Deeplearning/main.py

import datetime
from pathlib import Path

from Qtorch.Models.Qmlp import Qmlp
from Qfunctions.loadData import load_data
from Qfunctions.saveToXlsx import save_to_xlsx as save_to_xlsx


def _to_builtin(value):
  if isinstance(value, dict):
    return {str(k): _to_builtin(v) for k, v in value.items()}
  if isinstance(value, (list, tuple)):
    return [_to_builtin(v) for v in value]
  if hasattr(value, 'item'):
    try:
      return value.item()
    except Exception:
      return str(value)
  return value


def _yaml_scalar(value):
  if value is None:
    return 'null'
  if isinstance(value, bool):
    return 'true' if value else 'false'
  if isinstance(value, (int, float)):
    return str(value)
  text = str(value).replace('"', '\\"')
  return f'"{text}"'


def _yaml_lines(value, indent=0):
  space = ' ' * indent

  if isinstance(value, dict):
    if not value:
      return [space + '{}']
    lines = []
    for k, v in value.items():
      key = str(k)
      if isinstance(v, (dict, list, tuple)):
        lines.append(f'{space}{key}:')
        lines.extend(_yaml_lines(v, indent + 2))
      else:
        lines.append(f'{space}{key}: {_yaml_scalar(v)}')
    return lines

  if isinstance(value, (list, tuple)):
    if not value:
      return [space + '[]']
    lines = []
    for item in value:
      if isinstance(item, (dict, list, tuple)):
        lines.append(f'{space}-')
        lines.extend(_yaml_lines(item, indent + 2))
      else:
        lines.append(f'{space}- {_yaml_scalar(item)}')
    return lines

  return [space + _yaml_scalar(value)]


def _save_yaml(file_path, data):
  built_data = _to_builtin(data)
  text = '\n'.join(_yaml_lines(built_data)) + '\n'
  with open(file_path, 'w', encoding='utf-8') as f:
    f.write(text)

def main():
  # 输入元数据文件夹名称
  projet_name = '20260409 grap'
  # 请在[]内输入每一个分类的名称
  # label_names 是一个列表里面按顺序包含了小写的‘a'到‘z’
  label_names = list(range(1, 10))
  hidden_layers = [256, 128, 128, 128]
  test_size = 0.5
  dropout_rate = 0
  epochs = 300

  run_id = datetime.datetime.now().strftime('%Y%m%d_%H%M%S')
  run_project_name = f'{projet_name}/{run_id}'
  run_dir = Path('Result') / projet_name / run_id
  run_dir.mkdir(parents=True, exist_ok=True)

  print(label_names)
  data = load_data(projet_name, label_names)
  
  model = Qmlp(
    data=data,
    labels=label_names,
    hidden_layers=hidden_layers,
    test_size=test_size,
    dropout_rate=dropout_rate
    )
  # model = QCNN(
  #    data=data,
  #    labels=label_names,
  #    conv_channels=(16, 32),
  #    kernel_size=3,
  #    hidden_size=128,
  #    test_size=0.3,
  #    dropout_rate=0
  # )
  
  pca_2d, pca_3d = model.get_PCA()

  model.fit(epochs)
  
  cm = model.get_cm()
  cmn = model.get_cmn()
  epoch_data = model.get_epoch_data()

  save_to_xlsx(project_name=run_project_name, file_name='pca_2d', data=pca_2d)
  save_to_xlsx(project_name=run_project_name, file_name='pca_3d', data=pca_3d)
  save_to_xlsx(project_name=run_project_name, file_name='cm', data=cm)
  save_to_xlsx(project_name=run_project_name, file_name='cmn', data=cmn)
  save_to_xlsx(project_name=run_project_name, file_name='acc_and_loss', data=epoch_data)

  run_params = {
    'run_id': run_id,
    'project_name': projet_name,
    'label_names': label_names,
    'model': {
      'type': 'Qmlp',
      'hidden_layers': hidden_layers,
      'dropout_rate': dropout_rate,
    },
    'train': {
      'epochs': epochs,
      'test_size': test_size,
    },
    'created_at': datetime.datetime.now().isoformat(timespec='seconds'),
  }

  if not epoch_data.empty:
    best_idx = int(epoch_data['test_accuracy'].idxmax())
    best_row = epoch_data.loc[best_idx].to_dict()
    last_row = epoch_data.iloc[-1].to_dict()
    run_result = {
      'run_id': run_id,
      'result_dir': str(run_dir),
      'best_epoch': int(best_row.get('epoch', 0)),
      'best_metrics': best_row,
      'last_epoch_metrics': last_row,
      'generated_files': [
        'pca_2d.xlsx',
        'pca_2d.png',
        'pca_3d.xlsx',
        'pca_3d.png',
        'cm.xlsx',
        'cm.png',
        'cmn.xlsx',
        'cmn.png',
        'acc_and_loss.xlsx',
        'acc_and_loss_epoch.png',
        'acc_and_loss_last_epoch_bar.png',
      ],
      'created_at': datetime.datetime.now().isoformat(timespec='seconds'),
    }
  else:
    run_result = {
      'run_id': run_id,
      'result_dir': str(run_dir),
      'message': 'No epoch data generated.',
      'created_at': datetime.datetime.now().isoformat(timespec='seconds'),
    }

  _save_yaml(run_dir / 'run_params.yaml', run_params)
  _save_yaml(run_dir / 'run_result.yaml', run_result)

  print(f'Done. Result saved in: {run_dir}')
  
if __name__ == '__main__':
  main()