horizonnet/misc/utils.py

import torch
import torch.nn as nn
from collections import OrderedDict


def group_weight(module):
    # Group module parameters into two group
    # One need weight_decay and the other doesn't
    group_decay = []
    group_no_decay = []
    for m in module.modules():
        if isinstance(m, nn.Linear):
            group_decay.append(m.weight)
            if m.bias is not None:
                group_no_decay.append(m.bias)
        elif isinstance(m, nn.modules.conv._ConvNd):
            group_decay.append(m.weight)
            if m.bias is not None:
                group_no_decay.append(m.bias)
        elif isinstance(m, nn.modules.batchnorm._BatchNorm):
            if m.weight is not None:
                group_no_decay.append(m.weight)
            if m.bias is not None:
                group_no_decay.append(m.bias)
        elif isinstance(m, nn.GroupNorm):
            if m.weight is not None:
                group_no_decay.append(m.weight)
            if m.bias is not None:
                group_no_decay.append(m.bias)

    assert len(list(module.parameters())) == len(group_decay) + len(group_no_decay)
    return [dict(params=group_decay), dict(params=group_no_decay, weight_decay=.0)]


def adjust_learning_rate(optimizer, args):
    if args.cur_iter < args.warmup_iters:
        frac = args.cur_iter / args.warmup_iters
        step = args.lr - args.warmup_lr
        args.running_lr = args.warmup_lr + step * frac
    else:
        frac = (float(args.cur_iter) - args.warmup_iters) / (args.max_iters - args.warmup_iters)
        scale_running_lr = max((1. - frac), 0.) ** args.lr_pow
        args.running_lr = args.lr * scale_running_lr

    for param_group in optimizer.param_groups:
        param_group['lr'] = args.running_lr


def save_model(net, path, args):
    state_dict = OrderedDict({
        'args': args.__dict__,
        'kwargs': {
            'backbone': net.backbone,
            'use_rnn': net.use_rnn,
        },
        'state_dict': net.state_dict(),
    })
    torch.save(state_dict, path)


def load_trained_model(Net, path):
    state_dict = torch.load(path, map_location='cpu')
    
    # Compatibilidad con diferentes formatos de checkpoint
    if 'kwargs' in state_dict:
        # Formato antiguo: con kwargs
        net = Net(**state_dict['kwargs'])
    elif 'backbone' in state_dict:
        # Formato de entrenamiento: con backbone directo
        backbone = state_dict.get('backbone', 'resnet50')
        net = Net(backbone, use_rnn=True)
    else:
        # Fallback: usar valores por defecto
        net = Net('resnet50', use_rnn=True)
    
    net.load_state_dict(state_dict['state_dict'])
    return net