运行RCAN项目代码遇坑总结

一.【Pytorch1.1】ImportError: cannot import name ‘_update_worker_pids’ from ‘torch._C’
遇到这个问题首先将自己的pytorch版本改为1.1版本，然后改如下代码：
1.

from torch._C import _set_worker_signal_handlers, _update_worker_pids, \
    _remove_worker_pids, _error_if_any_worker_fails
from torch.utils.data.dataloader import DataLoader
from torch.utils.data.dataloader import _DataLoaderIter
from torch.utils.data.dataloader import ManagerWatchdog
from torch.utils.data.dataloader import _pin_memory_loop
from torch.utils.data.dataloader import MP_STATUS_CHECK_INTERVAL
 
from torch.utils.data.dataloader import ExceptionWrapper
from torch.utils.data.dataloader import _use_shared_memory
from torch.utils.data.dataloader import numpy_type_map
from torch.utils.data.dataloader import default_collate
from torch.utils.data.dataloader import pin_memory_batch
from torch.utils.data.dataloader import _SIGCHLD_handler_set
from torch.utils.data.dataloader import _set_SIGCHLD_handler

改为：

from torch._C import _set_worker_signal_handlers
from torch.utils.data import _utils
from torch.utils.data.dataloader import DataLoader
from torch.utils.data.dataloader import _DataLoaderIter
 
_use_shared_memory = False

except Exception:
    data_queue.put((idx, ExceptionWrapper(sys.exc_info())))

改为：

except Exception:
   data_queue.put((idx, _utils.ExceptionWrapper(sys.exc_info())))

首先将：

self.worker_result_queue = multiprocessing.SimpleQueue()

改为：

self.worker_result_queue = multiprocessing.Queue()

再将：

self.worker_manager_thread = threading.Thread(
    target=_pin_memory_loop,
    args=(self.worker_result_queue, self.data_queue, self.done_event, self.pin_memory,
          maybe_device_id))
self.worker_manager_thread.daemon = True
self.worker_manager_thread.start()

改为：

self.pin_memory_thread = threading.Thread(
    target=_pin_memory_loop,
    args=(self.worker_result_queue, self.data_queue, maybe_device_id, self.done_event))
self.pin_memory_thread.daemon = True
self.pin_memory_thread.start()

最后将：

target=_pin_memory_loop,

改为：

 target=_utils.pin_memory._pin_memory_loop,

_update_worker_pids(id(self), tuple(w.pid for w in self.workers))
_set_SIGCHLD_handler()

改为：

_utils.signal_handling._set_worker_pids(id(self), tuple(w.pid for w in self.workers))
_utils.signal_handling._set_SIGCHLD_handler()

collate_fn=default_collate

改为：

collate_fn=_utils.collate.default_collate

二.如果又出现：

ValueError: x and y must have same first dimension, but have shapes (59,) and (58,)

则将：

将trainer.py中所有的：
self.scheduler.get_last_epoch() + 1

改为：

self.scheduler.get_last_epoch()