multiple fixes

modules/cmd_args.py

@@ -114,7 +114,7 @@ def compatibility_args(opts, args):
     opts.sd_vae_as_default = True
     opts.enable_emphasis = True
     opts.enable_batch_seeds = True
-    opts.multiple_tqdm = False
+    # opts.multiple_tqdm = False
     opts.print_hypernet_extra = False
     opts.dimensions_and_batch_together = True
     opts.enable_pnginfo = True

modules/loader.py

@@ -9,7 +9,11 @@ initialized = False
 logging.getLogger("DeepSpeed").disabled = True
 # os.environ.setdefault('OMP_NUM_THREADS', 1)
 # os.environ.setdefault('MKL_NUM_THREADS', 1)
+
+# import tensorflow as tf # pylint: disable=C0411
+
 import torch # pylint: disable=C0411
+
 # torch.set_num_threads(1)
 try:
     import intel_extension_for_pytorch as ipex # pylint: disable=import-error, unused-import
@@ -19,14 +23,14 @@ except Exception:
 urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
 import torchvision # pylint: disable=W0611,C0411
 import pytorch_lightning # pytorch_lightning should be imported after torch, but it re-enables warnings on import so import once to disable them # pylint: disable=W0611,C0411
-if ".dev" in torch.__version__ or "+git" in torch.__version__:
-    torch.__long_version__ = torch.__version__
-    torch.__version__ = re.search(r'[\d.]+[\d]', torch.__version__).group(0)
 logging.getLogger("xformers").addFilter(lambda record: 'A matching Triton is not available' not in record.getMessage())
 logging.getLogger("pytorch_lightning").disabled = True
 warnings.filterwarnings(action="ignore", category=DeprecationWarning)
 warnings.filterwarnings(action="ignore", category=FutureWarning)
 warnings.filterwarnings(action="ignore", category=UserWarning, module="torchvision")
+if ".dev" in torch.__version__ or "+git" in torch.__version__:
+    torch.__long_version__ = torch.__version__
+    torch.__version__ = re.search(r'[\d.]+[\d]', torch.__version__).group(0)
 timer.startup.record("torch")
 
 from fastapi import FastAPI # pylint: disable=W0611,C0411

modules/processing.py

@@ -163,6 +163,7 @@ class StableDiffusionProcessing:
         self.refiner_steps = 5
         self.refiner_start = 0
         self.ops = []
+        self.resize_mode: int = 0
         self.ddim_discretize = shared.opts.ddim_discretize
         self.s_min_uncond = shared.opts.s_min_uncond
         self.s_churn = shared.opts.s_churn
@@ -172,9 +173,10 @@ class StableDiffusionProcessing:
         self.s_tmin = shared.opts.s_tmin
         self.s_tmax = float('inf')  # not representable as a standard ui option
         self.refiner_switch_at = 0 # a1111 compatibility item
-        self.comments = {}
-        self.is_api = False
-        self.resize_mode: int = 0
+        self.all_hr_prompts = [] # a1111 compatibility item
+        self.hr_prompt = '' # a1111 compatibility item
+        self.comments = {} # a1111 compatibility item
+        self.is_api = False # a1111 compatibility item
         shared.opts.data['clip_skip'] = clip_skip
 
     @property
@@ -1180,6 +1182,8 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
         crop_region = None
         image_mask = self.image_mask
         if image_mask is not None:
+            if type(image_mask) == list:
+                image_mask = image_mask[0]
             image_mask = image_mask.convert('L')
             if self.inpainting_mask_invert:
                 image_mask = ImageOps.invert(image_mask)

modules/sd_models.py

@@ -51,6 +51,7 @@ class CheckpointInfo:
         self.filename = filename
         self.type = ''
         relname = filename
+        app_path = os.path.abspath(script_path)
 
         def rel(fn, path):
             try:
@@ -58,6 +59,8 @@ class CheckpointInfo:
             except Exception:
                 return fn
 
+        if relname.startswith('..'):
+            relname = os.path.abspath(relname)
         if relname.startswith(shared.opts.ckpt_dir):
             relname = rel(filename, shared.opts.ckpt_dir)
         elif relname.startswith(shared.opts.diffusers_dir):
@@ -66,6 +69,8 @@ class CheckpointInfo:
             relname = rel(filename, model_path)
         elif relname.startswith(script_path):
             relname = rel(filename, script_path)
+        elif relname.startswith(app_path):
+            relname = rel(filename, app_path)
         else:
             relname = os.path.abspath(relname)
         relname, ext = os.path.splitext(relname)
@@ -705,6 +710,9 @@ def compile_diffusers(sd_model):
 
 
 def set_diffuser_options(sd_model, vae, op: str):
+    if sd_model is None:
+        shared.log.warning(f'{op} is not loaded')
+        return
     if (shared.opts.diffusers_model_cpu_offload or shared.cmd_opts.medvram) and (shared.opts.diffusers_seq_cpu_offload or shared.cmd_opts.lowvram):
         shared.log.warning(f'Setting {op}: Model CPU offload and Sequential CPU offload are not compatible')
         shared.log.debug(f'Setting {op}: disabling model CPU offload')

modules/sd_vae_approx.py

@@ -5,8 +5,6 @@ from modules import devices, paths, shared
 
 
 sd_vae_approx_model = None
-simple_weights = None
-simple_bias = None
 
 
 class VAEApprox(nn.Module):
@@ -58,24 +56,23 @@ def nn_approximation(sample): # Approximate NN
 
 def cheap_approximation(sample): # Approximate simple
     # https://discuss.huggingface.co/t/decoding-latents-to-rgb-without-upscaling/23204/2
-    global simple_weights, simple_bias # pylint: disable=global-statement
-    if simple_weights is None or simple_bias is None:
-        if shared.sd_model_type == "sdxl":
-            simple_weights = torch.tensor([
-                [0.4543,-0.2868, 0.1566,-0.4748],
-                [0.5008, 0.0952, 0.2155,-0.3268],
-                [0.5294, 0.1625,-0.0624,-0.3793]
-            ]).reshape(3, 4, 1, 1).to(sample.device)
-            simple_bias = torch.tensor([0.1375, 0.0144, -0.0675]).to(sample.device)
-        else:
-            simple_weights = torch.tensor([
-                [0.298, 0.187,-0.158,-0.184],
-                [0.207, 0.286, 0.189,-0.271],
-                [0.208, 0.173, 0.264,-0.473],
-            ]).reshape(3, 4, 1, 1).to(sample.device)
-            simple_bias = None
+    if shared.sd_model_type == "sdxl":
+        simple_weights = torch.tensor([
+            [0.4543,-0.2868, 0.1566,-0.4748],
+            [0.5008, 0.0952, 0.2155,-0.3268],
+            [0.5294, 0.1625,-0.0624,-0.3793]
+        ]).reshape(3, 4, 1, 1)
+        simple_bias = torch.tensor([0.1375, 0.0144, -0.0675])
+    else:
+        simple_weights = torch.tensor([
+            [0.298, 0.187,-0.158,-0.184],
+            [0.207, 0.286, 0.189,-0.271],
+            [0.208, 0.173, 0.264,-0.473],
+        ]).reshape(3, 4, 1, 1)
+        simple_bias = None
     try:
-        x_sample = nn.functional.conv2d(sample, simple_weights, simple_bias) # pylint: disable=not-callable
+        print('HERE', sample.dtype, sample.device)
+        x_sample = nn.functional.conv2d(sample, simple_weights.to(sample.device, sample.dtype), simple_bias.to(sample.device, sample.dtype)) # pylint: disable=not-callable
         return x_sample
     except Exception as e:
         shared.log.error(f'Decode simple: {e}')