2023-01-03 06:53:32 +00:00
|
|
|
import torch
|
2023-02-17 20:31:38 +00:00
|
|
|
import contextlib
|
2023-03-31 21:19:58 +00:00
|
|
|
import copy
|
2023-01-03 06:53:32 +00:00
|
|
|
|
|
|
|
import sd1_clip
|
|
|
|
import sd2_clip
|
2023-02-08 08:17:54 +00:00
|
|
|
import model_management
|
2023-03-07 16:00:35 +00:00
|
|
|
from .ldm.util import instantiate_from_config
|
|
|
|
from .ldm.models.autoencoder import AutoencoderKL
|
2023-03-13 18:49:18 +00:00
|
|
|
import yaml
|
2023-02-16 15:38:08 +00:00
|
|
|
from .cldm import cldm
|
2023-02-25 05:55:42 +00:00
|
|
|
from .t2i_adapter import adapter
|
2023-02-16 15:38:08 +00:00
|
|
|
|
|
|
|
from . import utils
|
2023-04-02 03:19:15 +00:00
|
|
|
from . import clip_vision
|
2023-02-03 07:06:34 +00:00
|
|
|
|
2023-02-26 22:04:35 +00:00
|
|
|
def load_model_weights(model, sd, verbose=False, load_state_dict_to=[]):
|
2023-01-03 06:53:32 +00:00
|
|
|
m, u = model.load_state_dict(sd, strict=False)
|
|
|
|
|
|
|
|
k = list(sd.keys())
|
|
|
|
for x in k:
|
|
|
|
# print(x)
|
|
|
|
if x.startswith("cond_stage_model.transformer.") and not x.startswith("cond_stage_model.transformer.text_model."):
|
|
|
|
y = x.replace("cond_stage_model.transformer.", "cond_stage_model.transformer.text_model.")
|
|
|
|
sd[y] = sd.pop(x)
|
|
|
|
|
2023-01-28 07:14:22 +00:00
|
|
|
if 'cond_stage_model.transformer.text_model.embeddings.position_ids' in sd:
|
|
|
|
ids = sd['cond_stage_model.transformer.text_model.embeddings.position_ids']
|
|
|
|
if ids.dtype == torch.float32:
|
|
|
|
sd['cond_stage_model.transformer.text_model.embeddings.position_ids'] = ids.round()
|
2023-01-28 05:19:33 +00:00
|
|
|
|
2023-02-05 19:36:28 +00:00
|
|
|
keys_to_replace = {
|
|
|
|
"cond_stage_model.model.positional_embedding": "cond_stage_model.transformer.text_model.embeddings.position_embedding.weight",
|
|
|
|
"cond_stage_model.model.token_embedding.weight": "cond_stage_model.transformer.text_model.embeddings.token_embedding.weight",
|
|
|
|
"cond_stage_model.model.ln_final.weight": "cond_stage_model.transformer.text_model.final_layer_norm.weight",
|
|
|
|
"cond_stage_model.model.ln_final.bias": "cond_stage_model.transformer.text_model.final_layer_norm.bias",
|
|
|
|
}
|
|
|
|
|
|
|
|
for x in keys_to_replace:
|
|
|
|
if x in sd:
|
|
|
|
sd[keys_to_replace[x]] = sd.pop(x)
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
sd = utils.transformers_convert(sd, "cond_stage_model.model", "cond_stage_model.transformer.text_model", 24)
|
2023-02-05 19:36:28 +00:00
|
|
|
|
2023-01-03 06:53:32 +00:00
|
|
|
for x in load_state_dict_to:
|
|
|
|
x.load_state_dict(sd, strict=False)
|
|
|
|
|
|
|
|
if len(m) > 0 and verbose:
|
|
|
|
print("missing keys:")
|
|
|
|
print(m)
|
|
|
|
if len(u) > 0 and verbose:
|
|
|
|
print("unexpected keys:")
|
|
|
|
print(u)
|
|
|
|
|
|
|
|
model.eval()
|
|
|
|
return model
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
LORA_CLIP_MAP = {
|
|
|
|
"mlp.fc1": "mlp_fc1",
|
|
|
|
"mlp.fc2": "mlp_fc2",
|
|
|
|
"self_attn.k_proj": "self_attn_k_proj",
|
|
|
|
"self_attn.q_proj": "self_attn_q_proj",
|
|
|
|
"self_attn.v_proj": "self_attn_v_proj",
|
|
|
|
"self_attn.out_proj": "self_attn_out_proj",
|
|
|
|
}
|
|
|
|
|
2023-03-10 02:41:24 +00:00
|
|
|
LORA_UNET_MAP_ATTENTIONS = {
|
2023-02-03 07:06:34 +00:00
|
|
|
"proj_in": "proj_in",
|
|
|
|
"proj_out": "proj_out",
|
|
|
|
"transformer_blocks.0.attn1.to_q": "transformer_blocks_0_attn1_to_q",
|
|
|
|
"transformer_blocks.0.attn1.to_k": "transformer_blocks_0_attn1_to_k",
|
|
|
|
"transformer_blocks.0.attn1.to_v": "transformer_blocks_0_attn1_to_v",
|
|
|
|
"transformer_blocks.0.attn1.to_out.0": "transformer_blocks_0_attn1_to_out_0",
|
|
|
|
"transformer_blocks.0.attn2.to_q": "transformer_blocks_0_attn2_to_q",
|
|
|
|
"transformer_blocks.0.attn2.to_k": "transformer_blocks_0_attn2_to_k",
|
|
|
|
"transformer_blocks.0.attn2.to_v": "transformer_blocks_0_attn2_to_v",
|
|
|
|
"transformer_blocks.0.attn2.to_out.0": "transformer_blocks_0_attn2_to_out_0",
|
|
|
|
"transformer_blocks.0.ff.net.0.proj": "transformer_blocks_0_ff_net_0_proj",
|
|
|
|
"transformer_blocks.0.ff.net.2": "transformer_blocks_0_ff_net_2",
|
|
|
|
}
|
|
|
|
|
2023-03-10 02:41:24 +00:00
|
|
|
LORA_UNET_MAP_RESNET = {
|
|
|
|
"in_layers.2": "resnets_{}_conv1",
|
|
|
|
"emb_layers.1": "resnets_{}_time_emb_proj",
|
|
|
|
"out_layers.3": "resnets_{}_conv2",
|
|
|
|
"skip_connection": "resnets_{}_conv_shortcut"
|
|
|
|
}
|
2023-02-03 07:06:34 +00:00
|
|
|
|
|
|
|
def load_lora(path, to_load):
|
2023-04-02 03:19:15 +00:00
|
|
|
lora = utils.load_torch_file(path)
|
2023-02-03 07:06:34 +00:00
|
|
|
patch_dict = {}
|
|
|
|
loaded_keys = set()
|
|
|
|
for x in to_load:
|
2023-03-23 07:40:12 +00:00
|
|
|
alpha_name = "{}.alpha".format(x)
|
|
|
|
alpha = None
|
|
|
|
if alpha_name in lora.keys():
|
|
|
|
alpha = lora[alpha_name].item()
|
|
|
|
loaded_keys.add(alpha_name)
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
A_name = "{}.lora_up.weight".format(x)
|
|
|
|
B_name = "{}.lora_down.weight".format(x)
|
2023-03-21 18:51:51 +00:00
|
|
|
mid_name = "{}.lora_mid.weight".format(x)
|
2023-03-23 07:40:12 +00:00
