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

comfy/ldm/modules/attention.py

@@ -9,7 +9,7 @@ from typing import Optional, Any
 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 einops import rearrange
 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 @@ except ImportError:
 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 .k_diffusion import external as k_diffusion_external
 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 copy
 
 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 @@ class CLIP:
         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: