Merge remote-tracking branch 'upstream/main' into hf_model_download
This commit is contained in:
commit
4e6d8f1157
|
@ -41,7 +41,8 @@ Resuming from a checkpoint, 50 epochs, 6 batch size, 3e-6 learning rate, cosine
|
|||
--sample_steps 200 ^
|
||||
--lr 3e-6 ^
|
||||
--ckpt_every_n_minutes 10 ^
|
||||
--useadam8bit
|
||||
--useadam8bit ^
|
||||
--ed1_mode
|
||||
|
||||
Training from SD2 512 base model, 18 epochs, 4 batch size, 1.2e-6 learning rate, constant LR, generate samples evern 100 steps, 30 minute checkpoint interval, adam8bit, using imagesin the x:\mydata folder, training at resolution class of 640:
|
||||
|
||||
|
|
36
train.py
36
train.py
|
@ -46,13 +46,12 @@ from accelerate.utils import set_seed
|
|||
|
||||
import wandb
|
||||
from torch.utils.tensorboard import SummaryWriter
|
||||
from tqdm.auto import tqdm
|
||||
|
||||
import keyboard
|
||||
|
||||
from data.every_dream import EveryDreamBatch
|
||||
from utils.huggingface_downloader import try_download_model_from_hf
|
||||
from utils.convert_diffusers_to_stable_diffusion import convert as converter
|
||||
from utils.convert_diff_to_ckpt import convert as converter
|
||||
from utils.gpu import GPU
|
||||
|
||||
|
||||
|
@ -281,7 +280,12 @@ def main(args):
|
|||
"""
|
||||
log_time = setup_local_logger(args)
|
||||
args = setup_args(args)
|
||||
|
||||
|
||||
if args.notebook:
|
||||
from tqdm.notebook import tqdm
|
||||
else:
|
||||
from tqdm.auto import tqdm
|
||||
|
||||
seed = args.seed if args.seed != -1 else random.randint(0, 2**30)
|
||||
set_seed(seed)
|
||||
gpu = GPU()
|
||||
|
@ -295,7 +299,7 @@ def main(args):
|
|||
os.makedirs(log_folder)
|
||||
|
||||
@torch.no_grad()
|
||||
def __save_model(save_path, unet, text_encoder, tokenizer, scheduler, vae, save_ckpt_dir):
|
||||
def __save_model(save_path, unet, text_encoder, tokenizer, scheduler, vae, save_ckpt_dir, save_full_precision=False):
|
||||
"""
|
||||
Save the model to disk
|
||||
"""
|
||||
|
@ -320,9 +324,11 @@ def main(args):
|
|||
sd_ckpt_full = os.path.join(save_ckpt_dir, sd_ckpt_path)
|
||||
else:
|
||||
sd_ckpt_full = os.path.join(os.curdir, sd_ckpt_path)
|
||||
|
||||
half = not save_full_precision
|
||||
|
||||
logging.info(f" * Saving SD model to {sd_ckpt_full}")
|
||||
converter(model_path=save_path, checkpoint_path=sd_ckpt_full, half=True)
|
||||
converter(model_path=save_path, checkpoint_path=sd_ckpt_full, half=half)
|
||||
# optimizer_path = os.path.join(save_path, "optimizer.pt")
|
||||
|
||||
# if self.save_optimizer_flag:
|
||||
|
@ -585,7 +591,7 @@ def main(args):
|
|||
logging.error(f"{Fore.LIGHTRED_EX} CTRL-C received, attempting to save model to {interrupted_checkpoint_path}{Style.RESET_ALL}")
|
||||
logging.error(f"{Fore.LIGHTRED_EX} ************************************************************************{Style.RESET_ALL}")
|
||||
time.sleep(2) # give opportunity to ctrl-C again to cancel save
|
||||
__save_model(interrupted_checkpoint_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir)
|
||||
__save_model(interrupted_checkpoint_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir, args.save_full_precision)
|
||||
exit(_SIGTERM_EXIT_CODE)
|
||||
|
||||
signal.signal(signal.SIGINT, sigterm_handler)
|
||||
|
@ -781,7 +787,7 @@ def main(args):
|
|||
append_epoch_log(global_step=global_step, epoch_pbar=epoch_pbar, gpu=gpu, log_writer=log_writer, **logs)
