head_node.py

import ray
import torch
import torch.nn as nn

@ray.remote
class HeadNode:
    def __init__(self, num_parts):
        self.num_parts = num_parts
        self.model_parts = [None] * num_parts

    def receive_weights(self, part_id, weights):
        print(f"📩 Received weights from part {part_id}")
        self.model_parts[part_id] = weights

        if all(self.model_parts):
            print("🧠 All parts received, combining full model...")
            self.combine_model()

    def combine_model(self):
        # Full architecture (must match model split)
        full_model = nn.Sequential(
            nn.Linear(10, 64),
            nn.ReLU(),
            nn.Linear(64, 64),
            nn.ReLU(),
            nn.Linear(64, 1)
        )

        full_state_dict = full_model.state_dict()
        part_keys = list(full_state_dict.keys())

        # Flatten part weights into full model
        i = 0
        for part in self.model_parts:
            for key in part.keys():
                full_state_dict[part_keys[i]] = part[key]
                i += 1

        full_model.load_state_dict(full_state_dict)
        torch.save(full_model.state_dict(), "final_model.pt")
        print("✅ Full model saved as final_model.pt")