Update app.py

app.py

@@ -146,12 +146,6 @@ theme = gr.themes.Monochrome(
     ],
 )
 
-def stream(model, code, generate_kwargs):
-    input_ids = tokenizer(code, return_tensors="pt").to(device)
-    # generated_ids = model.generate(**input_ids, **generate_kwargs)
-    generated_ids = model.generate(**input_ids)
-    return tokenizer.decode(generated_ids[0][input_ids["input_ids"].shape[1]:], skip_special_tokens=True).strip()
-
 @spaces.GPU
 def generate(
     prompt, temperature=0.6, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0, library="LangChain", method="Prefix"
@@ -174,20 +168,25 @@ def generate(
     if method == "Base":
         output = stream(basemodel, prompt, generate_kwargs)
     elif method == "Prefix":
-        output = stream(model_map[library + " Prefix"], prompt, generate_kwargs)
+        output = model_map[library + " Prefix"]
     elif method == "Evo Prefix" and library in ["SQLModel", "SfePy", "MegEngine"]:
-        output = stream(model_map["Main Evo Prefix"], prompt, generate_kwargs)
+        output = model_map["Main Evo Prefix"]
     elif method == "FFT" and library in ["SQLModel", "SfePy", "MegEngine"]:
-        output = stream(model_map[library + " FFT"], prompt, generate_kwargs)
+        output = model_map[library + " FFT"]
     elif method == "Evo FFT" and library in ["SQLModel", "SfePy", "MegEngine"]:
-        output = stream(model_map["Main Evo FFT"], prompt, generate_kwargs)
+        output = model_map["Main Evo FFT"]
     elif method == "Full Data FFT" and library in ["SQLModel", "SfePy", "MegEngine"]:
-        output = stream(model_map["Main FD FFT"], prompt, generate_kwargs)
+        output = model_map["Main FD FFT"]
     elif method == "Evo Prefix" and library in ["LangChain", "LlamaIndex", "DSPy"]:
-        output = stream(model_map["CS Evo Prefix"], prompt, generate_kwargs)
+        model = model_map["CS Evo Prefix"]
     else:
         output = ""
-    return output
+    model.to(device)
+    input_ids = tokenizer(code, return_tensors="pt").to(device)
+    # generated_ids = model.generate(**input_ids, **generate_kwargs)
+    generated_ids = model.generate(**input_ids)
+    
+    return tokenizer.decode(generated_ids[0][input_ids["input_ids"].shape[1]:], skip_special_tokens=True).strip()
 
 
 examples = [