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
if A_name in lora.keys():
|
2023-03-21 18:51:51 +00:00
|
|
|
mid = None
|
|
|
|
if mid_name in lora.keys():
|
|
|
|
mid = lora[mid_name]
|
|
|
|
loaded_keys.add(mid_name)
|
|
|
|
patch_dict[to_load[x]] = (lora[A_name], lora[B_name], alpha, mid)
|
2023-02-03 07:06:34 +00:00
|
|
|
loaded_keys.add(A_name)
|
|
|
|
loaded_keys.add(B_name)
|
2023-03-23 07:40:12 +00:00
|
|
|
|
|
|
|
hada_w1_a_name = "{}.hada_w1_a".format(x)
|
|
|
|
hada_w1_b_name = "{}.hada_w1_b".format(x)
|
|
|
|
hada_w2_a_name = "{}.hada_w2_a".format(x)
|
|
|
|
hada_w2_b_name = "{}.hada_w2_b".format(x)
|
2023-03-23 08:32:25 +00:00
|
|
|
hada_t1_name = "{}.hada_t1".format(x)
|
|
|
|
hada_t2_name = "{}.hada_t2".format(x)
|
2023-03-23 07:40:12 +00:00
|
|
|
if hada_w1_a_name in lora.keys():
|
2023-03-23 08:32:25 +00:00
|
|
|
hada_t1 = None
|
|
|
|
hada_t2 = None
|
|
|
|
if hada_t1_name in lora.keys():
|
|
|
|
hada_t1 = lora[hada_t1_name]
|
|
|
|
hada_t2 = lora[hada_t2_name]
|
|
|
|
loaded_keys.add(hada_t1_name)
|
|
|
|
loaded_keys.add(hada_t2_name)
|
|
|
|
|
|
|
|
patch_dict[to_load[x]] = (lora[hada_w1_a_name], lora[hada_w1_b_name], alpha, lora[hada_w2_a_name], lora[hada_w2_b_name], hada_t1, hada_t2)
|
2023-03-23 07:40:12 +00:00
|
|
|
loaded_keys.add(hada_w1_a_name)
|
|
|
|
loaded_keys.add(hada_w1_b_name)
|
|
|
|
loaded_keys.add(hada_w2_a_name)
|
|
|
|
loaded_keys.add(hada_w2_b_name)
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
for x in lora.keys():
|
|
|
|
if x not in loaded_keys:
|
|
|
|
print("lora key not loaded", x)
|
|
|
|
return patch_dict
|
|
|
|
|
|
|
|
def model_lora_keys(model, key_map={}):
|
|
|
|
sdk = model.state_dict().keys()
|
|
|
|
|
|
|
|
counter = 0
|
|
|
|
for b in range(12):
|
|
|
|
tk = "model.diffusion_model.input_blocks.{}.1".format(b)
|
|
|
|
up_counter = 0
|
2023-03-10 02:41:24 +00:00
|
|
|
for c in LORA_UNET_MAP_ATTENTIONS:
|
2023-02-03 07:06:34 +00:00
|
|
|
k = "{}.{}.weight".format(tk, c)
|
|
|
|
if k in sdk:
|
2023-03-10 02:41:24 +00:00
|
|
|
lora_key = "lora_unet_down_blocks_{}_attentions_{}_{}".format(counter // 2, counter % 2, LORA_UNET_MAP_ATTENTIONS[c])
|
2023-02-05 19:36:28 +00:00
|
|
|
key_map[lora_key] = k
|
2023-02-03 07:06:34 +00:00
|
|
|
up_counter += 1
|
|
|
|
if up_counter >= 4:
|
|
|
|
counter += 1
|
2023-03-10 02:41:24 +00:00
|
|
|
for c in LORA_UNET_MAP_ATTENTIONS:
|
2023-02-03 07:06:34 +00:00
|
|
|
k = "model.diffusion_model.middle_block.1.{}.weight".format(c)
|
|
|
|
if k in sdk:
|
2023-03-10 02:41:24 +00:00
|
|
|
lora_key = "lora_unet_mid_block_attentions_0_{}".format(LORA_UNET_MAP_ATTENTIONS[c])
|
2023-02-05 19:36:28 +00:00
|
|
|
key_map[lora_key] = k
|
2023-02-03 07:06:34 +00:00
|
|
|
counter = 3
|
|
|
|
for b in range(12):
|
|
|
|
tk = "model.diffusion_model.output_blocks.{}.1".format(b)
|
|
|
|
up_counter = 0
|
2023-03-10 02:41:24 +00:00
|
|
|
for c in LORA_UNET_MAP_ATTENTIONS:
|
2023-02-03 07:06:34 +00:00
|
|
|
k = "{}.{}.weight".format(tk, c)
|
|
|
|
if k in sdk:
|
2023-03-10 02:41:24 +00:00
|
|
|
lora_key = "lora_unet_up_blocks_{}_attentions_{}_{}".format(counter // 3, counter % 3, LORA_UNET_MAP_ATTENTIONS[c])
|
2023-02-05 19:36:28 +00:00
|
|
|
key_map[lora_key] = k
|
2023-02-03 07:06:34 +00:00
|
|
|
up_counter += 1
|
|
|
|
if up_counter >= 4:
|
|
|
|
counter += 1
|
|
|
|
counter = 0
|
2023-02-05 06:54:09 +00:00
|
|
|
text_model_lora_key = "lora_te_text_model_encoder_layers_{}_{}"
|
2023-02-05 19:36:28 +00:00
|
|
|
for b in range(24):
|
2023-02-03 07:06:34 +00:00
|
|
|
for c in LORA_CLIP_MAP:
|
|
|
|
k = "transformer.text_model.encoder.layers.{}.{}.weight".format(b, c)
|
|
|
|
if k in sdk:
|
2023-02-05 06:54:09 +00:00
|
|
|
lora_key = text_model_lora_key.format(b, LORA_CLIP_MAP[c])
|
2023-02-05 19:36:28 +00:00
|
|
|
key_map[lora_key] = k
|
2023-02-05 06:54:09 +00:00
|
|
|
|
2023-03-10 02:41:24 +00:00
|
|
|
|
|
|
|
#Locon stuff
|
|
|
|
ds_counter = 0
|
|
|
|
counter = 0
|
|
|
|
for b in range(12):
|
|
|
|
tk = "model.diffusion_model.input_blocks.{}.0".format(b)
|
|
|
|
key_in = False
|
|
|
|
for c in LORA_UNET_MAP_RESNET:
|
|
|
|
k = "{}.{}.weight".format(tk, c)
|
|
|
|
if k in sdk:
|
|
|
|
lora_key = "lora_unet_down_blocks_{}_{}".format(counter // 2, LORA_UNET_MAP_RESNET[c].format(counter % 2))
|
|
|
|
key_map[lora_key] = k
|
|
|
|
key_in = True
|
|
|
|
for bb in range(3):
|
|
|
|
k = "{}.{}.op.weight".format(tk[:-2], bb)
|
|
|
|
if k in sdk:
|
|
|
|
lora_key = "lora_unet_down_blocks_{}_downsamplers_0_conv".format(ds_counter)
|
|
|
|
key_map[lora_key] = k
|
|
|
|
ds_counter += 1
|
|
|
|
if key_in:
|
|
|
|
counter += 1
|
|
|
|
|
|
|
|
counter = 0
|
|
|
|
for b in range(3):
|
|
|
|
tk = "model.diffusion_model.middle_block.{}".format(b)
|
|
|
|
key_in = False
|
|
|
|
for c in LORA_UNET_MAP_RESNET:
|
|
|
|
k = "{}.{}.weight".format(tk, c)
|
|
|
|
if k in sdk:
|
|
|
|
lora_key = "lora_unet_mid_block_{}".format(LORA_UNET_MAP_RESNET[c].format(counter))
|
|
|
|
key_map[lora_key] = k
|
|
|
|
key_in = True
|
|
|
|
if key_in:
|
|
|
|
counter += 1
|
|
|
|
|
|
|
|
counter = 0
|
|
|
|
us_counter = 0
|
|
|
|
for b in range(12):
|
|
|
|
tk = "model.diffusion_model.output_blocks.{}.0".format(b)