|
||||
torch.cuda.empty_cache()
|
||||
|
||||
if keyboard.is_pressed("ctrl+alt+page up") or ((global_step + 1) % args.sample_steps == 0):
|
||||
if (not args.notebook and keyboard.is_pressed("ctrl+alt+page up")) or ((global_step + 1) % args.sample_steps == 0):
|
||||
pipe = __create_inference_pipe(unet=unet, text_encoder=text_encoder, tokenizer=tokenizer, scheduler=sample_scheduler, vae=vae)
|
||||
pipe = pipe.to(device)
|
||||
|
||||
|
@ -803,12 +809,12 @@ def main(args):
|
|||
last_epoch_saved_time = time.time()
|
||||
logging.info(f"Saving model, {args.ckpt_every_n_minutes} mins at step {global_step}")
|
||||
save_path = os.path.join(f"{log_folder}/ckpts/{args.project_name}-ep{epoch:02}-gs{global_step:05}")
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir)
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir, args.save_full_precision)
|
||||
|
||||
if epoch > 0 and epoch % args.save_every_n_epochs == 0 and step == 1 and epoch < args.max_epochs - 1:
|
||||
logging.info(f" Saving model, {args.save_every_n_epochs} epochs at step {global_step}")
|
||||
save_path = os.path.join(f"{log_folder}/ckpts/{args.project_name}-ep{epoch:02}-gs{global_step:05}")
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir)
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir, args.save_full_precision)
|
||||
|
||||
del batch
|
||||
global_step += 1
|
||||
|
@ -829,7 +835,7 @@ def main(args):
|
|||
# end of training
|
||||
|
||||
save_path = os.path.join(f"{log_folder}/ckpts/last-{args.project_name}-ep{epoch:02}-gs{global_step:05}")
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir)
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir, args.save_full_precision)
|
||||
|
||||
total_elapsed_time = time.time() - training_start_time
|
||||
logging.info(f"{Fore.CYAN}Training complete{Style.RESET_ALL}")
|
||||
|
@ -839,7 +845,7 @@ def main(args):
|
|||
except Exception as ex:
|
||||
logging.error(f"{Fore.LIGHTYELLOW_EX}Something went wrong, attempting to save model{Style.RESET_ALL}")
|
||||
save_path = os.path.join(f"{log_folder}/ckpts/errored-{args.project_name}-ep{epoch:02}-gs{global_step:05}")
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir)
|
||||
__save_model(save_path, unet, text_encoder, tokenizer, noise_scheduler, vae, args.save_ckpt_dir, args.save_full_precision)
|
||||
raise ex
|
||||
|
||||
logging.info(f"{Fore.LIGHTWHITE_EX} ***************************{Style.RESET_ALL}")
|
||||
|
@ -854,6 +860,12 @@ def update_old_args(t_args):
|
|||
if not hasattr(t_args, "shuffle_tags"):
|
||||
print(f" Config json is missing 'shuffle_tags'")
|
||||
t_args.__dict__["shuffle_tags"] = False
|
||||
if not hasattr(t_args, "save_full_precision"):
|
||||
print(f" Config json is missing 'save_full_precision'")
|
||||
t_args.__dict__["save_full_precision"] = False
|
||||
if not hasattr(t_args, "notebook"):
|
||||
print(f" Config json is missing 'notebook'")
|
||||
t_args.__dict__["notebook"] = False
|
||||
|
||||
if __name__ == "__main__":
|
||||
supported_resolutions = [256, 384, 448, 512, 576, 640, 704, 768, 832, 896, 960, 1024, 1088, 1152]
|
||||
|
@ -909,6 +921,8 @@ if __name__ == "__main__":
|
|||
argparser.add_argument("--useadam8bit", action="store_true", default=False, help="Use AdamW 8-Bit optimizer, recommended!")
|
||||
argparser.add_argument("--wandb", action="store_true", default=False, help="enable wandb logging instead of tensorboard, requires env var WANDB_API_KEY")
|
||||
argparser.add_argument("--write_schedule", action="store_true", default=False, help="write schedule of images and their batches to file (def: False)")
|
||||
argparser.add_argument("--save_full_precision", action="store_true", default=False, help="save ckpts at full FP32")
|
||||
argparser.add_argument("--notebook", action="store_true", default=False, help="disable keypresses and uses tqdm.notebook for jupyter notebook (def: False)")
|
||||
|
||||
args = argparser.parse_args()
|
||||
|
||||
|
|
|
@ -0,0 +1,321 @@
|
|||
# from https://github.com/huggingface/diffusers/blob/main/scripts/convert_diffusers_to_original_stable_diffusion.py
|
||||
# modified to be callable
|
||||
|
||||
# coding=utf-8
|
||||
# Copyright 2022 The HuggingFace Inc. team.
|
||||
# Modifications for EveryDream Copyright 2022 Victor C Hall
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
# Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint.
|
||||
# *Only* converts the UNet, VAE, and Text Encoder.
|
||||
# Does not convert optimizer state or any other thing.
|
||||
|
||||
import os.path as osp
|
||||
import re
|
||||
|
||||
import torch
|
||||
|
||||
|
||||
# =================#
|
||||
# UNet Conversion #
|
||||
# =================#
|
||||
|
||||
unet_conversion_map = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("time_embed.0.weight", "time_embedding.linear_1.weight"),
|
||||
("time_embed.0.bias", "time_embedding.linear_1.bias"),
|
||||
("time_embed.2.weight", "time_embedding.linear_2.weight"),
|
||||
("time_embed.2.bias", "time_embedding.linear_2.bias"),
|
||||
("input_blocks.0.0.weight", "conv_in.weight"),
|
||||
("input_blocks.0.0.bias", "conv_in.bias"),
|
||||
("out.0.weight", "conv_norm_out.weight"),
|
||||
("out.0.bias", "conv_norm_out.bias"),
|
||||
("out.2.weight", "conv_out.weight"),
|
||||
("out.2.bias", "conv_out.bias"),
|
||||
]
|
||||
|
||||
unet_conversion_map_resnet = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("in_layers.0", "norm1"),
|
||||
("in_layers.2", "conv1"),
|
||||
("out_layers.0", "norm2"),
|
||||
("out_layers.3", "conv2"),
|
||||
("emb_layers.1", "time_emb_proj"),
|
||||
("skip_connection", "conv_shortcut"),
|
||||
]
|
||||
|
||||
unet_conversion_map_layer = []
|
||||
# hardcoded number of downblocks and resnets/attentions...
|
||||
# would need smarter logic for other networks.
|
||||
for i in range(4):
|
||||
# loop over downblocks/upblocks
|
||||
|
||||
for j in range(2):
|
||||
# loop over resnets/attentions for downblocks
|
||||
hf_down_res_prefix = f"down_blocks.{i}.resnets.{j}."
|
||||
sd_down_res_prefix = f"input_blocks.{3*i + j + 1}.0."
|
||||
unet_conversion_map_layer.append((sd_down_res_prefix, hf_down_res_prefix))
|
||||
|
||||
if i < 3:
|
||||
# no attention layers in down_blocks.3
|
||||
hf_down_atn_prefix = f"down_blocks.{i}.attentions.{j}."
|
||||
sd_down_atn_prefix = f"input_blocks.{3*i + j + 1}.1."
|
||||
unet_conversion_map_layer.append((sd_down_atn_prefix, hf_down_atn_prefix))
|
||||
|
||||
for j in range(3):
|
||||
# loop over resnets/attentions for upblocks
|
||||
hf_up_res_prefix = f"up_blocks.{i}.resnets.{j}."
|
||||
sd_up_res_prefix = f"output_blocks.{3*i + j}.0."
|
||||
unet_conversion_map_layer.append((sd_up_res_prefix, hf_up_res_prefix))
|
||||
|
||||
if i > 0:
|
||||
# no attention layers in up_blocks.0
|
||||
hf_up_atn_prefix = f"up_blocks.{i}.attentions.{j}."
|
||||
sd_up_atn_prefix = f"output_blocks.{3*i + j}.1."
