Spaces:
Sleeping
Sleeping
File size: 1,352 Bytes
10e9b7d e39d8d6 10e9b7d e39d8d6 e80aab9 e39d8d6 7637e78 e39d8d6 860a4ac d20fc11 e39d8d6 e80aab9 e39d8d6 7637e78 e39d8d6 7637e78 e39d8d6 49ad7de e39d8d6 49ad7de e39d8d6 e80aab9 7637e78 49ad7de e80aab9 e39d8d6 d20fc11 e39d8d6 |
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 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
import gradio as gr
import json
import os
from transformers import pipeline
# Load model (lightweight & allowed)
qa_pipeline = pipeline(
"text2text-generation",
model="google/flan-t5-base",
max_new_tokens=64
)
DATA_PATH = "/data/gaia_validation_questions.json"
def solve_question(question: str) -> str:
"""
Very simple baseline solver.
GAIA Unit 4 rewards formatting + correctness, not fancy agents.
"""
try:
result = qa_pipeline(question)
return result[0]["generated_text"].strip()
except Exception:
return "unknown"
def run_evaluation():
"""
Runs GAIA evaluation and returns answers in correct format
"""
with open(DATA_PATH, "r") as f:
data = json.load(f)
answers = {}
for item in data:
qid = item["id"]
question = item["question"]
answers[qid] = solve_question(question)
return answers
def submit():
"""
This function is REQUIRED by GAIA.
It must return a dict of {question_id: answer}
"""
return run_evaluation()
with gr.Blocks() as demo:
gr.Markdown("# GAIA Unit 4 – Basic Agent Runner")
run_btn = gr.Button("Run Evaluation & Submit")
output = gr.JSON(label="Submission Result")
run_btn.click(fn=submit, outputs=output)
demo.launch(server_name="0.0.0.0", server_port=7860)
|