|
|
|
|
key_in = False
|
|
|
|
for c in LORA_UNET_MAP_RESNET:
|
|
|
|
k = "{}.{}.weight".format(tk, c)
|
|
|
|
if k in sdk:
|
|
|
|
lora_key = "lora_unet_up_blocks_{}_{}".format(counter // 3, LORA_UNET_MAP_RESNET[c].format(counter % 3))
|
|
|
|
key_map[lora_key] = k
|
|
|
|
key_in = True
|
|
|
|
for bb in range(3):
|
|
|
|
k = "{}.{}.conv.weight".format(tk[:-2], bb)
|
|
|
|
if k in sdk:
|
|
|
|
lora_key = "lora_unet_up_blocks_{}_upsamplers_0_conv".format(us_counter)
|
|
|
|
key_map[lora_key] = k
|
|
|
|
us_counter += 1
|
|
|
|
if key_in:
|
|
|
|
counter += 1
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
return key_map
|
|
|
|
|
|
|
|
class ModelPatcher:
|
|
|
|
def __init__(self, model):
|
|
|
|
self.model = model
|
|
|
|
self.patches = []
|
|
|
|
self.backup = {}
|
2023-03-31 21:19:58 +00:00
|
|
|
self.model_options = {"transformer_options":{}}
|
2023-02-03 07:06:34 +00:00
|
|
|
|
|
|
|
def clone(self):
|
|
|
|
n = ModelPatcher(self.model)
|
|
|
|
n.patches = self.patches[:]
|
2023-03-31 21:19:58 +00:00
|
|
|
n.model_options = copy.deepcopy(self.model_options)
|
2023-02-03 07:06:34 +00:00
|
|
|
return n
|
|
|
|
|
2023-03-31 21:19:58 +00:00
|
|
|
def set_model_tomesd(self, ratio):
|
|
|
|
self.model_options["transformer_options"]["tomesd"] = {"ratio": ratio}
|
|
|
|
|
|
|
|
def model_dtype(self):
|
|
|
|
return self.model.diffusion_model.dtype
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
def add_patches(self, patches, strength=1.0):
|
|
|
|
p = {}
|
|
|
|
model_sd = self.model.state_dict()
|
|
|
|
for k in patches:
|
2023-02-05 19:36:28 +00:00
|
|
|
if k in model_sd:
|
2023-02-03 07:06:34 +00:00
|
|
|
p[k] = patches[k]
|
|
|
|
self.patches += [(strength, p)]
|
|
|
|
return p.keys()
|
|
|
|
|
|
|
|
def patch_model(self):
|
|
|
|
model_sd = self.model.state_dict()
|
|
|
|
for p in self.patches:
|
|
|
|
for k in p[1]:
|
|
|
|
v = p[1][k]
|
2023-02-05 19:36:28 +00:00
|
|
|
key = k
|
2023-02-05 06:54:09 +00:00
|
|
|
if key not in model_sd:
|
2023-02-03 07:06:34 +00:00
|
|
|
print("could not patch. key doesn't exist in model:", k)
|
|
|
|
continue
|
|
|
|
|
2023-02-05 06:54:09 +00:00
|
|
|
weight = model_sd[key]
|
|
|
|
if key not in self.backup:
|
|
|
|
self.backup[key] = weight.clone()
|
2023-02-03 07:06:34 +00:00
|
|
|
|
|
|
|
alpha = p[0]
|
2023-03-23 07:40:12 +00:00
|
|
|
|
|
|
|
if len(v) == 4: #lora/locon
|
|
|
|
mat1 = v[0]
|
|
|
|
mat2 = v[1]
|
|
|
|
if v[2] is not None:
|
|
|
|
alpha *= v[2] / mat2.shape[0]
|
|
|
|
if v[3] is not None:
|
|
|
|
#locon mid weights, hopefully the math is fine because I didn't properly test it
|
|
|
|
final_shape = [mat2.shape[1], mat2.shape[0], v[3].shape[2], v[3].shape[3]]
|
|
|
|
mat2 = torch.mm(mat2.transpose(0, 1).flatten(start_dim=1).float(), v[3].transpose(0, 1).flatten(start_dim=1).float()).reshape(final_shape).transpose(0, 1)
|
|
|
|
weight += (alpha * torch.mm(mat1.flatten(start_dim=1).float(), mat2.flatten(start_dim=1).float())).reshape(weight.shape).type(weight.dtype).to(weight.device)
|
|
|
|
else: #loha
|
|
|
|
w1a = v[0]
|
|
|
|
w1b = v[1]
|
|
|
|
if v[2] is not None:
|
|
|
|
alpha *= v[2] / w1b.shape[0]
|
|
|
|
w2a = v[3]
|
|
|
|
w2b = v[4]
|
2023-03-23 08:32:25 +00:00
|
|
|
if v[5] is not None: #cp decomposition
|
|
|
|
t1 = v[5]
|
|
|
|
t2 = v[6]
|
|
|
|
m1 = torch.einsum('i j k l, j r, i p -> p r k l', t1.float(), w1b.float(), w1a.float())
|
|
|
|
m2 = torch.einsum('i j k l, j r, i p -> p r k l', t2.float(), w2b.float(), w2a.float())
|
|
|
|
else:
|
|
|
|
m1 = torch.mm(w1a.float(), w1b.float())
|
|
|
|
m2 = torch.mm(w2a.float(), w2b.float())
|
|
|
|
|
|
|
|
weight += (alpha * m1 * m2).reshape(weight.shape).type(weight.dtype).to(weight.device)
|
2023-02-03 07:06:34 +00:00
|
|
|
return self.model
|
|
|
|
def unpatch_model(self):
|
|
|
|
model_sd = self.model.state_dict()
|
2023-02-28 17:27:43 +00:00
|
|
|
keys = list(self.backup.keys())
|
|
|
|
for k in keys:
|
2023-02-03 07:06:34 +00:00
|
|
|
model_sd[k][:] = self.backup[k]
|
2023-02-28 17:27:43 +00:00
|
|
|
del self.backup[k]
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
self.backup = {}
|
|
|
|
|
|
|
|
def load_lora_for_models(model, clip, lora_path, strength_model, strength_clip):
|
|
|
|
key_map = model_lora_keys(model.model)
|
|
|
|
key_map = model_lora_keys(clip.cond_stage_model, key_map)
|
|
|
|
loaded = load_lora(lora_path, key_map)
|
|
|
|
new_modelpatcher = model.clone()
|
|
|
|
k = new_modelpatcher.add_patches(loaded, strength_model)
|
|
|
|
new_clip = clip.clone()
|
|
|
|
k1 = new_clip.add_patches(loaded, strength_clip)
|
|
|
|
k = set(k)
|
|
|
|
k1 = set(k1)
|
|
|
|
for x in loaded:
|
|
|
|
if (x not in k) and (x not in k1):
|
|
|
|
print("NOT LOADED", x)
|
|
|
|
|
|
|
|
return (new_modelpatcher, new_clip)
|
2023-01-03 06:53:32 +00:00
|
|
|
|
|
|
|
|
|
|
|
class CLIP:
|
2023-02-03 07:06:34 +00:00
|
|
|
def __init__(self, config={}, embedding_directory=None, no_init=False):
|
|
|
|
if no_init:
|
|
|
|
return
|
2023-01-03 06:53:32 +00:00
|
|
|
self.target_clip = config["target"]
|
2023-01-29 23:46:44 +00:00
|
|
|
if "params" in config:
|
|
|
|
params = config["params"]
|
|
|
|
else:
|
|
