stable-diffusion-webui/modules/sd_vae.py

244 lines
8.4 KiB
Python
Raw Normal View History

2022-10-30 08:54:31 -06:00
import torch
import safetensors.torch
2022-10-30 08:54:31 -06:00
import os
2022-12-25 05:49:25 -07:00
import collections
2022-10-30 08:54:31 -06:00
from collections import namedtuple
from modules import shared, devices, script_callbacks, sd_models
2022-10-30 08:54:31 -06:00
from modules.paths import models_path
import glob
2022-11-17 04:04:10 -07:00
from copy import deepcopy
2022-10-30 08:54:31 -06:00
2022-10-30 08:54:31 -06:00
model_dir = "Stable-diffusion"
model_path = os.path.abspath(os.path.join(models_path, model_dir))
vae_dir = "VAE"
vae_path = os.path.abspath(os.path.join(models_path, vae_dir))
2022-10-30 08:54:31 -06:00
vae_ignore_keys = {"model_ema.decay", "model_ema.num_updates"}
default_vae_dict = {"auto": "auto", "None": None, None: None}
2022-10-30 08:54:31 -06:00
default_vae_list = ["auto", "None"]
2022-10-30 08:54:31 -06:00
default_vae_values = [default_vae_dict[x] for x in default_vae_list]
vae_dict = dict(default_vae_dict)
vae_list = list(default_vae_list)
first_load = True
base_vae = None
loaded_vae_file = None
checkpoint_info = None
2022-12-25 05:49:25 -07:00
checkpoints_loaded = collections.OrderedDict()
def get_base_vae(model):
if base_vae is not None and checkpoint_info == model.sd_checkpoint_info and model:
return base_vae
return None
def store_base_vae(model):
global base_vae, checkpoint_info
if checkpoint_info != model.sd_checkpoint_info:
assert not loaded_vae_file, "Trying to store non-base VAE!"
2022-11-17 04:04:10 -07:00
base_vae = deepcopy(model.first_stage_model.state_dict())
checkpoint_info = model.sd_checkpoint_info
def delete_base_vae():
global base_vae, checkpoint_info
base_vae = None
checkpoint_info = None
def restore_base_vae(model):
global loaded_vae_file
if base_vae is not None and checkpoint_info == model.sd_checkpoint_info:
print("Restoring base VAE")
_load_vae_dict(model, base_vae)
loaded_vae_file = None
delete_base_vae()
2022-10-30 08:54:31 -06:00
def get_filename(filepath):
return os.path.splitext(os.path.basename(filepath))[0]
2022-10-30 08:54:31 -06:00
def refresh_vae_list(vae_path=vae_path, model_path=model_path):
global vae_dict, vae_list
res = {}
candidates = [
*glob.iglob(os.path.join(model_path, '**/*.vae.ckpt'), recursive=True),
*glob.iglob(os.path.join(model_path, '**/*.vae.pt'), recursive=True),
*glob.iglob(os.path.join(model_path, '**/*.vae.safetensors'), recursive=True),
2022-10-31 01:42:28 -06:00
*glob.iglob(os.path.join(vae_path, '**/*.ckpt'), recursive=True),
*glob.iglob(os.path.join(vae_path, '**/*.pt'), recursive=True),
*glob.iglob(os.path.join(vae_path, '**/*.safetensors'), recursive=True),
2022-10-30 08:54:31 -06:00
]
if shared.cmd_opts.vae_path is not None and os.path.isfile(shared.cmd_opts.vae_path):
candidates.append(shared.cmd_opts.vae_path)
for filepath in candidates:
name = get_filename(filepath)
res[name] = filepath
vae_list.clear()
vae_list.extend(default_vae_list)
vae_list.extend(list(res.keys()))
vae_dict.clear()
vae_dict.update(res)
2022-11-02 01:37:22 -06:00
vae_dict.update(default_vae_dict)
2022-10-30 08:54:31 -06:00
return vae_list
def get_vae_from_settings(vae_file="auto"):
# else, we load from settings, if not set to be default
if vae_file == "auto" and shared.opts.sd_vae is not None:
# if saved VAE settings isn't recognized, fallback to auto
vae_file = vae_dict.get(shared.opts.sd_vae, "auto")
# if VAE selected but not found, fallback to auto
if vae_file not in default_vae_values and not os.path.isfile(vae_file):
vae_file = "auto"
2022-11-19 01:20:07 -07:00
print(f"Selected VAE doesn't exist: {vae_file}")
return vae_file
def resolve_vae(checkpoint_file=None, vae_file="auto"):
2022-10-30 08:54:31 -06:00
global first_load, vae_dict, vae_list
2022-11-02 01:37:22 -06:00
# if vae_file argument is provided, it takes priority, but not saved
2022-10-30 08:54:31 -06:00
if vae_file and vae_file not in default_vae_list:
if not os.path.isfile(vae_file):
2022-11-12 20:58:15 -07:00
print(f"VAE provided as function argument doesn't exist: {vae_file}")
2022-10-30 08:54:31 -06:00
vae_file = "auto"
2022-11-02 01:37:22 -06:00
# for the first load, if vae-path is provided, it takes priority, saved, and failure is reported
2022-10-30 08:54:31 -06:00
if first_load and shared.cmd_opts.vae_path is not None:
if os.path.isfile(shared.cmd_opts.vae_path):
vae_file = shared.cmd_opts.vae_path
