fix

app.py

@@ -1,8 +1,8 @@
-import gc, json, torch, gradio as gr, time
+import gc, json, torch, gradio as gr
 from huggingface_hub import hf_hub_download
 import tiktoken
 
-from mingpt.model import GPT
+from mingpt.model import GPT 
 
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 REPO_ID = "to0ony/final-thesis-plotgen"
@@ -21,6 +21,7 @@ def load_model():
 
     gcfg = GPT.get_default_config()
     gcfg.model_type = None
+
     gcfg.vocab_size = int(cfg["vocab_size"])
     gcfg.block_size = int(cfg["block_size"])
     gcfg.n_layer    = int(cfg["n_layer"])
@@ -36,6 +37,7 @@ def load_model():
     state["model"] = model
     return model
 
+
 @torch.inference_mode()
 def generate(prompt, max_new_tokens=200, temperature=0.9, top_k=50):
     """Generiranje teksta iz prompta"""
@@ -53,18 +55,6 @@ def generate(prompt, max_new_tokens=200, temperature=0.9, top_k=50):
 
     return enc.decode(y[0].tolist())
 
-def typewriter(prompt, max_new_tokens=200, temperature=0.9, top_k=50, delay=0.02):
-    """
-    UI-streaming: pozove generate() jednom, a zatim ispisuje
-    rezultat riječ-po-riječ radi efekta tipkanja.
-    """
-    full = generate(prompt, max_new_tokens, temperature, top_k)
-    acc = []
-    for word in full.split(" "):
-        acc.append(word)
-        yield " ".join(acc)
-        time.sleep(delay)
-
 # Gradio UI
 with gr.Blocks(title="🎬 Final Thesis Plot Generator") as demo:
     gr.Markdown("## 🎬 Film Plot Generator\nUnesi prompt i generiraj radnju filma.")
@@ -76,8 +66,7 @@ with gr.Blocks(title="🎬 Final Thesis Plot Generator") as demo:
     btn = gr.Button("Generate")
     output = gr.Textbox(label="Output", lines=15)
 
-    # koristimo typewriter za streaming outputa; generate() ostaje netaknut
-    btn.click(typewriter, [prompt, max_new_tokens, temperature, top_k], output)
+    btn.click(generate, [prompt, max_new_tokens, temperature, top_k], output)
 
 if __name__ == "__main__":
-    demo.queue().launch()
+    demo.launch()