|
|
params = {}
|
|
|
|
|
2023-01-03 06:53:32 +00:00
|
|
|
if self.target_clip == "ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder":
|
|
|
|
clip = sd2_clip.SD2ClipModel
|
|
|
|
tokenizer = sd2_clip.SD2Tokenizer
|
|
|
|
elif self.target_clip == "ldm.modules.encoders.modules.FrozenCLIPEmbedder":
|
|
|
|
clip = sd1_clip.SD1ClipModel
|
|
|
|
tokenizer = sd1_clip.SD1Tokenizer
|
2023-01-29 23:46:44 +00:00
|
|
|
|
|
|
|
self.cond_stage_model = clip(**(params))
|
2023-02-05 20:49:03 +00:00
|
|
|
self.tokenizer = tokenizer(embedding_directory=embedding_directory)
|
2023-02-03 07:06:34 +00:00
|
|
|
self.patcher = ModelPatcher(self.cond_stage_model)
|
2023-03-06 16:34:02 +00:00
|
|
|
self.layer_idx = None
|
2023-02-03 07:06:34 +00:00
|
|
|
|
|
|
|
def clone(self):
|
|
|
|
n = CLIP(no_init=True)
|
|
|
|
n.target_clip = self.target_clip
|
|
|
|
n.patcher = self.patcher.clone()
|
|
|
|
n.cond_stage_model = self.cond_stage_model
|
|
|
|
n.tokenizer = self.tokenizer
|
2023-03-03 18:04:36 +00:00
|
|
|
n.layer_idx = self.layer_idx
|
2023-02-03 07:06:34 +00:00
|
|
|
return n
|
|
|
|
|
2023-02-05 20:20:18 +00:00
|
|
|
def load_from_state_dict(self, sd):
|
|
|
|
self.cond_stage_model.transformer.load_state_dict(sd, strict=False)
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
def add_patches(self, patches, strength=1.0):
|
|
|
|
return self.patcher.add_patches(patches, strength)
|
2023-01-03 06:53:32 +00:00
|
|
|
|
2023-02-05 20:20:18 +00:00
|
|
|
def clip_layer(self, layer_idx):
|
2023-03-03 18:04:36 +00:00
|
|
|
self.layer_idx = layer_idx
|
2023-02-05 20:20:18 +00:00
|
|
|
|
2023-04-14 19:16:55 +00:00
|
|
|
def tokenize(self, text, return_word_ids=False):
|
|
|
|
return self.tokenizer.tokenize_with_weights(text, return_word_ids)
|
2023-04-13 20:06:50 +00:00
|
|
|
|
2023-04-14 19:16:55 +00:00
|
|
|
def encode(self, text, from_tokens=False):
|
2023-03-06 16:34:02 +00:00
|
|
|
if self.layer_idx is not None:
|
|
|
|
self.cond_stage_model.clip_layer(self.layer_idx)
|
2023-04-14 19:16:55 +00:00
|
|
|
if from_tokens:
|
|
|
|
tokens = text
|
|
|
|
else:
|
|
|
|
tokens = self.tokenizer.tokenize_with_weights(text)
|
2023-02-03 07:06:34 +00:00
|
|
|
try:
|
|
|
|
self.patcher.patch_model()
|
|
|
|
cond = self.cond_stage_model.encode_token_weights(tokens)
|
|
|
|
self.patcher.unpatch_model()
|
|
|
|
except Exception as e:
|
|
|
|
self.patcher.unpatch_model()
|
|
|
|
raise e
|
2023-01-03 06:53:32 +00:00
|
|
|
return cond
|
|
|
|
|
|
|
|
class VAE:
|
2023-03-06 15:50:50 +00:00
|
|
|
def __init__(self, ckpt_path=None, scale_factor=0.18215, device=None, config=None):
|
2023-01-03 06:53:32 +00:00
|
|
|
if config is None:
|
|
|
|
#default SD1.x/SD2.x VAE parameters
|
|
|
|
ddconfig = {'double_z': True, 'z_channels': 4, 'resolution': 256, 'in_channels': 3, 'out_ch': 3, 'ch': 128, 'ch_mult': [1, 2, 4, 4], 'num_res_blocks': 2, 'attn_resolutions': [], 'dropout': 0.0}
|
|
|
|
self.first_stage_model = AutoencoderKL(ddconfig, {'target': 'torch.nn.Identity'}, 4, monitor="val/rec_loss", ckpt_path=ckpt_path)
|
|
|
|
else:
|
|
|
|
self.first_stage_model = AutoencoderKL(**(config['params']), ckpt_path=ckpt_path)
|
|
|
|
self.first_stage_model = self.first_stage_model.eval()
|
|
|
|
self.scale_factor = scale_factor
|
2023-03-06 15:50:50 +00:00
|
|
|
if device is None:
|
|
|
|
device = model_management.get_torch_device()
|
2023-01-03 06:53:32 +00:00
|
|
|
self.device = device
|
|
|
|
|
2023-03-22 18:49:00 +00:00
|
|
|
def decode_tiled_(self, samples, tile_x=64, tile_y=64, overlap = 16):
|
|
|
|
decode_fn = lambda a: (self.first_stage_model.decode(1. / self.scale_factor * a.to(self.device)) + 1.0)
|
|
|
|
output = torch.clamp((
|
|
|
|
(utils.tiled_scale(samples, decode_fn, tile_x // 2, tile_y * 2, overlap, upscale_amount = 8) +
|
|
|
|
utils.tiled_scale(samples, decode_fn, tile_x * 2, tile_y // 2, overlap, upscale_amount = 8) +
|
|
|
|
utils.tiled_scale(samples, decode_fn, tile_x, tile_y, overlap, upscale_amount = 8))
|
|
|
|
/ 3.0) / 2.0, min=0.0, max=1.0)
|
|
|
|
return output
|
|
|
|
|
|
|
|
def decode(self, samples_in):
|
2023-02-08 08:17:54 +00:00
|
|
|
model_management.unload_model()
|
2023-01-03 06:53:32 +00:00
|
|
|
self.first_stage_model = self.first_stage_model.to(self.device)
|
2023-03-22 18:49:00 +00:00
|
|
|
try:
|
2023-03-29 06:24:37 +00:00
|
|
|
free_memory = model_management.get_free_memory(self.device)
|
|
|
|
batch_number = int((free_memory * 0.7) / (2562 * samples_in.shape[2] * samples_in.shape[3] * 64))
|
|
|
|
batch_number = max(1, batch_number)
|
|
|
|
|
|
|
|
pixel_samples = torch.empty((samples_in.shape[0], 3, round(samples_in.shape[2] * 8), round(samples_in.shape[3] * 8)), device="cpu")
|
|
|
|
for x in range(0, samples_in.shape[0], batch_number):
|
|
|
|
samples = samples_in[x:x+batch_number].to(self.device)
|
|
|
|
pixel_samples[x:x+batch_number] = torch.clamp((self.first_stage_model.decode(1. / self.scale_factor * samples) + 1.0) / 2.0, min=0.0, max=1.0).cpu()
|
2023-03-22 18:49:00 +00:00
|
|
|
except model_management.OOM_EXCEPTION as e:
|
|
|
|
print("Warning: Ran out of memory when regular VAE decoding, retrying with tiled VAE decoding.")
