Update app.py

app.py

@@ -1,30 +1,39 @@
-import streamlit as st
-from transformers import pipeline
+from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
+import torch
 
-# Load the text classification model pipeline
-classifier = pipeline("text-classification",model='isom5240ust/bert-base-uncased-emotion', return_all_scores=True)
+# ✅ Step 1: Emoji 翻译模型（你自己训练的模型）
+emoji_model_id = "JenniferHJF/qwen1.5-emoji-finetuned"
+emoji_tokenizer = AutoTokenizer.from_pretrained(emoji_model_id, trust_remote_code=True)
+emoji_model = AutoModelForCausalLM.from_pretrained(
+    emoji_model_id,
+    trust_remote_code=True,
+    torch_dtype=torch.float16
+).to("cuda" if torch.cuda.is_available() else "cpu")
+emoji_model.eval()
 
-# Streamlit application title
-st.title("Text Classification for you")
-st.write("Classification for 6 emotions: sadness, joy, love, anger, fear, surprise")
+# ✅ Step 2: 冒犯性文本识别模型
+classifier = pipeline("text-classification", model="unitary/toxic-bert", device=0 if torch.cuda.is_available() else -1)
 
-# Text input for user to enter the text to classify
-text = st.text_area("Enter the text to classify", "")
+def classify_emoji_text(text: str):
+    """
+    Step 1: 翻译文本中的 emoji
+    Step 2: 使用分类器判断是否冒犯
+    """
+    prompt = f"""请判断下面的文本是否具有冒犯性。
+这里的“冒犯性”主要指包含人身攻击、侮辱、歧视、仇恨言论或极端粗俗的内容。
+如果文本具有冒犯性，请仅回复冒犯；如果不具有冒犯性，请仅回复不冒犯。
+文本如下：
+{text}
+"""
 
-# Perform text classification when the user clicks the "Classify" button
-if st.button("Classify"):
-    # Perform text classification on the input text
-    results = classifier(text)[0]
+    input_ids = emoji_tokenizer(prompt, return_tensors="pt").to(emoji_model.device)
+    with torch.no_grad():
+        output_ids = emoji_model.generate(**input_ids, max_new_tokens=50, do_sample=False)
+    decoded = emoji_tokenizer.decode(output_ids[0], skip_special_tokens=True)
+    translated_text = decoded.strip().split("文本如下：")[-1].strip()
 
-    # Display the classification result
-    max_score = float('-inf')
-    max_label = ''
+    result = classifier(translated_text)[0]
+    label = result["label"]
+    score = result["score"]
 
-    for result in results:
-        if result['score'] > max_score:
-            max_score = result['score']
-            max_label = result['label']
-
-    st.write("Text:", text)
-    st.write("Label:", max_label)
-    st.write("Score:", max_score)
+    return translated_text, label, score