pytorch_lightning Default process group is not initialized

目录

 关于验证精度

关于分布式训练

---

pytorch_lightning Default process group is not initialized

学习nanodet训练自己的数据，

  关于验证精度

# self.metric_names = ["mAP", "AP_50", "AP_75", "AP_small", "AP_m", "AP_l"]
self.metric_names = ["mAP", "AP_50", "AP_75", "AP_small", "AP_m"]

保存依据：

evaluator:
  name: CocoDetectionEvaluator
  save_key: mAP

关于分布式训练

    accelerator = None if len(cfg.device.gpu_ids) <= 1 else "ddp"

    trainer = pl.Trainer(
        default_root_dir=my_save_dir,
        max_epochs=cfg.schedule.total_epochs,