|
|
|
|
pixel_samples = self.decode_tiled_(samples_in)
|
|
|
|
|
2023-01-03 06:53:32 +00:00
|
|
|
self.first_stage_model = self.first_stage_model.cpu()
|
|
|
|
pixel_samples = pixel_samples.cpu().movedim(1,-1)
|
|
|
|
return pixel_samples
|
|
|
|
|
2023-03-22 07:29:09 +00:00
|
|
|
def decode_tiled(self, samples, tile_x=64, tile_y=64, overlap = 16):
|
2023-02-24 07:10:10 +00:00
|
|
|
model_management.unload_model()
|
|
|
|
self.first_stage_model = self.first_stage_model.to(self.device)
|
2023-03-22 18:49:00 +00:00
|
|
|
output = self.decode_tiled_(samples, tile_x, tile_y, overlap)
|
2023-02-24 07:10:10 +00:00
|
|
|
self.first_stage_model = self.first_stage_model.cpu()
|
|
|
|
return output.movedim(1,-1)
|
|
|
|
|
2023-01-03 06:53:32 +00:00
|
|
|
def encode(self, pixel_samples):
|
2023-02-08 08:17:54 +00:00
|
|
|
model_management.unload_model()
|
2023-01-03 06:53:32 +00:00
|
|
|
self.first_stage_model = self.first_stage_model.to(self.device)
|
|
|
|
pixel_samples = pixel_samples.movedim(-1,1).to(self.device)
|
|
|
|
samples = self.first_stage_model.encode(2. * pixel_samples - 1.).sample() * self.scale_factor
|
|
|
|
self.first_stage_model = self.first_stage_model.cpu()
|
|
|
|
samples = samples.cpu()
|
|
|
|
return samples
|
|
|
|
|
2023-03-11 20:28:15 +00:00
|
|
|
def encode_tiled(self, pixel_samples, tile_x=512, tile_y=512, overlap = 64):
|
|
|
|
model_management.unload_model()
|
|
|
|
self.first_stage_model = self.first_stage_model.to(self.device)
|
|
|
|
pixel_samples = pixel_samples.movedim(-1,1).to(self.device)
|
|
|
|
samples = utils.tiled_scale(pixel_samples, lambda a: self.first_stage_model.encode(2. * a - 1.).sample() * self.scale_factor, tile_x, tile_y, overlap, upscale_amount = (1/8), out_channels=4)
|
2023-03-22 07:29:09 +00:00
|
|
|
samples += utils.tiled_scale(pixel_samples, lambda a: self.first_stage_model.encode(2. * a - 1.).sample() * self.scale_factor, tile_x * 2, tile_y // 2, overlap, upscale_amount = (1/8), out_channels=4)
|
|
|
|
samples += utils.tiled_scale(pixel_samples, lambda a: self.first_stage_model.encode(2. * a - 1.).sample() * self.scale_factor, tile_x // 2, tile_y * 2, overlap, upscale_amount = (1/8), out_channels=4)
|
|
|
|
samples /= 3.0
|
2023-03-11 20:28:15 +00:00
|
|
|
self.first_stage_model = self.first_stage_model.cpu()
|
|
|
|
samples = samples.cpu()
|
|
|
|
return samples
|
2023-02-25 19:57:28 +00:00
|
|
|
|
|
|
|
def resize_image_to(tensor, target_latent_tensor, batched_number):
|
|
|
|
tensor = utils.common_upscale(tensor, target_latent_tensor.shape[3] * 8, target_latent_tensor.shape[2] * 8, 'nearest-exact', "center")
|
|
|
|
target_batch_size = target_latent_tensor.shape[0]
|
|
|
|
|
|
|
|
current_batch_size = tensor.shape[0]
|
|
|
|
print(current_batch_size, target_batch_size)
|
|
|
|
if current_batch_size == 1:
|
|
|
|
return tensor
|
|
|
|
|
|
|
|
per_batch = target_batch_size // batched_number
|
|
|
|
tensor = tensor[:per_batch]
|
|
|
|
|
|
|
|
if per_batch > tensor.shape[0]:
|
|
|
|
tensor = torch.cat([tensor] * (per_batch // tensor.shape[0]) + [tensor[:(per_batch % tensor.shape[0])]], dim=0)
|
|
|
|
|
|
|
|
current_batch_size = tensor.shape[0]
|
|
|
|
if current_batch_size == target_batch_size:
|
|
|
|
return tensor
|
|
|
|
else:
|
|
|
|
return torch.cat([tensor] * batched_number, dim=0)
|
|
|
|
|
2023-02-16 15:38:08 +00:00
|
|
|
class ControlNet:
|
2023-03-06 15:50:50 +00:00
|
|
|
def __init__(self, control_model, device=None):
|
2023-02-16 15:38:08 +00:00
|
|
|
self.control_model = control_model
|
|
|
|
self.cond_hint_original = None
|
|
|
|
self.cond_hint = None
|
2023-02-16 23:08:01 +00:00
|
|
|
self.strength = 1.0
|
2023-03-06 15:50:50 +00:00
|
|
|
if device is None:
|
|
|
|
device = model_management.get_torch_device()
|
2023-02-17 20:31:38 +00:00
|
|
|
self.device = device
|
2023-02-21 06:18:53 +00:00
|
|
|
self.previous_controlnet = None
|
2023-02-16 15:38:08 +00:00
|
|
|
|
2023-02-25 19:57:28 +00:00
|
|
|
def get_control(self, x_noisy, t, cond_txt, batched_number):
|
2023-02-21 06:18:53 +00:00
|
|
|
control_prev = None
|
|
|
|
if self.previous_controlnet is not None:
|
2023-02-26 03:12:22 +00:00
|
|
|
control_prev = self.previous_controlnet.get_control(x_noisy, t, cond_txt, batched_number)
|
2023-02-21 06:18:53 +00:00
|
|
|
|
2023-02-17 20:31:38 +00:00
|
|
|
output_dtype = x_noisy.dtype
|
2023-02-16 15:38:08 +00:00
|
|
|
if self.cond_hint is None or x_noisy.shape[2] * 8 != self.cond_hint.shape[2] or x_noisy.shape[3] * 8 != self.cond_hint.shape[3]:
|
|
|
|
if self.cond_hint is not None:
|
|
|
|
del self.cond_hint
|
|
|
|
self.cond_hint = None
|
2023-02-25 19:57:28 +00:00
|
|
|
self.cond_hint = resize_image_to(self.cond_hint_original, x_noisy, batched_number).to(self.control_model.dtype).to(self.device)
|
2023-02-17 20:31:38 +00:00
|
|
|
|
|
|
|
if self.control_model.dtype == torch.float16:
|
|
|
|
precision_scope = torch.autocast
|
|
|
|
else:
|
|
|
|
precision_scope = contextlib.nullcontext
|
|
|
|
|
2023-03-06 15:50:50 +00:00
|
|
|
with precision_scope(model_management.get_autocast_device(self.device)):
|
2023-02-17 20:45:29 +00:00
|
|
|
self.control_model = model_management.load_if_low_vram(self.control_model)
|
2023-02-17 20:31:38 +00:00
|
|
|
control = self.control_model(x=x_noisy, hint=self.cond_hint, timesteps=t, context=cond_txt)
|
2023-02-17 20:45:29 +00:00
|
|
|
self.control_model = model_management.unload_if_low_vram(self.control_model)
|
2023-02-25 05:55:42 +00:00
|
|
|
out = {'middle':[], 'output': []}
|
2023-02-17 20:31:38 +00:00
|
|
|
autocast_enabled = torch.is_autocast_enabled()
|
2023-02-21 06:18:53 +00:00
|
|
|
|
|
|
|
for i in range(len(control)):
|
2023-02-25 04:36:17 +00:00
|
|
|
if i == (len(control) - 1):
|
|
|
|
key = 'middle'
|
|
|
|
index = 0
|
|
|
|
else:
|
|
|
|
key = 'output'
|
|
|
|
index = i
|
2023-02-21 06:18:53 +00:00
|
|
|
x = control[i]
|
2023-02-16 23:08:01 +00:00
|
|
|
x *= self.strength
|
2023-02-17 20:31:38 +00:00
|
|
|
if x.dtype != output_dtype and not autocast_enabled:
|
|
|
|
x = x.to(output_dtype)
|
2023-02-21 06:18:53 +00:00
|
|
|
|
2023-02-25 04:36:17 +00:00
|
|
|
if control_prev is not None and key in control_prev:
|
|
|
|
