use prompt embeddings rather than prompt strings

app.py

@@ -11,11 +11,39 @@ pipe = DiffusionPipeline.from_pretrained(
 )
 pipe.to('cuda')
 
+
+def build_embeddings(enhanced_prompt, negative_prompt=None):
+    max_length = pipe.tokenizer.model_max_length
+
+    input_ids = pipe.tokenizer(enhanced_prompt, return_tensors="pt").input_ids
+    input_ids = input_ids.to("cuda")
+
+    negative_ids = pipe.tokenizer(
+        negative_prompt or "",
+        truncation=False,
+        padding="max_length",
+        max_length=input_ids.shape[-1],
+        return_tensors="pt"
+    ).input_ids
+    negative_ids = negative_ids.to("cuda")
+
+    concat_embeds = []
+    neg_embeds = []
+    for i in range(0, input_ids.shape[-1], max_length):
+        concat_embeds.append(pipe.text_encoder(input_ids[:, i: i + max_length])[0])
+        neg_embeds.append(pipe.text_encoder(negative_ids[:, i: i + max_length])[0])
+
+    prompt_embeds = torch.cat(concat_embeds, dim=1)
+    negative_prompt_embeds = torch.cat(neg_embeds, dim=1)
+    return prompt_embeds, negative_prompt_embeds
+
+
 @spaces.GPU
 def generate(prompt, negative_prompt, num_inference_steps, guidance_scale, width, height, num_samples):
+    prompt_embeds, neg_prompt_embeds = build_embeddings(prompt, negative_prompt)
     return pipe(
-        prompt,
-        negative_prompt=negative_prompt,
+        prompt_embeds=prompt_embeds,
+        negative_prompt_embeds=negative_prompt_embeds,
         num_inference_steps=num_inference_steps,
         guidance_scale=guidance_scale,
         width=width,