diff options
Diffstat (limited to 'modules/sd_models.py')
| -rw-r--r-- | modules/sd_models.py | 108 |
1 files changed, 85 insertions, 23 deletions
diff --git a/modules/sd_models.py b/modules/sd_models.py index 34c57bfa..5b37f3fe 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -5,7 +5,11 @@ import gc from collections import namedtuple
import torch
import re
+import safetensors.torch
from omegaconf import OmegaConf
+from os import mkdir
+from urllib import request
+import ldm.modules.midas as midas
from ldm.util import instantiate_from_config
@@ -35,6 +39,7 @@ def setup_model(): os.makedirs(model_path)
list_models()
+ enable_midas_autodownload()
def checkpoint_tiles():
@@ -45,7 +50,7 @@ def checkpoint_tiles(): def list_models():
checkpoints_list.clear()
- model_list = modelloader.load_models(model_path=model_path, command_path=shared.cmd_opts.ckpt_dir, ext_filter=[".ckpt"])
+ model_list = modelloader.load_models(model_path=model_path, command_path=shared.cmd_opts.ckpt_dir, ext_filter=[".ckpt", ".safetensors"])
def modeltitle(path, shorthash):
abspath = os.path.abspath(path)
@@ -143,8 +148,8 @@ def transform_checkpoint_dict_key(k): def get_state_dict_from_checkpoint(pl_sd):
- if "state_dict" in pl_sd:
- pl_sd = pl_sd["state_dict"]
+ pl_sd = pl_sd.pop("state_dict", pl_sd)
+ pl_sd.pop("state_dict", None)
sd = {}
for k, v in pl_sd.items():
@@ -159,27 +164,41 @@ def get_state_dict_from_checkpoint(pl_sd): return pl_sd
+def read_state_dict(checkpoint_file, print_global_state=False, map_location=None):
+ _, extension = os.path.splitext(checkpoint_file)
+ if extension.lower() == ".safetensors":
+ pl_sd = safetensors.torch.load_file(checkpoint_file, device=map_location or shared.weight_load_location)
+ else:
+ pl_sd = torch.load(checkpoint_file, map_location=map_location or shared.weight_load_location)
+
+ if print_global_state and "global_step" in pl_sd:
+ print(f"Global Step: {pl_sd['global_step']}")
+
+ sd = get_state_dict_from_checkpoint(pl_sd)
+ return sd
+
+
def load_model_weights(model, checkpoint_info, vae_file="auto"):
checkpoint_file = checkpoint_info.filename
sd_model_hash = checkpoint_info.hash
- if shared.opts.sd_checkpoint_cache > 0 and hasattr(model, "sd_checkpoint_info"):
- sd_vae.restore_base_vae(model)
- checkpoints_loaded[model.sd_checkpoint_info] = model.state_dict().copy()
-
- vae_file = sd_vae.resolve_vae(checkpoint_file, vae_file=vae_file)
+ cache_enabled = shared.opts.sd_checkpoint_cache > 0
- if checkpoint_info not in checkpoints_loaded:
+ if cache_enabled and checkpoint_info in checkpoints_loaded:
+ # use checkpoint cache
+ print(f"Loading weights [{sd_model_hash}] from cache")
+ model.load_state_dict(checkpoints_loaded[checkpoint_info])
+ else:
+ # load from file
print(f"Loading weights [{sd_model_hash}] from {checkpoint_file}")
- pl_sd = torch.load(checkpoint_file, map_location=shared.weight_load_location)
- if "global_step" in pl_sd:
- print(f"Global Step: {pl_sd['global_step']}")
-
- sd = get_state_dict_from_checkpoint(pl_sd)
- del pl_sd
+ sd = read_state_dict(checkpoint_file)
model.load_state_dict(sd, strict=False)
del sd
+
+ if cache_enabled:
+ # cache newly loaded model
+ checkpoints_loaded[checkpoint_info] = model.state_dict().copy()
if shared.cmd_opts.opt_channelslast:
model.to(memory_format=torch.channels_last)
@@ -199,23 +218,63 @@ def load_model_weights(model, checkpoint_info, vae_file="auto"): model.first_stage_model.to(devices.dtype_vae)
- else:
- vae_name = sd_vae.get_filename(vae_file) if vae_file else None
- vae_message = f" with {vae_name} VAE" if vae_name else ""
- print(f"Loading weights [{sd_model_hash}]{vae_message} from cache")
- model.load_state_dict(checkpoints_loaded[checkpoint_info])
-
- if shared.opts.sd_checkpoint_cache > 0:
- while len(checkpoints_loaded) > shared.opts.sd_checkpoint_cache:
+ # clean up cache if limit is reached
+ if cache_enabled:
+ while len(checkpoints_loaded) > shared.opts.sd_checkpoint_cache + 1: # we need to count the current model
checkpoints_loaded.popitem(last=False) # LRU
model.sd_model_hash = sd_model_hash
model.sd_model_checkpoint = checkpoint_file
model.sd_checkpoint_info = checkpoint_info
+ sd_vae.delete_base_vae()
+ sd_vae.clear_loaded_vae()
+ vae_file = sd_vae.resolve_vae(checkpoint_file, vae_file=vae_file)
sd_vae.load_vae(model, vae_file)
+def enable_midas_autodownload():
+ """
+ Gives the ldm.modules.midas.api.load_model function automatic downloading.
+
+ When the 512-depth-ema model, and other future models like it, is loaded,
+ it calls midas.api.load_model to load the associated midas depth model.
+ This function applies a wrapper to download the model to the correct
+ location automatically.
+ """
+
+ midas_path = os.path.join(models_path, 'midas')
+
+ # stable-diffusion-stability-ai hard-codes the midas model path to
+ # a location that differs from where other scripts using this model look.
+ # HACK: Overriding the path here.
+ for k, v in midas.api.ISL_PATHS.items():
+ file_name = os.path.basename(v)
+ midas.api.ISL_PATHS[k] = os.path.join(midas_path, file_name)
+
+ midas_urls = {
+ "dpt_large": "https://github.com/intel-isl/DPT/releases/download/1_0/dpt_large-midas-2f21e586.pt",
+ "dpt_hybrid": "https://github.com/intel-isl/DPT/releases/download/1_0/dpt_hybrid-midas-501f0c75.pt",
+ "midas_v21": "https://github.com/AlexeyAB/MiDaS/releases/download/midas_dpt/midas_v21-f6b98070.pt",
+ "midas_v21_small": "https://github.com/AlexeyAB/MiDaS/releases/download/midas_dpt/midas_v21_small-70d6b9c8.pt",
+ }
+
+ midas.api.load_model_inner = midas.api.load_model
+
+ def load_model_wrapper(model_type):
+ path = midas.api.ISL_PATHS[model_type]
+ if not os.path.exists(path):
+ if not os.path.exists(midas_path):
+ mkdir(midas_path)
+
+ print(f"Downloading midas model weights for {model_type} to {path}")
+ request.urlretrieve(midas_urls[model_type], path)
+ print(f"{model_type} downloaded")
+
+ return midas.api.load_model_inner(model_type)
+
+ midas.api.load_model = load_model_wrapper
+
def load_model(checkpoint_info=None):
from modules import lowvram, sd_hijack
checkpoint_info = checkpoint_info or select_checkpoint()
@@ -243,6 +302,9 @@ def load_model(checkpoint_info=None): do_inpainting_hijack()
+ if shared.cmd_opts.no_half:
+ sd_config.model.params.unet_config.params.use_fp16 = False
+
sd_model = instantiate_from_config(sd_config.model)
load_model_weights(sd_model, checkpoint_info)
|