prev = control_prev[key][index]
|
|
|
|
if prev is not None:
|
|
|
|
x += prev
|
|
|
|
out[key].append(x)
|
|
|
|
if control_prev is not None and 'input' in control_prev:
|
|
|
|
out['input'] = control_prev['input']
|
2023-02-17 20:31:38 +00:00
|
|
|
return out
|
2023-02-16 15:38:08 +00:00
|
|
|
|
2023-02-16 23:08:01 +00:00
|
|
|
def set_cond_hint(self, cond_hint, strength=1.0):
|
2023-02-16 15:38:08 +00:00
|
|
|
self.cond_hint_original = cond_hint
|
2023-02-16 23:08:01 +00:00
|
|
|
self.strength = strength
|
2023-02-16 15:38:08 +00:00
|
|
|
return self
|
|
|
|
|
2023-02-21 06:18:53 +00:00
|
|
|
def set_previous_controlnet(self, controlnet):
|
|
|
|
self.previous_controlnet = controlnet
|
|
|
|
return self
|
|
|
|
|
2023-02-16 15:38:08 +00:00
|
|
|
def cleanup(self):
|
2023-02-21 06:18:53 +00:00
|
|
|
if self.previous_controlnet is not None:
|
|
|
|
self.previous_controlnet.cleanup()
|
2023-02-16 15:38:08 +00:00
|
|
|
if self.cond_hint is not None:
|
|
|
|
del self.cond_hint
|
|
|
|
self.cond_hint = None
|
|
|
|
|
|
|
|
def copy(self):
|
|
|
|
c = ControlNet(self.control_model)
|
|
|
|
c.cond_hint_original = self.cond_hint_original
|
2023-02-16 23:08:01 +00:00
|
|
|
c.strength = self.strength
|
2023-02-16 15:38:08 +00:00
|
|
|
return c
|
|
|
|
|
2023-02-21 06:18:53 +00:00
|
|
|
def get_control_models(self):
|
|
|
|
out = []
|
|
|
|
if self.previous_controlnet is not None:
|
|
|
|
out += self.previous_controlnet.get_control_models()
|
|
|
|
out.append(self.control_model)
|
|
|
|
return out
|
|
|
|
|
2023-02-23 04:22:03 +00:00
|
|
|
def load_controlnet(ckpt_path, model=None):
|
2023-04-02 03:19:15 +00:00
|
|
|
controlnet_data = utils.load_torch_file(ckpt_path)
|
2023-02-16 15:38:08 +00:00
|
|
|
pth_key = 'control_model.input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight'
|
|
|
|
pth = False
|
|
|
|
sd2 = False
|
|
|
|
key = 'input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight'
|
|
|
|
if pth_key in controlnet_data:
|
|
|
|
pth = True
|
|
|
|
key = pth_key
|
|
|
|
elif key in controlnet_data:
|
|
|
|
pass
|
|
|
|
else:
|
2023-03-17 22:17:59 +00:00
|
|
|
net = load_t2i_adapter(controlnet_data)
|
|
|
|
if net is None:
|
|
|
|
print("error checkpoint does not contain controlnet or t2i adapter data", ckpt_path)
|
|
|
|
return net
|
2023-02-16 15:38:08 +00:00
|
|
|
|
|
|
|
context_dim = controlnet_data[key].shape[1]
|
2023-02-17 20:31:38 +00:00
|
|
|
|
|
|
|
use_fp16 = False
|
2023-03-06 15:50:50 +00:00
|
|
|
if model_management.should_use_fp16() and controlnet_data[key].dtype == torch.float16:
|
2023-02-17 20:31:38 +00:00
|
|
|
use_fp16 = True
|
|
|
|
|
2023-03-08 06:13:38 +00:00
|
|
|
if context_dim == 768:
|
|
|
|
#SD1.x
|
|
|
|
control_model = cldm.ControlNet(image_size=32,
|
|
|
|
in_channels=4,
|
|
|
|
hint_channels=3,
|
|
|
|
model_channels=320,
|
|
|
|
attention_resolutions=[ 4, 2, 1 ],
|
|
|
|
num_res_blocks=2,
|
|
|
|
channel_mult=[ 1, 2, 4, 4 ],
|
|
|
|
num_heads=8,
|
|
|
|
use_spatial_transformer=True,
|
|
|
|
transformer_depth=1,
|
|
|
|
context_dim=context_dim,
|
|
|
|
use_checkpoint=True,
|
|
|
|
legacy=False,
|
|
|
|
use_fp16=use_fp16)
|
|
|
|
else:
|
|
|
|
#SD2.x
|
|
|
|
control_model = cldm.ControlNet(image_size=32,
|
|
|
|
in_channels=4,
|
|
|
|
hint_channels=3,
|
|
|
|
model_channels=320,
|
|
|
|
attention_resolutions=[ 4, 2, 1 ],
|
|
|
|
num_res_blocks=2,
|
|
|
|
channel_mult=[ 1, 2, 4, 4 ],
|
|
|
|
num_head_channels=64,
|
|
|
|
use_spatial_transformer=True,
|
|
|
|
use_linear_in_transformer=True,
|
|
|
|
transformer_depth=1,
|
|
|
|
context_dim=context_dim,
|
|
|
|
use_checkpoint=True,
|
|
|
|
legacy=False,
|
|
|
|
use_fp16=use_fp16)
|
2023-02-16 15:38:08 +00:00
|
|
|
if pth:
|
2023-02-23 04:22:03 +00:00
|
|
|
if 'difference' in controlnet_data:
|
|
|
|
if model is not None:
|
|
|
|
m = model.patch_model()
|
|
|
|
model_sd = m.state_dict()
|
|
|
|
for x in controlnet_data:
|
|
|
|
c_m = "control_model."
|
|
|
|
if x.startswith(c_m):
|
|
|
|
sd_key = "model.diffusion_model.{}".format(x[len(c_m):])
|
|
|
|
if sd_key in model_sd:
|
|
|
|
cd = controlnet_data[x]
|
|
|
|
cd += model_sd[sd_key].type(cd.dtype).to(cd.device)
|
|
|
|
model.unpatch_model()
|
|
|
|
else:
|
|
|
|
print("WARNING: Loaded a diff controlnet without a model. It will very likely not work.")
|
|
|
|
|
2023-02-16 15:38:08 +00:00
|
|
|
class WeightsLoader(torch.nn.Module):
|
|
|
|
pass
|
|
|
|
w = WeightsLoader()
|
|
|
|
w.control_model = control_model
|
|
|
|
w.load_state_dict(controlnet_data, strict=False)
|
|
|
|
else:
|
|
|
|
control_model.load_state_dict(controlnet_data, strict=False)
|
|
|
|
|
2023-03-19 14:50:38 +00:00
|
|
|
if use_fp16:
|
|
|
|
control_model = control_model.half()
|
|
|
|
|
2023-02-16 15:38:08 +00:00
|
|
|
control = ControlNet(control_model)
|
|
|
|
return control
|
|
|
|
|
2023-02-25 05:55:42 +00:00
|
|
|
class T2IAdapter:
|
2023-03-06 15:50:50 +00:00
|
|
|
def __init__(self, t2i_model, channels_in, device=None):
|
2023-02-25 05:55:42 +00:00
|
|
|
self.t2i_model = t2i_model
|
|
|
|
self.channels_in = channels_in
|
|
|
|
self.strength = 1.0
|
2023-03-06 15:50:50 +00:00
|
|
|
if device is None:
|
|
|
|
device = model_management.get_torch_device()
|
2023-02-25 05:55:42 +00:00
|
|
|
self.device = device
|
|
|
|
self.previous_controlnet = None
|
|
|
|
self.control_input = None
|
|
|
|
self.cond_hint_original = None
|
|
|
|
self.cond_hint = None
|
|
|
|
|
2023-02-25 19:57:28 +00:00
|
|
|
def get_control(self, x_noisy, t, cond_txt, batched_number):
|
2023-02-25 05:55:42 +00:00
|
|
|
control_prev = None
|
|
|
|
if self.previous_controlnet is not None:
|
2023-02-26 03:12:22 +00:00
|
|
|
control_prev = self.previous_controlnet.get_control(x_noisy, t, cond_txt, batched_number)
|
2023-02-25 05:55:42 +00:00
|
|
|
|
|
|
|
if self.cond_hint is None or x_noisy.shape[2] * 8 != self.cond_hint.shape[2] or x_noisy.shape[3] * 8 != self.cond_hint.shape[3]:
|
|
|
|
if self.cond_hint is not None:
|
|
|
|
del self.cond_hint
|
|
|
|
self.cond_hint = None
|
2023-02-25 19:57:28 +00:00
|
|
|
self.cond_hint = resize_image_to(self.cond_hint_original, x_noisy, batched_number).float().to(self.device)
|
2023-02-25 05:55:42 +00:00
|
|
|
