Spaces:
Sleeping
Sleeping
File size: 1,806 Bytes
785c4f7 11f2d5b 5b3d26d 7a7ebad 453eba8 5b3d26d 6a1a7af 11f2d5b 453eba8 7d3dbed 5b3d26d 7a7ebad 5b3d26d 7a7ebad 5b3d26d 7a7ebad 5b3d26d 7a7ebad 5b3d26d |
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 |
import torch
from transformers import AutoTokenizer
from model import EvoTransformerV22
from search_utils import web_search
import openai
# Load Evo model and tokenizer
model = EvoTransformerV22()
model.load_state_dict(torch.load("evo_hellaswag.pt", map_location="cpu"))
model.eval()
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
# GPT Setup
openai.api_key = "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" # 🔑 Set your actual key securely
def get_evo_response(query, options, user_context=""):
context_texts = web_search(query) + ([user_context] if user_context else [])
context_str = "\n".join(context_texts)
input_pairs = [f"{query} [SEP] {opt} [CTX] {context_str}" for opt in options]
scores = []
for pair in input_pairs:
encoded = tokenizer(pair, return_tensors="pt", truncation=True, padding="max_length", max_length=128)
with torch.no_grad():
output = model(encoded["input_ids"])
score = torch.sigmoid(output).item()
scores.append(score)
best_idx = int(scores[1] > scores[0])
return (
options[best_idx],
f"{options[0]}: {scores[0]:.3f} vs {options[1]}: {scores[1]:.3f}",
max(scores),
context_str
)
def get_gpt_response(query, user_context=""):
try:
context_block = f"\n\nContext:\n{user_context}" if user_context else ""
completion = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user", "content": query + context_block}]
)
return completion.choices[0].message.content.strip()
except Exception as e:
return f"⚠️ GPT error: {str(e)}"
|