inference.py

import torch
from torch.utils.data import DataLoader
import json
import os

from nets.model import Model
from Actor.actor import Actor
from dataloader import VRP_Dataset

# --- تحميل الإعدادات ---
with open('/data/params_saved.json', 'r') as f:
    params = json.load(f)

# --- تعيين الجهاز ---
device = params['device']
dataset_path = params['dataset_path']
input_size = None  # سيتم تحديده بعد تحميل البيانات

# --- تحميل نموذج مدرب ---
model_path = "/data/model_state_dict.pt"
if not os.path.exists(model_path):
    raise FileNotFoundError(f"Model not found at {model_path}")

# --- إعداد بيانات عشوائية للاختبار ---
inference_dataset = VRP_Dataset(
    size=1,
    num_nodes=params['num_nodes'],
    num_depots=params['num_depots'],
    path=dataset_path,
    device=device
)

input_size = inference_dataset.model_input_length()

# --- تحميل النموذج ---
model = Model(
    input_size=input_size,
    embedding_size=params["embedding_size"],
    decoder_input_size=params["decoder_input_size"]
)
model.load_state_dict(torch.load(model_path, map_location=device))

# --- تهيئة الممثل (Actor) والـ NN Actor ---
actor = Actor(model=model,
              num_movers=params['num_movers'],
              num_neighbors_encoder=params['num_neighbors_encoder'],
              num_neighbors_action=params['num_neighbors_action'],
              device=device,
              normalize=False)
actor.eval_mode()

nn_actor = Actor(model=None, num_movers=1, num_neighbors_action=1, device=device)
nn_actor.nearest_neighbors()

# --- تنفيذ الاستدلال على دفعة واحدة ---
dataloader = DataLoader(inference_dataset, batch_size=1, collate_fn=inference_dataset.collate)
for batch in dataloader:
    with torch.no_grad():
        actor.greedy_search()
        actor_output = actor(batch)
        total_time = actor_output['total_time'].item()

        nn_output = nn_actor(batch)
        nn_time = nn_output['total_time'].item()

        print("\n===== INFERENCE RESULT =====")
        print(f"Actor Model Total Cost: {total_time:.4f}")
        print(f"Nearest Neighbor Cost : {nn_time:.4f}")
        print(f"Improvement over NN    : {(nn_time - total_time) / nn_time * 100:.2f}%")