if self.channels_in == 1 and self.cond_hint.shape[1] > 1:
|
|
|
|
self.cond_hint = torch.mean(self.cond_hint, 1, keepdim=True)
|
|
|
|
self.t2i_model.to(self.device)
|
|
|
|
self.control_input = self.t2i_model(self.cond_hint)
|
|
|
|
self.t2i_model.cpu()
|
|
|
|
|
|
|
|
output_dtype = x_noisy.dtype
|
|
|
|
out = {'input':[]}
|
|
|
|
|
2023-02-25 17:19:03 +00:00
|
|
|
autocast_enabled = torch.is_autocast_enabled()
|
2023-02-25 05:55:42 +00:00
|
|
|
for i in range(len(self.control_input)):
|
|
|
|
key = 'input'
|
|
|
|
x = self.control_input[i] * self.strength
|
|
|
|
if x.dtype != output_dtype and not autocast_enabled:
|
|
|
|
x = x.to(output_dtype)
|
|
|
|
|
|
|
|
if control_prev is not None and key in control_prev:
|
|
|
|
index = len(control_prev[key]) - i * 3 - 3
|
|
|
|
prev = control_prev[key][index]
|
|
|
|
if prev is not None:
|
|
|
|
x += prev
|
|
|
|
out[key].insert(0, None)
|
|
|
|
out[key].insert(0, None)
|
|
|
|
out[key].insert(0, x)
|
|
|
|
|
|
|
|
if control_prev is not None and 'input' in control_prev:
|
|
|
|
for i in range(len(out['input'])):
|
|
|
|
if out['input'][i] is None:
|
|
|
|
out['input'][i] = control_prev['input'][i]
|
|
|
|
if control_prev is not None and 'middle' in control_prev:
|
|
|
|
out['middle'] = control_prev['middle']
|
|
|
|
if control_prev is not None and 'output' in control_prev:
|
|
|
|
out['output'] = control_prev['output']
|
|
|
|
return out
|
|
|
|
|
|
|
|
def set_cond_hint(self, cond_hint, strength=1.0):
|
|
|
|
self.cond_hint_original = cond_hint
|
|
|
|
self.strength = strength
|
|
|
|
return self
|
|
|
|
|
|
|
|
def set_previous_controlnet(self, controlnet):
|
|
|
|
self.previous_controlnet = controlnet
|
|
|
|
return self
|
|
|
|
|
|
|
|
def copy(self):
|
|
|
|
c = T2IAdapter(self.t2i_model, self.channels_in)
|
|
|
|
c.cond_hint_original = self.cond_hint_original
|
|
|
|
c.strength = self.strength
|
|
|
|
return c
|
|
|
|
|
|
|
|
def cleanup(self):
|
|
|
|
if self.previous_controlnet is not None:
|
|
|
|
self.previous_controlnet.cleanup()
|
|
|
|
if self.cond_hint is not None:
|
|
|
|
del self.cond_hint
|
|
|
|
self.cond_hint = None
|
|
|
|
|
|
|
|
def get_control_models(self):
|
|
|
|
out = []
|
|
|
|
if self.previous_controlnet is not None:
|
|
|
|
out += self.previous_controlnet.get_control_models()
|
|
|
|
return out
|
|
|
|
|
2023-03-17 22:17:59 +00:00
|
|
|
def load_t2i_adapter(t2i_data):
|
2023-03-03 23:58:22 +00:00
|
|
|
keys = t2i_data.keys()
|
2023-03-05 23:39:25 +00:00
|
|
|
if "body.0.in_conv.weight" in keys:
|
2023-03-03 23:58:22 +00:00
|
|
|
cin = t2i_data['body.0.in_conv.weight'].shape[1]
|
|
|
|
model_ad = adapter.Adapter_light(cin=cin, channels=[320, 640, 1280, 1280], nums_rb=4)
|
2023-03-17 22:17:59 +00:00
|
|
|
elif 'conv_in.weight' in keys:
|
2023-03-03 23:58:22 +00:00
|
|
|
cin = t2i_data['conv_in.weight'].shape[1]
|
|
|
|
model_ad = adapter.Adapter(cin=cin, channels=[320, 640, 1280, 1280][:4], nums_rb=2, ksize=1, sk=True, use_conv=False)
|
2023-03-17 22:17:59 +00:00
|
|
|
else:
|
|
|
|
return None
|
2023-02-25 05:55:42 +00:00
|
|
|
model_ad.load_state_dict(t2i_data)
|
|
|
|
return T2IAdapter(model_ad, cin // 64)
|
2023-02-16 15:38:08 +00:00
|
|
|
|
2023-03-05 23:39:25 +00:00
|
|
|
|
|
|
|
class StyleModel:
|
|
|
|
def __init__(self, model, device="cpu"):
|
|
|
|
self.model = model
|
|
|
|
|
|
|
|
def get_cond(self, input):
|
|
|
|
return self.model(input.last_hidden_state)
|
|
|
|
|
|
|
|
|
|
|
|
def load_style_model(ckpt_path):
|
2023-04-02 03:19:15 +00:00
|
|
|
model_data = utils.load_torch_file(ckpt_path)
|
2023-03-05 23:39:25 +00:00
|
|
|
keys = model_data.keys()
|
|
|
|
if "style_embedding" in keys:
|
|
|
|
model = adapter.StyleAdapter(width=1024, context_dim=768, num_head=8, n_layes=3, num_token=8)
|
|
|
|
else:
|
|
|
|
raise Exception("invalid style model {}".format(ckpt_path))
|
|
|
|
model.load_state_dict(model_data)
|
|
|
|
return StyleModel(model)
|
|
|
|
|
|
|
|
|
2023-02-05 20:20:18 +00:00
|
|
|
def load_clip(ckpt_path, embedding_directory=None):
|
2023-04-02 03:19:15 +00:00
|
|
|
clip_data = utils.load_torch_file(ckpt_path)
|
2023-02-05 20:20:18 +00:00
|
|
|
config = {}
|
|
|
|
if "text_model.encoder.layers.22.mlp.fc1.weight" in clip_data:
|
|
|
|
config['target'] = 'ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
|
|
|
|
else:
|
|
|
|
config['target'] = 'ldm.modules.encoders.modules.FrozenCLIPEmbedder'
|
|
|
|
clip = CLIP(config=config, embedding_directory=embedding_directory)
|
|
|
|
clip.load_from_state_dict(clip_data)
|
|
|
|
return clip
|
2023-01-03 06:53:32 +00:00
|
|
|
|
2023-01-29 23:46:44 +00:00
|
|
|
def load_checkpoint(config_path, ckpt_path, output_vae=True, output_clip=True, embedding_directory=None):
|
2023-03-13 18:49:18 +00:00
|
|
|
with open(config_path, 'r') as stream:
|
|
|
|
config = yaml.safe_load(stream)
|
2023-01-03 06:53:32 +00:00
|
|
|
model_config_params = config['model']['params']
|
|
|
|
clip_config = model_config_params['cond_stage_config']
|
|
|
|
scale_factor = model_config_params['scale_factor']
|
|
|
|
vae_config = model_config_params['first_stage_config']
|
|
|
|
|
2023-03-14 01:12:48 +00:00
|
|
|
fp16 = False
|
|
|
|
if "unet_config" in model_config_params:
|
|
|
|
if "params" in model_config_params["unet_config"]:
|
|
|
|
if "use_fp16" in model_config_params["unet_config"]["params"]:
|
|
|
|
fp16 = model_config_params["unet_config"]["params"]["use_fp16"]
|
|
|
|
|
2023-01-03 06:53:32 +00:00
|
|
|
clip = None
|
|
|
|
vae = None
|
|
|
|
|
|
|
|
class WeightsLoader(torch.nn.Module):
|
|
|
|
pass
|
|
|
|
|
|
|
|
w = WeightsLoader()
|
|
|
|
load_state_dict_to = []
|
|
|
|
if output_vae:
|
|
|
|
vae = VAE(scale_factor=scale_factor, config=vae_config)
|
|
|
|
w.first_stage_model = vae.first_stage_model
|
|
|
|
load_state_dict_to = [w]
|
|
|
|
|
|
|
|
if output_clip:
|
2023-01-29 23:46:44 +00:00
|
|
|
clip = CLIP(config=clip_config, embedding_directory=embedding_directory)
|
2023-01-03 06:53:32 +00:00
|
|
|
w.cond_stage_model = clip.cond_stage_model
|
|
|
|
load_state_dict_to = [w]
|
|
|
|
|
2023-03-13 18:49:18 +00:00
|
|
|
model = instantiate_from_config(config["model"])
|
2023-04-02 03:19:15 +00:00
|
|
|
