fix 'No candidate PyTorch version found for ZeroGPU'

app.py

@@ -4,7 +4,6 @@ import gradio as gr
 from functools import lru_cache
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
-# 可選模型列表
 MODEL_LIST = [
     "ckiplab/gpt2-tiny-chinese",
     "ckiplab/gpt2-base-chinese",
@@ -18,43 +17,33 @@ MODEL_LIST = [
 @lru_cache(maxsize=None)
 def get_pipeline(model_name):
     tok = AutoTokenizer.from_pretrained(model_name)
-    mdl = AutoModelForCausalLM.from_pretrained(model_name)
+    # By setting weights_only=False we bypass the torch.load(weights_only=True)
+    # path that is disallowed for torch<2.6 due to CVE-2025-32434 :contentReference[oaicite:1]{index=1}.
+    mdl = AutoModelForCausalLM.from_pretrained(model_name, weights_only=False)
     mdl.to("cuda")
     return pipeline("text-generation", model=mdl, tokenizer=tok, device=0)
 
 @spaces.GPU
 def suggest_next(text, model_name, k, m):
-    """在 GPU（ZeroGPU H200）上執行推理，產生 M 個長度上限 K 的下段建議。"""
-    pipe = get_pipeline(model_name)
-    outs = pipe(text, max_new_tokens=k, num_return_sequences=m, do_sample=False)
+    outs = get_pipeline(model_name)(
+        text, max_new_tokens=k, num_return_sequences=m, do_sample=False
+    )
     return [out["generated_text"][len(text):] for out in outs]
 
 def append_suggestion(current, choice):
     return current + choice
 
 with gr.Blocks() as demo:
-    gr.Markdown(
-        "## 🇹🇼 台灣中文下段預測\n"
-        "結合小型語言模型與 ZeroGPU，提供即時 IME 風格的下段文字建議。"
-    )
+    gr.Markdown("## 🇹🇼 台灣中文下段預測（ZeroGPU + Gradio v5）")
 
-    input_text = gr.TextArea(
-        label="輸入文字", lines=4, placeholder="請在此輸入起始片段…"
-    )
+    input_text = gr.TextArea(label="輸入文字", lines=4, placeholder="請在此輸入起始片段…")
 
     with gr.Row():
-        model_selector = gr.Dropdown(
-            MODEL_LIST, value=MODEL_LIST[0], label="選擇模型"
-        )
-        k_slider = gr.Slider(
-            minimum=1, maximum=50, step=1, value=5, label="K（最大新生成詞元）"
-        )
-        m_slider = gr.Slider(
-            minimum=1, maximum=10, step=1, value=5, label="M（建議數量）"
-        )
+        model_selector = gr.Dropdown(MODEL_LIST, value=MODEL_LIST[0], label="選擇模型")
+        k_slider = gr.Slider(1, 50, value=5, label="K（最大新生成詞元）")
+        m_slider = gr.Slider(1, 10, value=5, label="M（建議數量）")
 
     suggestions = gr.Dropdown([], label="建議清單", interactive=True)
-
     gpu_button = gr.Button("使用 GPU 生成建議")
 
     gpu_button.click(
@@ -62,7 +51,6 @@ with gr.Blocks() as demo:
         inputs=[input_text, model_selector, k_slider, m_slider],
         outputs=suggestions,
     )
-
     suggestions.change(
         fn=append_suggestion,
         inputs=[input_text, suggestions],