not in DO_NOT_TRAIN_WEIGHTS:
params.append(state_dict[diffusers_key])
if diffusers_key in named_params and diffusers_key not in DO_NOT_TRAIN_WEIGHTS:
if named_params[diffusers_key].requires_grad:
params.append(named_params[diffusers_key])
param_data = {"params": params, "lr": text_encoder_lr}
trainable_parameters.append(param_data)
print(f"Found {len(params)} trainable parameter in text encoder")
return trainable_parameters