sd = utils.load_torch_file(ckpt_path)
|
2023-02-26 22:04:35 +00:00
|
|
|
model = load_model_weights(model, sd, verbose=False, load_state_dict_to=load_state_dict_to)
|
2023-03-14 01:12:48 +00:00
|
|
|
|
|
|
|
if fp16:
|
|
|
|
model = model.half()
|
|
|
|
|
2023-02-03 07:06:34 +00:00
|
|
|
return (ModelPatcher(model), clip, vae)
|
2023-03-03 08:37:35 +00:00
|
|
|
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, output_clipvision=False, embedding_directory=None):
|
|
|
|
sd = utils.load_torch_file(ckpt_path)
|
2023-03-03 08:37:35 +00:00
|
|
|
sd_keys = sd.keys()
|
|
|
|
clip = None
|
2023-04-02 03:19:15 +00:00
|
|
|
clipvision = None
|
2023-03-03 08:37:35 +00:00
|
|
|
vae = None
|
|
|
|
|
2023-03-03 16:07:10 +00:00
|
|
|
fp16 = model_management.should_use_fp16()
|
|
|
|
|
2023-03-03 08:37:35 +00:00
|
|
|
class WeightsLoader(torch.nn.Module):
|
|
|
|
pass
|
|
|
|
|
|
|
|
w = WeightsLoader()
|
|
|
|
load_state_dict_to = []
|
|
|
|
if output_vae:
|
|
|
|
vae = VAE()
|
|
|
|
w.first_stage_model = vae.first_stage_model
|
|
|
|
load_state_dict_to = [w]
|
|
|
|
|
|
|
|
if output_clip:
|
|
|
|
clip_config = {}
|
|
|
|
if "cond_stage_model.model.transformer.resblocks.22.attn.out_proj.weight" in sd_keys:
|
|
|
|
clip_config['target'] = 'ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
|
|
|
|
else:
|
|
|
|
clip_config['target'] = 'ldm.modules.encoders.modules.FrozenCLIPEmbedder'
|
|
|
|
clip = CLIP(config=clip_config, embedding_directory=embedding_directory)
|
|
|
|
w.cond_stage_model = clip.cond_stage_model
|
|
|
|
load_state_dict_to = [w]
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
clipvision_key = "embedder.model.visual.transformer.resblocks.0.attn.in_proj_weight"
|
|
|
|
noise_aug_config = None
|
|
|
|
if clipvision_key in sd_keys:
|
|
|
|
size = sd[clipvision_key].shape[1]
|
|
|
|
|
|
|
|
if output_clipvision:
|
|
|
|
clipvision = clip_vision.load_clipvision_from_sd(sd)
|
|
|
|
|
|
|
|
noise_aug_key = "noise_augmentor.betas"
|
|
|
|
if noise_aug_key in sd_keys:
|
|
|
|
noise_aug_config = {}
|
|
|
|
params = {}
|
|
|
|
noise_schedule_config = {}
|
|
|
|
noise_schedule_config["timesteps"] = sd[noise_aug_key].shape[0]
|
|
|
|
noise_schedule_config["beta_schedule"] = "squaredcos_cap_v2"
|
|
|
|
params["noise_schedule_config"] = noise_schedule_config
|
|
|
|
noise_aug_config['target'] = "ldm.modules.encoders.noise_aug_modules.CLIPEmbeddingNoiseAugmentation"
|
|
|
|
if size == 1280: #h
|
|
|
|
params["timestep_dim"] = 1024
|
|
|
|
elif size == 1024: #l
|
|
|
|
params["timestep_dim"] = 768
|
|
|
|
noise_aug_config['params'] = params
|
|
|
|
|
2023-03-03 08:37:35 +00:00
|
|
|
sd_config = {
|
|
|
|
"linear_start": 0.00085,
|
|
|
|
"linear_end": 0.012,
|
|
|
|
"num_timesteps_cond": 1,
|
|
|
|
"log_every_t": 200,
|
|
|
|
"timesteps": 1000,
|
|
|
|
"first_stage_key": "jpg",
|
|
|
|
"cond_stage_key": "txt",
|
|
|
|
"image_size": 64,
|
|
|
|
"channels": 4,
|
|
|
|
"cond_stage_trainable": False,
|
|
|
|
"monitor": "val/loss_simple_ema",
|
|
|
|
"scale_factor": 0.18215,
|
|
|
|
"use_ema": False,
|
|
|
|
}
|
|
|
|
|
|
|
|
unet_config = {
|
|
|
|
"use_checkpoint": True,
|
|
|
|
"image_size": 32,
|
|
|
|
"out_channels": 4,
|
|
|
|
"attention_resolutions": [
|
|
|
|
4,
|
|
|
|
2,
|
|
|
|
1
|
|
|
|
],
|
|
|
|
"num_res_blocks": 2,
|
|
|
|
"channel_mult": [
|
|
|
|
1,
|
|
|
|
2,
|
|
|
|
4,
|
|
|
|
4
|
|
|
|
],
|
|
|
|
"use_spatial_transformer": True,
|
|
|
|
"transformer_depth": 1,
|
|
|
|
"legacy": False
|
|
|
|
}
|
|
|
|
|
|
|
|
if len(sd['model.diffusion_model.input_blocks.1.1.proj_in.weight'].shape) == 2:
|
|
|
|
unet_config['use_linear_in_transformer'] = True
|
|
|
|
|
|
|
|
unet_config["use_fp16"] = fp16
|
|
|
|
unet_config["model_channels"] = sd['model.diffusion_model.input_blocks.0.0.weight'].shape[0]
|
|
|
|
unet_config["in_channels"] = sd['model.diffusion_model.input_blocks.0.0.weight'].shape[1]
|
|
|
|
unet_config["context_dim"] = sd['model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight'].shape[1]
|
|
|
|
|
|
|
|
sd_config["unet_config"] = {"target": "ldm.modules.diffusionmodules.openaimodel.UNetModel", "params": unet_config}
|
|
|
|
model_config = {"target": "ldm.models.diffusion.ddpm.LatentDiffusion", "params": sd_config}
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
if noise_aug_config is not None: #SD2.x unclip model
|
|
|
|
sd_config["noise_aug_config"] = noise_aug_config
|
|
|
|
sd_config["image_size"] = 96
|
|
|
|
sd_config["embedding_dropout"] = 0.25
|
|
|
|
sd_config["conditioning_key"] = 'crossattn-adm'
|
|
|
|
model_config["target"] = "ldm.models.diffusion.ddpm.ImageEmbeddingConditionedLatentDiffusion"
|
|
|
|
elif unet_config["in_channels"] > 4: #inpainting model
|
2023-03-03 08:37:35 +00:00
|
|
|
sd_config["conditioning_key"] = "hybrid"
|
|
|
|
sd_config["finetune_keys"] = None
|
|
|
|
model_config["target"] = "ldm.models.diffusion.ddpm.LatentInpaintDiffusion"
|
|
|
|
else:
|
|
|
|
sd_config["conditioning_key"] = "crossattn"
|
|
|
|
|
|
|
|
if unet_config["context_dim"] == 1024:
|
|
|
|
unet_config["num_head_channels"] = 64 #SD2.x
|
|
|
|
else:
|
|
|
|
unet_config["num_heads"] = 8 #SD1.x
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
unclip = 'model.diffusion_model.label_emb.0.0.weight'
|
|
|
|
if unclip in sd_keys:
|
|
|
|
unet_config["num_classes"] = "sequential"
|
|
|
|
unet_config["adm_in_channels"] = sd[unclip].shape[1]
|
|
|
|
|
2023-03-03 18:04:36 +00:00
|
|
|
if unet_config["context_dim"] == 1024 and unet_config["in_channels"] == 4: #only SD2.x non inpainting models are v prediction
|
|
|
|
k = "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm1.bias"
|
|
|
|
out = sd[k]
|
|
|
|
if torch.std(out, unbiased=False) > 0.09: # not sure how well this will actually work. I guess we will find out.
|
|
|
|
sd_config["parameterization"] = 'v'
|
2023-03-03 08:37:35 +00:00
|
|
|
|
|
|
|
model = instantiate_from_config(model_config)
|
|
|
|
model = load_model_weights(model, sd, verbose=False, load_state_dict_to=load_state_dict_to)
|
|
|
|
|
2023-03-14 00:58:09 +00:00
|
|
|
if fp16:
|
|
|
|
model = model.half()
|
|
|
|
|
2023-04-02 03:19:15 +00:00
|
|
|
return (ModelPatcher(model), clip, vae, clipvision)
|