Spaces:
Sleeping
Sleeping
File size: 2,008 Bytes
e65d0ad 0bae633 e65d0ad 3d71f2e e65d0ad 3d71f2e e65d0ad 3d71f2e 0bae633 3d71f2e e65d0ad 0bae633 3d71f2e 0bae633 e65d0ad 0bae633 e65d0ad 0bae633 |
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 |
import json
import re
from transformers import AutoTokenizer, AutoModelForCausalLM
# Global variables to cache the model and tokenizer
tokenizer, model = None, None
def load_model():
global tokenizer, model
if tokenizer is None or model is None:
model_name = "Salesforce/codegen-350M-mono"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
return tokenizer, model
def evaluate_code(question, code):
# Refined prompt instructing the model to output exactly valid JSON.
prompt = f"""You are an expert code evaluator.
Evaluate the following solution for the given problem.
Return ONLY a JSON object with exactly two keys:
"stars": an integer between 0 and 5 (0 means completely incorrect, 5 means excellent).
"feedback": a concise message.
Output must be exactly valid JSON and nothing else.
Problem: "{question}"
Solution: "{code}"
"""
tokenizer, model = load_model()
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(
**inputs,
max_new_tokens=50,
temperature=0.0,
pad_token_id=tokenizer.eos_token_id,
do_sample=False
)
response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
match = re.search(r'\{.*\}', response_text)
if match:
json_text = match.group(0)
try:
result = json.loads(json_text)
except Exception as e:
result = {"stars": 0, "feedback": "Evaluation failed. Unable to parse AI response."}
else:
result = {"stars": 0, "feedback": "Evaluation failed. Unable to extract JSON from AI response."}
return result
if __name__ == "__main__":
import sys
if len(sys.argv) < 3:
print(json.dumps({"error": "Please provide a question and code as arguments"}))
sys.exit(1)
question = sys.argv[1]
code = sys.argv[2]
result = evaluate_code(question, code)
print(json.dumps(result))
|