QAway-to commited on
Commit ·
3c8cb2b
1
Parent(s): 4236d9d
google/flan-t5-small . app.py v2.2
Browse files- core/interviewer.py +42 -53
core/interviewer.py
CHANGED
|
@@ -1,91 +1,80 @@
|
|
| 1 |
# core/interviewer.py
|
| 2 |
-
"""
|
| 3 |
-
🇬🇧 Interviewer logic module
|
| 4 |
-
Generates context-aware MBTI interview questions using Flan-T5 model!.
|
| 5 |
-
|
| 6 |
-
🇷🇺 Модуль логики интервьюера
|
| 7 |
-
Генерирует вопросы по категориям MBTI с использованием Flan-T5.
|
| 8 |
-
"""
|
| 9 |
|
| 10 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
| 11 |
-
import random
|
| 12 |
|
| 13 |
-
# --------------------------------------------------------------
|
| 14 |
-
# 1️⃣ Настройки
|
| 15 |
-
# --------------------------------------------------------------
|
| 16 |
QG_MODEL = "google/flan-t5-small"
|
| 17 |
|
| 18 |
tokenizer = AutoTokenizer.from_pretrained(QG_MODEL)
|
| 19 |
-
|
| 20 |
|
| 21 |
-
|
|
|
|
| 22 |
"text2text-generation",
|
| 23 |
-
model=
|
| 24 |
tokenizer=tokenizer,
|
| 25 |
max_new_tokens=40,
|
| 26 |
num_beams=4,
|
| 27 |
no_repeat_ngram_size=4,
|
| 28 |
)
|
| 29 |
|
| 30 |
-
# --------------------------------------------------------------
|
| 31 |
-
# 2️⃣ Состояние сессии
|
| 32 |
-
# --------------------------------------------------------------
|
| 33 |
session_state = {
|
| 34 |
"history": {},
|
| 35 |
"categories": [
|
| 36 |
"Extroversion", "Introversion",
|
| 37 |
"Sensing", "Intuition",
|
| 38 |
"Thinking", "Feeling",
|
| 39 |
-
"Judging", "Perceiving"
|
| 40 |
],
|
| 41 |
}
|
| 42 |
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
#
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
"""
|
| 49 |
-
|
| 50 |
-
|
|
|
|
| 51 |
"""
|
| 52 |
history = session_state["history"].get(user_id, {"asked": []})
|
| 53 |
asked = history["asked"]
|
| 54 |
-
|
| 55 |
|
| 56 |
-
|
| 57 |
-
if len(asked) >= len(all_cats):
|
| 58 |
return "✅ All MBTI axes covered."
|
| 59 |
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
history["asked"].append(next_cat)
|
| 63 |
session_state["history"][user_id] = history
|
| 64 |
|
| 65 |
-
# Промпт для T5
|
| 66 |
prompt = (
|
| 67 |
-
f"
|
| 68 |
-
f"
|
|
|
|
|
|
|
|
|
|
| 69 |
)
|
| 70 |
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
return f"⚠️ Generation error: {e}"
