Bugfixes for slider reference

extensions_built_in/image_reference_slider_trainer/ImageReferenceSliderTrainerProcess.py

@@ -159,7 +159,13 @@ class ImageReferenceSliderTrainerProcess(BaseSDTrainProcess):
 
         # if training text encoder enable grads, else do context of no grad
         with torch.set_grad_enabled(self.train_config.train_text_encoder):
-            conditional_embeds = self.sd.encode_prompt(prompts).to(self.device_torch, dtype=dtype)
+            # fix issue with them being tuples sometimes
+            prompt_list = []
+            for prompt in prompts:
+                if isinstance(prompt, tuple):
+                    prompt = prompt[0]
+                prompt_list.append(prompt)
+            conditional_embeds = self.sd.encode_prompt(prompt_list).to(self.device_torch, dtype=dtype)
             conditional_embeds = concat_prompt_embeds([conditional_embeds, conditional_embeds])
 
         # if self.model_config.is_xl:

toolkit/network_mixins.py

@@ -55,7 +55,7 @@ class ToolkitModuleMixin:
         self.is_checkpointing = False
         self.is_normalizing = False
         self.normalize_scaler = 1.0
-        self._multiplier: Union[float, list, torch.Tensor] = 1.0
+        self._multiplier: Union[float, list, torch.Tensor] = None
 
     # this allows us to set different multipliers on a per item in a batch basis
     # allowing us to run positive and negative weights in the same batch
@@ -123,6 +123,8 @@ class ToolkitModuleMixin:
         return lx * scale
 
     def forward(self: Module, x):
+        if self._multiplier is None:
+            self.set_multiplier(0.0)
 
         org_forwarded = self.org_forward(x)
         lora_output = self._call_forward(x)