|
||||
unet_conversion_map_layer.append((sd_up_atn_prefix, hf_up_atn_prefix))
|
||||
|
||||
if i < 3:
|
||||
# no downsample in down_blocks.3
|
||||
hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0.conv."
|
||||
sd_downsample_prefix = f"input_blocks.{3*(i+1)}.0.op."
|
||||
unet_conversion_map_layer.append((sd_downsample_prefix, hf_downsample_prefix))
|
||||
|
||||
# no upsample in up_blocks.3
|
||||
hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
|
||||
sd_upsample_prefix = f"output_blocks.{3*i + 2}.{1 if i == 0 else 2}."
|
||||
unet_conversion_map_layer.append((sd_upsample_prefix, hf_upsample_prefix))
|
||||
|
||||
hf_mid_atn_prefix = "mid_block.attentions.0."
|
||||
sd_mid_atn_prefix = "middle_block.1."
|
||||
unet_conversion_map_layer.append((sd_mid_atn_prefix, hf_mid_atn_prefix))
|
||||
|
||||
for j in range(2):
|
||||
hf_mid_res_prefix = f"mid_block.resnets.{j}."
|
||||
sd_mid_res_prefix = f"middle_block.{2*j}."
|
||||
unet_conversion_map_layer.append((sd_mid_res_prefix, hf_mid_res_prefix))
|
||||
|
||||
|
||||
def convert_unet_state_dict(unet_state_dict):
|
||||
# buyer beware: this is a *brittle* function,
|
||||
# and correct output requires that all of these pieces interact in
|
||||
# the exact order in which I have arranged them.
|
||||
mapping = {k: k for k in unet_state_dict.keys()}
|
||||
for sd_name, hf_name in unet_conversion_map:
|
||||
mapping[hf_name] = sd_name
|
||||
for k, v in mapping.items():
|
||||
if "resnets" in k:
|
||||
for sd_part, hf_part in unet_conversion_map_resnet:
|
||||
v = v.replace(hf_part, sd_part)
|
||||
mapping[k] = v
|
||||
for k, v in mapping.items():
|
||||
for sd_part, hf_part in unet_conversion_map_layer:
|
||||
v = v.replace(hf_part, sd_part)
|
||||
mapping[k] = v
|
||||
new_state_dict = {v: unet_state_dict[k] for k, v in mapping.items()}
|
||||
return new_state_dict
|
||||
|
||||
|
||||
# ================#
|
||||
# VAE Conversion #
|
||||
# ================#
|
||||
|
||||
vae_conversion_map = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("nin_shortcut", "conv_shortcut"),
|
||||
("norm_out", "conv_norm_out"),
|
||||
("mid.attn_1.", "mid_block.attentions.0."),
|
||||
]
|
||||
|
||||
for i in range(4):
|
||||
# down_blocks have two resnets
|
||||
for j in range(2):
|
||||
hf_down_prefix = f"encoder.down_blocks.{i}.resnets.{j}."
|
||||
sd_down_prefix = f"encoder.down.{i}.block.{j}."
|
||||
vae_conversion_map.append((sd_down_prefix, hf_down_prefix))
|
||||
|
||||
if i < 3:
|
||||
hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0."
|
||||
sd_downsample_prefix = f"down.{i}.downsample."
|
||||
vae_conversion_map.append((sd_downsample_prefix, hf_downsample_prefix))
|
||||
|
||||
hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
|
||||
sd_upsample_prefix = f"up.{3-i}.upsample."
|
||||
vae_conversion_map.append((sd_upsample_prefix, hf_upsample_prefix))
|
||||
|
||||
# up_blocks have three resnets
|
||||
# also, up blocks in hf are numbered in reverse from sd
|
||||
for j in range(3):
|
||||
hf_up_prefix = f"decoder.up_blocks.{i}.resnets.{j}."
|
||||
sd_up_prefix = f"decoder.up.{3-i}.block.{j}."
|
||||
vae_conversion_map.append((sd_up_prefix, hf_up_prefix))
|
||||
|
||||
# this part accounts for mid blocks in both the encoder and the decoder
|
||||
for i in range(2):
|
||||
hf_mid_res_prefix = f"mid_block.resnets.{i}."
|
||||
sd_mid_res_prefix = f"mid.block_{i+1}."
