Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -29,7 +29,108 @@ MODEL_CONFIGS = {
|
|
| 29 |
}
|
| 30 |
}
|
| 31 |
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
# 模型選擇切換函數
|
| 35 |
def update_model_choices(provider):
|
|
|
|
| 29 |
}
|
| 30 |
}
|
| 31 |
|
| 32 |
+
def get_llm_response(prompt, provider, model):
|
| 33 |
+
if provider == "openai":
|
| 34 |
+
response = openai_client.chat.completions.create(
|
| 35 |
+
model=model,
|
| 36 |
+
messages=[
|
| 37 |
+
{"role": "system", "content": "你是一位資深的國文作文評閱委員,請依據提供的評分規準進行評分。"},
|
| 38 |
+
{"role": "user", "content": prompt}
|
| 39 |
+
],
|
| 40 |
+
**MODEL_CONFIGS["openai"][model]
|
| 41 |
+
)
|
| 42 |
+
return response.choices[0].message.content
|
| 43 |
+
else: # groq
|
| 44 |
+
completion = groq_client.chat.completions.create(
|
| 45 |
+
model=model,
|
| 46 |
+
messages=[
|
| 47 |
+
{"role": "system", "content": "你是一位資深的國文作文評閱委員,請依據提供的評分規準進行評分。"},
|
| 48 |
+
{"role": "user", "content": prompt}
|
| 49 |
+
],
|
| 50 |
+
**MODEL_CONFIGS["groq"][model],
|
| 51 |
+
stream=False,
|
| 52 |
+
top_p=1,
|
| 53 |
+
stop=None
|
| 54 |
+
)
|
| 55 |
+
return completion.choices[0].message.content
|
| 56 |
+
|
| 57 |
+
def evaluate_essay(message, additional_prompt, provider, model):
|
| 58 |
+
if not message.strip():
|
| 59 |
+
return [], gr.Markdown("### 請輸入作文內容進行評分")
|
| 60 |
+
|
| 61 |
+
criteria = {
|
| 62 |
+
'題旨發揮': {'weight': 0.4, 'max_score': 40},
|
| 63 |
+
'資料掌握': {'weight': 0.2, 'max_score': 20},
|
| 64 |
+
'結構安排': {'weight': 0.2, 'max_score': 20},
|
| 65 |
+
'字句運用': {'weight': 0.2, 'max_score': 20}
|
| 66 |
+
}
|
| 67 |
+
|
| 68 |
+
grade_scores = {
|
| 69 |
+
'A+': 95, 'A': 90, 'A-': 85,
|
| 70 |
+
'B+': 80, 'B': 75, 'B-': 70,
|
| 71 |
+
'C+': 65, 'C': 60, 'C-': 55,
|
| 72 |
+
'0': 0
|
| 73 |
+
}
|
| 74 |
+
|
| 75 |
+
try:
|
| 76 |
+
history = []
|
| 77 |
+
total_score = 0
|
| 78 |
+
history.append(("作文內容:", message))
|
| 79 |
+
history.append(("", f"正在使用 {provider} ({model}) 進行評分分析..."))
|
| 80 |
+
|
| 81 |
+
all_feedback = {}
|
| 82 |
+
|
| 83 |
+
for criterion, details in criteria.items():
|
| 84 |
+
prompt = f"""評估以下作文的{criterion}(權重{details['weight']*100}%):
|
| 85 |
+
|
| 86 |
+
作文內容:
|
| 87 |
+
{message}
|
| 88 |
+
|
| 89 |
+
{additional_prompt if additional_prompt else ''}
|
| 90 |
+
|
| 91 |
+
請依三等九級制(A+、A、A-、B+、B、B-、C+、C、C-)評分,並提供簡短評語。
|
| 92 |
+
如果是缺考、未作答、完全文不對題或作答內容完全照抄試題,請給予0分。
|
| 93 |
+
|
| 94 |
+
請按以下格式回覆:
|
| 95 |
+
等第:
|
| 96 |
+
評語:"""
|
| 97 |
+
|
| 98 |
+
result = get_llm_response(prompt, provider, model)
|
| 99 |
+
lines = result.lower().split('\n')
|
| 100 |
+
grade = '0'
|
| 101 |
+
comment = ""
|
| 102 |
+
|
| 103 |
+
for line in lines:
|
| 104 |
+
if '等第:' in line or '等第:' in line:
|
| 105 |
+
grade_text = line.split(':')[-1].strip().upper()
|
| 106 |
+
if grade_text in grade_scores:
|
| 107 |
+
grade = grade_text
|
| 108 |
+
elif '評語:' in line or '評語:' in line:
|
| 109 |
+
comment = line.split(':')[-1].strip()
|
| 110 |
+
|
| 111 |
+
weighted_score = (grade_scores[grade] / 100) * details['max_score']
|
| 112 |
+
total_score += weighted_score
|
| 113 |
+
|
| 114 |
+
feedback = f"### {criterion}\n"
|
| 115 |
+
feedback += f"- **等第**:{grade}\n"
|
| 116 |
+
feedback += f"- **得分**:{weighted_score:.1f}/{details['max_score']}\n"
|
| 117 |
+
feedback += f"- **評語**:{comment}\n"
|
| 118 |
+
|
| 119 |
+
all_feedback[criterion] = feedback
|
| 120 |
+
|
| 121 |
+
for criterion in criteria:
|
| 122 |
+
history.append(("", all_feedback[criterion]))
|
| 123 |
+
|
| 124 |
+
total_score_display = f"""
|
| 125 |
+
# 總評分結果
|
| 126 |
+
## 使用模型:{provider} ({model})
|
| 127 |
+
## 總分:{total_score:.1f}/100
|
| 128 |
+
"""
|
| 129 |
+
|
| 130 |
+
return history, gr.Markdown(total_score_display)
|
| 131 |
+
|
| 132 |
+
except Exception as e:
|
| 133 |
+
return [("", f"評分過程發生錯誤:{str(e)}")], gr.Markdown("### ❌ 評分失敗")
|
| 134 |
|
| 135 |
# 模型選擇切換函數
|
| 136 |
def update_model_choices(provider):
|