input_ids to float32

app.py

@@ -19,20 +19,20 @@ model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-xl", device_m
 """###Interface"""
 
 def generate(input_text, minimum_length, maximum_length, temperature, repetition_penalty):
-  input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)
-  outputs = model.generate(input_ids,
-                          min_length=minimum_length,
-                          max_new_tokens=maximum_length,
-                          length_penalty=1.4,
-                          num_beams=12,
-                          no_repeat_ngram_size=3,
-                          temperature=temperature,
-                          top_k=100,
-                          top_p=0.9,
-                          repetition_penalty=repetition_penalty,
-                          )
-
-  return tokenizer.decode(outputs[0], skip_special_tokens=True).capitalize()
+    input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("float32").to(device)
+    outputs = model.generate(input_ids,
+                            min_length=minimum_length,
+                            max_new_tokens=maximum_length,
+                            length_penalty=1.4,
+                            num_beams=12,
+                            no_repeat_ngram_size=3,
+                            temperature=temperature,
+                            top_k=100,
+                            top_p=0.9,
+                            repetition_penalty=repetition_penalty,
+                            )
+
+    return tokenizer.decode(outputs[0], skip_special_tokens=True).capitalize()
 
 title = "Flan-T5-XL GRADIO GUI"