|
||||
vae_conversion_map.append((sd_mid_res_prefix, hf_mid_res_prefix))
|
||||
|
||||
|
||||
vae_conversion_map_attn = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("norm.", "group_norm."),
|
||||
("q.", "query."),
|
||||
("k.", "key."),
|
||||
("v.", "value."),
|
||||
("proj_out.", "proj_attn."),
|
||||
]
|
||||
|
||||
|
||||
def reshape_weight_for_sd(w):
|
||||
# convert HF linear weights to SD conv2d weights
|
||||
return w.reshape(*w.shape, 1, 1)
|
||||
|
||||
|
||||
def convert_vae_state_dict(vae_state_dict):
|
||||
mapping = {k: k for k in vae_state_dict.keys()}
|
||||
for k, v in mapping.items():
|
||||
for sd_part, hf_part in vae_conversion_map:
|
||||
v = v.replace(hf_part, sd_part)
|
||||
mapping[k] = v
|
||||
for k, v in mapping.items():
|
||||
if "attentions" in k:
|
||||
for sd_part, hf_part in vae_conversion_map_attn:
|
||||
v = v.replace(hf_part, sd_part)
|
||||
mapping[k] = v
|
||||
new_state_dict = {v: vae_state_dict[k] for k, v in mapping.items()}
|
||||
weights_to_convert = ["q", "k", "v", "proj_out"]
|
||||
for k, v in new_state_dict.items():
|
||||
for weight_name in weights_to_convert:
|
||||
if f"mid.attn_1.{weight_name}.weight" in k:
|
||||
#print(f"Reshaping {k} for SD format")
|
||||
new_state_dict[k] = reshape_weight_for_sd(v)
|
||||
return new_state_dict
|
||||
|
||||
|
||||
# =========================#
|
||||
# Text Encoder Conversion #
|
||||
# =========================#
|
||||
|
||||
|
||||
textenc_conversion_lst = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("resblocks.", "text_model.encoder.layers."),
|
||||
("ln_1", "layer_norm1"),
|
||||
("ln_2", "layer_norm2"),
|
||||
(".c_fc.", ".fc1."),
|
||||
(".c_proj.", ".fc2."),
|
||||
(".attn", ".self_attn"),
|
||||
("ln_final.", "transformer.text_model.final_layer_norm."),
|
||||
("token_embedding.weight", "transformer.text_model.embeddings.token_embedding.weight"),
|
||||
("positional_embedding", "transformer.text_model.embeddings.position_embedding.weight"),
|
||||
]
|
||||
protected = {re.escape(x[1]): x[0] for x in textenc_conversion_lst}
|
||||
textenc_pattern = re.compile("|".join(protected.keys()))
|
||||
|
||||
# Ordering is from https://github.com/pytorch/pytorch/blob/master/test/cpp/api/modules.cpp
|
||||
code2idx = {"q": 0, "k": 1, "v": 2}
|
||||
|
||||
|
||||
def convert_text_enc_state_dict_v20(text_enc_dict):
|
||||
new_state_dict = {}
|
||||
capture_qkv_weight = {}
|
||||
capture_qkv_bias = {}
|
||||
for k, v in text_enc_dict.items():
|
||||
if (
|
||||
k.endswith(".self_attn.q_proj.weight")
|
||||
or k.endswith(".self_attn.k_proj.weight")
|
||||
or k.endswith(".self_attn.v_proj.weight")
|
||||
):
|
||||
k_pre = k[: -len(".q_proj.weight")]
|
||||
k_code = k[-len("q_proj.weight")]
|
||||
if k_pre not in capture_qkv_weight:
|
||||
capture_qkv_weight[k_pre] = [None, None, None]
|
||||
capture_qkv_weight[k_pre][code2idx[k_code]] = v
|
||||
continue
|
||||
|
||||
if (
|
||||
k.endswith(".self_attn.q_proj.bias")
|
||||
or k.endswith(".self_attn.k_proj.bias")
|
||||
or k.endswith(".self_attn.v_proj.bias")
|
||||
):
|
||||
k_pre = k[: -len(".q_proj.bias")]
|
||||
k_code = k[-len("q_proj.bias")]
|
||||
if k_pre not in capture_qkv_bias:
|
||||
capture_qkv_bias[k_pre] = [None, None, None]
|
||||
capture_qkv_bias[k_pre][code2idx[k_code]] = v
|
||||
continue
|
||||
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k)
|
||||
new_state_dict[relabelled_key] = v
|
||||
|
||||
for k_pre, tensors in capture_qkv_weight.items():
|
||||
if None in tensors:
|
||||
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
|
||||
new_state_dict[relabelled_key + ".in_proj_weight"] = torch.cat(tensors)
|
||||
|
||||
for k_pre, tensors in capture_qkv_bias.items():
|
||||
if None in tensors:
|
||||
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
|
||||
new_state_dict[relabelled_key + ".in_proj_bias"] = torch.cat(tensors)
|
||||
|
||||
return new_state_dict
|
||||
|
||||
def convert_text_enc_state_dict(text_enc_dict):
|
||||
return text_enc_dict
|
||||
|
||||
|
||||
def convert(model_path: str, checkpoint_path: str, half: bool):
|
||||
|
||||
assert model_path is not None, "Must provide a model path!"
