diff --git a/modules/api/ray.py b/modules/api/ray.py
index fe5fa4d0dd5..c809b0dc830 100644
--- a/modules/api/ray.py
+++ b/modules/api/ray.py
@@ -7,9 +7,9 @@
 import os
 
 
-ray.init(os.environ.get("RAY_HEAD_ADDRESS", ""))
+#ray.init(os.environ.get("RAY_HEAD_ADDRESS", ""))
 #ray.init("ray://localhost:10001")
-
+ray.init()
 
 def ray_only():
     serve.shutdown()
diff --git a/modules/sd_models.py b/modules/sd_models.py
index 930d0bee5c8..d72a7b4a8a3 100644
--- a/modules/sd_models.py
+++ b/modules/sd_models.py
@@ -637,6 +637,13 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None):
     model_data.set_sd_model(sd_model)
     model_data.was_loaded_at_least_once = True
 
+    # Check if shared.sd_model is initialized and has 'cond_stage_model' attribute
+    if shared.sd_model is None or not hasattr(shared.sd_model, 'cond_stage_model'):
+        # Initialize shared.sd_model or add 'cond_stage_model' attribute
+        # The following is just a placeholder, replace it with actual initialization code
+        shared.sd_model = instantiate_from_config(sd_config.model)
+        shared.sd_model.cond_stage_model = instantiate_from_config(sd_config.cond_stage_model)
+
     sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings(force_reload=True)  # Reload embeddings after model load as they may or may not fit the model
 
     timer.record("load textual inversion embeddings")
@@ -654,7 +661,6 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None):
 
     return sd_model
 
-
 def reuse_model_from_already_loaded(sd_model, checkpoint_info, timer):
     """
     Checks if the desired checkpoint from checkpoint_info is not already loaded in model_data.loaded_sd_models.
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py
index aa79dc09843..7f1478f6de4 100644
--- a/modules/textual_inversion/textual_inversion.py
+++ b/modules/textual_inversion/textual_inversion.py
@@ -150,9 +150,17 @@ def register_embedding_by_name(self, embedding, model, name):
         self.word_embeddings[name] = embedding
         return embedding
 
+    # def get_expected_shape(self):
+    #     vec = shared.sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
+    #     return vec.shape[1]
+
     def get_expected_shape(self):
-        vec = shared.sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
-        return vec.shape[1]
+        if shared.sd_model is None or not hasattr(shared.sd_model, 'cond_stage_model'):
+            # Handle the error appropriately, e.g., by raising an exception or returning a default value
+            raise ValueError("shared.sd_model is not initialized or does not have a 'cond_stage_model' attribute")
+        else:
+            vec = shared.sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
+            return vec.shape[1]
 
     def load_from_file(self, path, filename):
         name, ext = os.path.splitext(filename)
diff --git a/webui-user.sh b/webui-user.sh
index d0302f940b9..2fe31534895 100644
--- a/webui-user.sh
+++ b/webui-user.sh
@@ -52,4 +52,4 @@ export RAY_MIN_REPLICAS=0
 export RAY_MAX_REPLICAS=2
 export RAY_PORT=7575
 export RAY_IP="0.0.0.0"
-#export RAY_HEAD_ADDRESS='ray://ray-head:10001'
+export RAY_HEAD_ADDRESS='ray://ray-head:10001'