Fix model_management import so it doesn't get executed twice.

comfy/ldm/modules/attention.py

@@ -9,7 +9,7 @@
 from ldm.modules.diffusionmodules.util import checkpoint
 from .sub_quadratic_attention import efficient_dot_product_attention
 
-import model_management
+from comfy import model_management
 
 from . import tomesd
 

comfy/ldm/modules/diffusionmodules/model.py

@@ -7,7 +7,7 @@
 from typing import Optional, Any
 
 from ldm.modules.attention import MemoryEfficientCrossAttention
-import model_management
+from comfy import model_management
 
 if model_management.xformers_enabled_vae():
     import xformers

comfy/ldm/modules/sub_quadratic_attention.py

@@ -24,7 +24,7 @@
 from torch import Tensor
 from typing import List
 
-import model_management
+from comfy import model_management
 
 def dynamic_slice(
     x: Tensor,

comfy/samplers.py

@@ -3,7 +3,7 @@
 from .extra_samplers import uni_pc
 import torch
 import contextlib
-import model_management
+from comfy import model_management
 from .ldm.models.diffusion.ddim import DDIMSampler
 from .ldm.modules.diffusionmodules.util import make_ddim_timesteps
 

comfy/sd.py

@@ -4,7 +4,7 @@
 
 import sd1_clip
 import sd2_clip
-import model_management
+from comfy import model_management
 from .ldm.util import instantiate_from_config
 from .ldm.models.autoencoder import AutoencoderKL
 import yaml
@@ -388,7 +388,7 @@ def encode_from_tokens(self, tokens):
         return cond
 
     def encode(self, text):
-        tokens = self.tokenizer.tokenize_with_weights(text)
+        tokens = self.tokenize(text)
         return self.encode_from_tokens(tokens)
 
 class VAE:

comfy_extras/nodes_upscale_model.py

@@ -1,6 +1,6 @@
 import os
 from comfy_extras.chainner_models import model_loading
-import model_management
+from comfy import model_management
 import torch
 import comfy.utils
 import folder_paths

nodes.py

@@ -21,16 +21,16 @@
 
 import comfy.clip_vision
 
-import model_management
+import comfy.model_management
 import importlib
 
 import folder_paths
 
 def before_node_execution():
-    model_management.throw_exception_if_processing_interrupted()
+    comfy.model_management.throw_exception_if_processing_interrupted()
 
 def interrupt_processing(value=True):
-    model_management.interrupt_current_processing(value)
+    comfy.model_management.interrupt_current_processing(value)
 
 MAX_RESOLUTION=8192
 
@@ -241,7 +241,7 @@ def load_checkpoint(self, model_path, output_vae=True, output_clip=True):
                     model_path = os.path.join(search_path, model_path)
                     break
 
-        return comfy.diffusers_convert.load_diffusers(model_path, fp16=model_management.should_use_fp16(), output_vae=output_vae, output_clip=output_clip, embedding_directory=folder_paths.get_folder_paths("embeddings"))
+        return comfy.diffusers_convert.load_diffusers(model_path, fp16=comfy.model_management.should_use_fp16(), output_vae=output_vae, output_clip=output_clip, embedding_directory=folder_paths.get_folder_paths("embeddings"))
 
 
 class unCLIPCheckpointLoader:
@@ -680,7 +680,7 @@ def set_mask(self, samples, mask):
 def common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive, negative, latent, denoise=1.0, disable_noise=False, start_step=None, last_step=None, force_full_denoise=False):
     latent_image = latent["samples"]
     noise_mask = None
-    device = model_management.get_torch_device()
+    device = comfy.model_management.get_torch_device()
 
     if disable_noise:
         noise = torch.zeros(latent_image.size(), dtype=latent_image.dtype, layout=latent_image.layout, device="cpu")
@@ -696,7 +696,7 @@ def common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive,
         noise_mask = noise_mask.to(device)
 
     real_model = None
-    model_management.load_model_gpu(model)
+    comfy.model_management.load_model_gpu(model)
     real_model = model.model
 
     noise = noise.to(device)
@@ -726,7 +726,7 @@ def common_ksampler(model, seed, steps, cfg, sampler_name, scheduler, positive,
     control_net_models = []
     for x in control_nets:
         control_net_models += x.get_control_models()
-    model_management.load_controlnet_gpu(control_net_models)
+    comfy.model_management.load_controlnet_gpu(control_net_models)
 
     if sampler_name in comfy.samplers.KSampler.SAMPLERS:
         sampler = comfy.samplers.KSampler(real_model, steps=steps, device=device, sampler=sampler_name, scheduler=scheduler, denoise=denoise, model_options=model.model_options)