|
||||
|
||||
assert checkpoint_path is not None, "Must provide a checkpoint path!"
|
||||
|
||||
unet_path = osp.join(model_path, "unet", "diffusion_pytorch_model.bin")
|
||||
vae_path = osp.join(model_path, "vae", "diffusion_pytorch_model.bin")
|
||||
text_enc_path = osp.join(model_path, "text_encoder", "pytorch_model.bin")
|
||||
|
||||
# Convert the UNet model
|
||||
unet_state_dict = torch.load(unet_path, map_location="cpu")
|
||||
unet_state_dict = convert_unet_state_dict(unet_state_dict)
|
||||
unet_state_dict = {"model.diffusion_model." + k: v for k, v in unet_state_dict.items()}
|
||||
|
||||
# Convert the VAE model
|
||||
vae_state_dict = torch.load(vae_path, map_location="cpu")
|
||||
vae_state_dict = convert_vae_state_dict(vae_state_dict)
|
||||
vae_state_dict = {"first_stage_model." + k: v for k, v in vae_state_dict.items()}
|
||||
|
||||
# Convert the text encoder model
|
||||
text_enc_dict = torch.load(text_enc_path, map_location="cpu")
|
||||
|
||||
# Easiest way to identify v2.0 model seems to be that the text encoder (OpenCLIP) is deeper
|
||||
is_v20_model = "text_model.encoder.layers.22.layer_norm2.bias" in text_enc_dict
|
||||
|
||||
if is_v20_model:
|
||||
# Need to add the tag 'transformer' in advance so we can knock it out from the final layer-norm
|
||||
text_enc_dict = {"transformer." + k: v for k, v in text_enc_dict.items()}
|
||||
text_enc_dict = convert_text_enc_state_dict_v20(text_enc_dict)
|
||||
text_enc_dict = {"cond_stage_model.model." + k: v for k, v in text_enc_dict.items()}
|
||||
else:
|
||||
text_enc_dict = convert_text_enc_state_dict(text_enc_dict)
|
||||
text_enc_dict = {"cond_stage_model.transformer." + k: v for k, v in text_enc_dict.items()}
|
||||
|
||||
# Put together new checkpoint
|
||||
state_dict = {**unet_state_dict, **vae_state_dict, **text_enc_dict}
|
||||
if half:
|
||||
state_dict = {k: v.half() for k, v in state_dict.items()}
|
||||
state_dict = {"state_dict": state_dict}
|
||||
torch.save(state_dict, checkpoint_path)
|
||||
|
|
@ -19,10 +19,11 @@
|
|||
|
||||
# Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint.
|
||||
# *Only* converts the UNet, VAE, and Text Encoder.
|
||||
# Does not convert optimizer state or any other thing.
|
||||
|
||||
# Does not convert optimizer state or any other thing.
|
||||
|
||||
import argparse
|
||||
import os.path as osp
|
||||
import re
|
||||
|
||||
import torch
|
||||
|
||||
|
@ -198,7 +199,7 @@ def convert_vae_state_dict(vae_state_dict):
|
|||
for k, v in new_state_dict.items():
|
||||
for weight_name in weights_to_convert:
|
||||
if f"mid.attn_1.{weight_name}.weight" in k:
|
||||
#print(f"Reshaping {k} for SD format")
|
||||
print(f"Reshaping {k} for SD format")
|
||||
new_state_dict[k] = reshape_weight_for_sd(v)
|
||||
return new_state_dict
|
||||
|
||||
|
@ -206,21 +207,95 @@ def convert_vae_state_dict(vae_state_dict):
|
|||
# =========================#
|
||||
# Text Encoder Conversion #
|
||||
# =========================#
|
||||
# pretty much a no-op
|
||||
|
||||
|
||||
textenc_conversion_lst = [
|
||||
# (stable-diffusion, HF Diffusers)
|
||||
("resblocks.", "text_model.encoder.layers."),
|
||||
("ln_1", "layer_norm1"),
|
||||
("ln_2", "layer_norm2"),
|
||||
(".c_fc.", ".fc1."),
|
||||
(".c_proj.", ".fc2."),
|
||||
(".attn", ".self_attn"),
|
||||
("ln_final.", "transformer.text_model.final_layer_norm."),
|
||||
("token_embedding.weight", "transformer.text_model.embeddings.token_embedding.weight"),
|
||||
("positional_embedding", "transformer.text_model.embeddings.position_embedding.weight"),
|
||||
]
|
||||
protected = {re.escape(x[1]): x[0] for x in textenc_conversion_lst}
|
||||
textenc_pattern = re.compile("|".join(protected.keys()))
|
||||
|
||||
# Ordering is from https://github.com/pytorch/pytorch/blob/master/test/cpp/api/modules.cpp
|
||||
code2idx = {"q": 0, "k": 1, "v": 2}
|
||||
|
||||
|
||||
def convert_text_enc_state_dict_v20(text_enc_dict):
|
||||
new_state_dict = {}
|
||||
capture_qkv_weight = {}
|
||||
capture_qkv_bias = {}
|
||||
for k, v in text_enc_dict.items():
|
||||
if (
|
||||
k.endswith(".self_attn.q_proj.weight")
|
||||
or k.endswith(".self_attn.k_proj.weight")
|
||||
or k.endswith(".self_attn.v_proj.weight")
|
||||
):
|
||||
k_pre = k[: -len(".q_proj.weight")]
|
||||
k_code = k[-len("q_proj.weight")]
|
||||
if k_pre not in capture_qkv_weight:
|
||||
capture_qkv_weight[k_pre] = [None, None, None]
|
||||
capture_qkv_weight[k_pre][code2idx[k_code]] = v
|
||||
continue
|
||||
|
||||
if (
|
||||
k.endswith(".self_attn.q_proj.bias")
|
||||
or k.endswith(".self_attn.k_proj.bias")
|
||||
or k.endswith(".self_attn.v_proj.bias")
|
||||
):
|
||||
k_pre = k[: -len(".q_proj.bias")]
|
||||
k_code = k[-len("q_proj.bias")]
|
||||
if k_pre not in capture_qkv_bias:
|
||||
capture_qkv_bias[k_pre] = [None, None, None]
|
||||
capture_qkv_bias[k_pre][code2idx[k_code]] = v
|
||||
continue
|
||||
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k)
|
||||
new_state_dict[relabelled_key] = v
|
||||
|
||||
for k_pre, tensors in capture_qkv_weight.items():
|
||||
if None in tensors:
|
||||
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
|
||||
new_state_dict[relabelled_key + ".in_proj_weight"] = torch.cat(tensors)
|
||||
|
||||
for k_pre, tensors in capture_qkv_bias.items():
|
||||
if None in tensors:
|
||||
raise Exception("CORRUPTED MODEL: one of the q-k-v values for the text encoder was missing")
|
||||
relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k_pre)
|
||||
new_state_dict[relabelled_key + ".in_proj_bias"] = torch.cat(tensors)
|
||||
|
||||
return new_state_dict
|
||||
|
||||
def convert_text_enc_state_dict(text_enc_dict):
|
||||
return text_enc_dict
|
||||
|
||||
def convert(model_path: str, checkpoint_path: str, half: bool):
|
||||
|
||||
assert model_path is not None, "Must provide a model path!"
