[IP2P] Make text encoder truly optional in InstructPi2Pix (#6995)

* make text encoder component truly optional.

* more fixes

* Apply suggestions from code review

Co-authored-by: YiYi Xu <[email protected]>

---------

Co-authored-by: YiYi Xu <[email protected]>

Author: sayakpaul

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_instruct_pix2pix.py

@@ -553,13 +553,15 @@ def _encode_prompt(
             else:
                 attention_mask = None
 
-            prompt_embeds = self.text_encoder(
-                text_input_ids.to(device),
-                attention_mask=attention_mask,
-            )
+            prompt_embeds = self.text_encoder(text_input_ids.to(device), attention_mask=attention_mask)
             prompt_embeds = prompt_embeds[0]
 
-        prompt_embeds = prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+        if self.text_encoder is not None:
+            prompt_embeds_dtype = self.text_encoder.dtype
+        else:
+            prompt_embeds_dtype = self.unet.dtype
+
+        prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
 
         bs_embed, seq_len, _ = prompt_embeds.shape
         # duplicate text embeddings for each generation per prompt, using mps friendly method
@@ -615,7 +617,7 @@ def _encode_prompt(
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = negative_prompt_embeds.shape[1]
 
-            negative_prompt_embeds = negative_prompt_embeds.to(dtype=self.text_encoder.dtype, device=device)
+            negative_prompt_embeds = negative_prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
 
             negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
             negative_prompt_embeds = negative_prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)