|
| 75 |
-
|
| 76 |
-
# Очистка мусора — чтобы не было “Generate a question about...”
|
| 77 |
-
bad_phrases = [
|
| 78 |
-
"generate", "question about", "output", "explain", "instruction", "user said"
|
| 79 |
-
]
|
| 80 |
-
for bp in bad_phrases:
|
| 81 |
-
if bp.lower() in output.lower():
|
| 82 |
-
output = output.split(bp, 1)[-1].strip().lstrip(":").strip()
|
| 83 |
-
|
| 84 |
-
# Убеждаемся, что начинается с нужного слова
|
| 85 |
-
if not output[0].isupper():
|
| 86 |
-
output = output.capitalize()
|
| 87 |
|
| 88 |
-
|
| 89 |
-
|
|
|
|
| 90 |
|
| 91 |
-
return f"({next_cat}) {
|
|
|
|
| 1 |
# core/interviewer.py
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
|
|
|
|
| 4 |
|
|
|
|
|
|
|
|
|
|
| 5 |
QG_MODEL = "google/flan-t5-small"
|
| 6 |
|
| 7 |
tokenizer = AutoTokenizer.from_pretrained(QG_MODEL)
|
| 8 |
+
_model = AutoModelForSeq2SeqLM.from_pretrained(QG_MODEL)
|
| 9 |
|
| 10 |
+
# используем заглавное имя, чтобы отличать от параметра в функции
|
| 11 |
+
QG_PIPE = pipeline(
|
| 12 |
"text2text-generation",
|
| 13 |
+
model=_model,
|
| 14 |
tokenizer=tokenizer,
|
| 15 |
max_new_tokens=40,
|
| 16 |
num_beams=4,
|
| 17 |
no_repeat_ngram_size=4,
|
| 18 |
)
|
| 19 |
|
|
|
|
|
|
|
|
|
|
| 20 |
session_state = {
|
| 21 |
"history": {},
|
| 22 |
"categories": [
|
| 23 |
"Extroversion", "Introversion",
|
| 24 |
"Sensing", "Intuition",
|
| 25 |
"Thinking", "Feeling",
|
| 26 |
+
"Judging", "Perceiving",
|
| 27 |
],
|
| 28 |
}
|
| 29 |
|
| 30 |
+
def _clean(q: str) -> str:
|
| 31 |
+
q = (q or "").strip()
|
| 32 |
+
# выбрасываем служебные словечки, если модель их вставила
|
| 33 |
+
bad = ["generate", "question", "output", "explain", "instruction", "user said", "based on"]
|
| 34 |
+
lower = q.lower()
|
| 35 |
+
for b in bad:
|
| 36 |
+
if b in lower:
|
| 37 |
+
# забираем всё после найденной подстроки
|
| 38 |
+
idx = lower.find(b) + len(b)
|
| 39 |
+
q = q[idx:].lstrip(":,. ").strip()
|
| 40 |
+
lower = q.lower()
|
| 41 |
+
if q and not q[0].isupper():
|
| 42 |
+
q = q.capitalize()
|
| 43 |
+
if "?" not in q:
|
| 44 |
+
q = q.rstrip(".") + "?"
|
| 45 |
+
return q
|
| 46 |
+
|
| 47 |
+
def generate_question(user_id: str, user_answer: str = None, qg_pipe=None, **kwargs) -> str:
|
| 48 |
"""
|
| 49 |
+
Возвращает один новый вопрос по следующей неиспользованной MBTI-оси.
|
| 50 |
+
Параметр qg_pipe — необязателен, игнорируем если не передали.
|
| 51 |
+
**kwargs проглатываем, чтобы не падать, если вызывающий код шлёт лишнее.
|
| 52 |
"""
|
| 53 |
history = session_state["history"].get(user_id, {"asked": []})
|
| 54 |
asked = history["asked"]
|
| 55 |
+
cats = session_state["categories"]
|
| 56 |
|
| 57 |
+
if len(asked) >= len(cats):
|
|
|
|
| 58 |
return "✅ All MBTI axes covered."
|
| 59 |
|
| 60 |
+
next_cat = next(c for c in cats if c not in asked)
|
| 61 |
+
asked.append(next_cat)
|
|
|
|
| 62 |
session_state["history"][user_id] = history
|
| 63 |
|
|
|
|
| 64 |
prompt = (
|
| 65 |
+
f"Ask one concise, open-ended question about {next_cat}. "
|
| 66 |
+
f"Start with What/Why/How/When. "
|
| 67 |
+
f"Do not mention instructions. "
|
| 68 |
+
f"Do not repeat or quote the user. "
|
| 69 |
+
f"User context: {user_answer or ''}"
|
| 70 |
)
|
| 71 |
|
| 72 |
+
pipe = qg_pipe or QG_PIPE
|
| 73 |
+
out = pipe(prompt)[0]["generated_text"]
|
| 74 |
+
question = _clean(out)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
+
# небольшой страховочный фоллбэк
|
| 77 |
+
if not question or len(question.split()) < 3:
|
| 78 |
+
question = f"What aspects of {next_cat.lower()} best describe you and why?"
|
| 79 |
|
| 80 |
+
return f"({next_cat}) {question}"
|