|
||||
|
||||
assert checkpoint_path is not None, "Must provide a checkpoint path!"
|
||||
if __name__ == "__main__":
|
||||
parser = argparse.ArgumentParser()
|
||||
|
||||
unet_path = osp.join(model_path, "unet", "diffusion_pytorch_model.bin")
|
||||
vae_path = osp.join(model_path, "vae", "diffusion_pytorch_model.bin")
|
||||
text_enc_path = osp.join(model_path, "text_encoder", "pytorch_model.bin")
|
||||
parser.add_argument("--model_path", default=None, type=str, required=True, help="Path to the model to convert.")
|
||||
parser.add_argument("--checkpoint_path", default=None, type=str, required=True, help="Path to the output model.")
|
||||
parser.add_argument("--half", action="store_true", help="Save weights in half precision.")
|
||||
|
||||
args = parser.parse_args()
|
||||
|
||||
assert args.model_path is not None, "Must provide a model path!"
|
||||
|
||||
assert args.checkpoint_path is not None, "Must provide a checkpoint path!"
|
||||
|
||||
|
||||
|
||||
unet_path = osp.join(args.model_path, "unet", "diffusion_pytorch_model.bin")
|
||||
vae_path = osp.join(args.model_path, "vae", "diffusion_pytorch_model.bin")
|
||||
text_enc_path = osp.join(args.model_path, "text_encoder", "pytorch_model.bin")
|
||||
|
||||
# Convert the UNet model
|
||||
unet_state_dict = torch.load(unet_path, map_location="cpu")
|
||||
|
@ -234,23 +309,22 @@ def convert(model_path: str, checkpoint_path: str, half: bool):
|
|||
|
||||
# Convert the text encoder model
|
||||
text_enc_dict = torch.load(text_enc_path, map_location="cpu")
|
||||
text_enc_dict = convert_text_enc_state_dict(text_enc_dict)
|
||||
text_enc_dict = {"cond_stage_model.transformer." + k: v for k, v in text_enc_dict.items()}
|
||||
|
||||
# Easiest way to identify v2.0 model seems to be that the text encoder (OpenCLIP) is deeper
|
||||
is_v20_model = "text_model.encoder.layers.22.layer_norm2.bias" in text_enc_dict
|
||||
|
||||
if is_v20_model:
|
||||
# Need to add the tag 'transformer' in advance so we can knock it out from the final layer-norm
|
||||
text_enc_dict = {"transformer." + k: v for k, v in text_enc_dict.items()}
|
||||
text_enc_dict = convert_text_enc_state_dict_v20(text_enc_dict)
|
||||
text_enc_dict = {"cond_stage_model.model." + k: v for k, v in text_enc_dict.items()}
|
||||
else:
|
||||
text_enc_dict = convert_text_enc_state_dict(text_enc_dict)
|
||||
text_enc_dict = {"cond_stage_model.transformer." + k: v for k, v in text_enc_dict.items()}
|
||||
|
||||
# Put together new checkpoint
|
||||
state_dict = {**unet_state_dict, **vae_state_dict, **text_enc_dict}
|
||||
if half:
|
||||
if args.half:
|
||||
state_dict = {k: v.half() for k, v in state_dict.items()}
|
||||
state_dict = {"state_dict": state_dict}
|
||||
torch.save(state_dict, checkpoint_path)
|
||||
print(" * Saved converted checkpoint to", checkpoint_path)
|
||||
|
||||
# if __name__ == "__main__":
|
||||
# parser = argparse.ArgumentParser()
|
||||
|
||||
# parser.add_argument("--model_path", default=None, type=str, required=True, help="Path to the model to convert.")
|
||||
# parser.add_argument("--checkpoint_path", default=None, type=str, required=True, help="Path to the output model.")
|
||||
# parser.add_argument("--half", action="store_true", help="Save weights in half precision.")
|
||||
|
||||
# args = parser.parse_args()
|
||||
# __convert(args)
|
||||
torch.save(state_dict, args.checkpoint_path)
|
||||
|
|
Loading…
Reference in New Issue