2022-11-02 01:37:22 -06:00
shared.opts.data['sd_vae'] = get_filename(vae_file)
2022-10-30 08:54:31 -06:00
else:
2022-11-12 20:58:15 -07:00
print(f"VAE provided as command line argument doesn't exist: {vae_file}")
# fallback to selector in settings, if vae selector not set to act as default fallback
if not shared.opts.sd_vae_as_default:
vae_file = get_vae_from_settings(vae_file)
2022-10-30 08:54:31 -06:00
# vae-path cmd arg takes priority for auto
if vae_file == "auto" and shared.cmd_opts.vae_path is not None:
if os.path.isfile(shared.cmd_opts.vae_path):
vae_file = shared.cmd_opts.vae_path
2022-11-12 20:58:15 -07:00
print(f"Using VAE provided as command line argument: {vae_file}")
2022-10-30 08:54:31 -06:00
# if still not found, try look for ".vae.pt" beside model
model_path = os.path.splitext(checkpoint_file)[0]
if vae_file == "auto":
vae_file_try = model_path + ".vae.pt"
if os.path.isfile(vae_file_try):
vae_file = vae_file_try
2022-11-12 20:58:15 -07:00
print(f"Using VAE found similar to selected model: {vae_file}")
2022-10-30 08:54:31 -06:00
# if still not found, try look for ".vae.ckpt" beside model
if vae_file == "auto":
vae_file_try = model_path + ".vae.ckpt"
if os.path.isfile(vae_file_try):
vae_file = vae_file_try
2022-11-12 20:58:15 -07:00
print(f"Using VAE found similar to selected model: {vae_file}")
# if still not found, try look for ".vae.safetensors" beside model
if vae_file == "auto":
vae_file_try = model_path + ".vae.safetensors"
if os.path.isfile(vae_file_try):
vae_file = vae_file_try
print(f"Using VAE found similar to selected model: {vae_file}")
2022-10-30 08:54:31 -06:00
# No more fallbacks for auto
if vae_file == "auto":
vae_file = None
# Last check, just because
if vae_file and not os.path.exists(vae_file):
vae_file = None
return vae_file
def load_vae(model, vae_file=None):
global first_load, vae_dict, vae_list, loaded_vae_file
# save_settings = False
2022-12-25 05:49:25 -07:00
cache_enabled = shared.opts.sd_vae_checkpoint_cache > 0
2022-10-30 08:54:31 -06:00
if vae_file:
2022-12-25 05:49:25 -07:00
if cache_enabled and vae_file in checkpoints_loaded:
# use vae checkpoint cache
print(f"Loading VAE weights [{get_filename(vae_file)}] from cache")
store_base_vae(model)
_load_vae_dict(model, checkpoints_loaded[vae_file])
else:
assert os.path.isfile(vae_file), f"VAE file doesn't exist: {vae_file}"
print(f"Loading VAE weights from: {vae_file}")
store_base_vae(model)
vae_ckpt = sd_models.read_state_dict(vae_file, map_location=shared.weight_load_location)
vae_dict_1 = {k: v for k, v in vae_ckpt.items() if k[0:4] != "loss" and k not in vae_ignore_keys}
2022-12-25 05:49:25 -07:00
_load_vae_dict(model, vae_dict_1)
if cache_enabled:
# cache newly loaded vae
checkpoints_loaded[vae_file] = vae_dict_1.copy()
# clean up cache if limit is reached
if cache_enabled:
while len(checkpoints_loaded) > shared.opts.sd_vae_checkpoint_cache + 1: # we need to count the current model
checkpoints_loaded.popitem(last=False) # LRU
2022-10-30 08:54:31 -06:00
# If vae used is not in dict, update it
# It will be removed on refresh though
2022-10-30 08:54:31 -06:00
vae_opt = get_filename(vae_file)
if vae_opt not in vae_dict:
vae_dict[vae_opt] = vae_file
vae_list.append(vae_opt)
2022-11-12 21:11:14 -07:00
elif loaded_vae_file:
restore_base_vae(model)
2022-10-30 08:54:31 -06:00
loaded_vae_file = vae_file
2022-10-30 08:54:31 -06:00
first_load = False
# don't call this from outside
def _load_vae_dict(model, vae_dict_1):
model.first_stage_model.load_state_dict(vae_dict_1)
2022-10-30 08:54:31 -06:00
model.first_stage_model.to(devices.dtype_vae)
def clear_loaded_vae():
global loaded_vae_file
loaded_vae_file = None
def reload_vae_weights(sd_model=None, vae_file="auto"):
from modules import lowvram, devices, sd_hijack
if not sd_model:
sd_model = shared.sd_model
checkpoint_info = sd_model.sd_checkpoint_info
checkpoint_file = checkpoint_info.filename
vae_file = resolve_vae(checkpoint_file, vae_file=vae_file)
if loaded_vae_file == vae_file:
return
if shared.cmd_opts.lowvram or shared.cmd_opts.medvram:
lowvram.send_everything_to_cpu()
else:
sd_model.to(devices.cpu)
sd_hijack.model_hijack.undo_hijack(sd_model)
load_vae(sd_model, vae_file)
sd_hijack.model_hijack.hijack(sd_model)
script_callbacks.model_loaded_callback(sd_model)
if not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram:
sd_model.to(devices.device)
print("VAE Weights loaded.")
return sd_model