Spaces:
Sleeping
Sleeping
File size: 1,554 Bytes
0b51cd7 917fc83 9f1f697 917fc83 0b51cd7 2441327 0b51cd7 2441327 0b51cd7 2441327 0b51cd7 2441327 0b51cd7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 |
import textwrap
from summarise import load_pipe
from scrape import scrape
from db import get_conn
from typing import Optional,List
from helpers import rows_by_tag
IDEA_PROMPT = (
" You are a senior ML researcher. CONTEXT provides a list of papers. From this list of papers, propose THREE new research projects."
"For each research project proposed, give a new Title, one-sentence on Motivation and background, two-sentences on the new method, "
"and one-sentence on Evaluation method.\n"
"===CONTEXT===\n"
"{context}\n"
"===PROJECT IDEAS===\n"
)
# ---------------------------------------------------------------------- #
def ideate_from_topic(topic: str, k: int = 8) -> Optional[str]:
rows = rows_by_tag(topic, k)
if not rows:
return None
ctx = "\n".join(f"- {t}: {s}" for t, _, s, _ in rows)
llm = load_pipe()
return llm(IDEA_PROMPT.format(context=ctx),
do_sample=False)[0]['generated_text'].strip()
# ------------------------------------------------------------------ #
def ideate_from_ids(ids: List[str]) -> Optional[str]:
from db import get_conn
conn = get_conn()
ctx = []
for pid in ids:
row = conn.execute(
"SELECT title, summary FROM papers WHERE id=?", (pid,)
).fetchone()
if row:
ctx.append(f"- {row[0]}: {row[1]}")
if not ctx:
return None
llm = load_pipe()
return llm(IDEA_PROMPT.format(context="\n".join(ctx)),
do_sample=False)[0]['generated_text'].strip() |