handle keyboard interrupt for ddp .test() (#1019)

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

* updated checkpoint docs

pytorch_lightning/trainer/distrib_data_parallel.py

@@ -350,8 +350,9 @@ class TrainerDDPMixin(ABC):
         :param model:
         :return:
         """
-        path = os.path.join(self.default_save_path, '__temp_weight_ddp_end.ckpt')
-        self.save_checkpoint(path)
+        if self.proc_rank == 0:
+            path = os.path.join(self.default_save_path, '__temp_weight_ddp_end.ckpt')
+            self.save_checkpoint(path)
 
     def load_spawn_weights(self, original_model):
         """
@@ -370,6 +371,8 @@ class TrainerDDPMixin(ABC):
         # remove ddp weights
         os.remove(path)
 
+        return loaded_model
+
     def resolve_root_node_address(self, root_node):
         if '[' in root_node:
             name = root_node.split('[')[0]

pytorch_lightning/trainer/distrib_parts.py

@@ -337,6 +337,7 @@ Here lightning distributes parts of your module across available GPUs to optimiz
 from abc import ABC, abstractmethod
 import logging as log
 import os
+import signal
 
 import torch
 
@@ -494,6 +495,8 @@ class TrainerDPMixin(ABC):
         m = f'INIT TPU local core: {self.tpu_local_core_rank}, ' \
             f'global rank: {self.tpu_global_core_rank}'
         log.info(m)
+
+        # continue training routine
         self.run_pretrain_routine(model)
 
         self.save_spawn_weights(model)

pytorch_lightning/trainer/trainer.py

@@ -959,7 +959,14 @@ class Trainer(TrainerIOMixin,
                 self.ddp_train(task, model)
             else:
                 self.__set_random_port()
+
+                # track for predict
+                self.model = model
+
+                # train
                 mp.spawn(self.ddp_train, nprocs=self.num_gpus, args=(model,))
+
+                # load weights if not interrupted
                 self.load_spawn_weights(model)
                 self.model = model
 
@@ -976,7 +983,14 @@ class Trainer(TrainerIOMixin,
 
             #  COLAB_GPU is an env var available by default in Colab environments.
             start_method = 'fork' if os.getenv('COLAB_GPU') else 'spawn'
+
+            # track for predict
+            self.model = model
+
+            # train
             xmp.spawn(self.tpu_train, args=(model,), nprocs=self.num_tpu_cores, start_method=start_method)
+
+            # load weights if not interrupted
             self.load_spawn_weights(model)
             self.model = model
 
@@ -1192,12 +1206,19 @@ class Trainer(TrainerIOMixin,
             trainer = Trainer()
             trainer.test(model)
         """
+
         self.testing = True
         if model is not None:
             self.model = model
             self.fit(model)
-        elif self.model is not None and (self.use_ddp or self.use_tpu):
-            self.fit(self.model)
+        elif self.use_ddp or self.use_tpu:
+            # attempt to load weights from a spawn
+            path = os.path.join(self.default_save_path, '__temp_weight_ddp_end.ckpt')
+            test_model = self.model
+            if os.path.exists(path):
+                test_model = self.load_spawn_weights(self.model)
+
+            self.fit(test_model)
         else:
             self.run_evaluation(test_mode=True)
 
@@ -1217,21 +1238,6 @@ class _PatchDataLoader(object):
         return self.dataloader
 
 
-class _PatchDataLoader(object):
-    r'''
-    Callable object for patching dataloaders passed into trainer.fit().
-    Use this class to override model.*_dataloader() and be pickle-compatible.
-
-    Args:
-        dataloader: Dataloader object to return when called.
-    '''
-    def __init__(self, dataloader: Union[List[DataLoader], DataLoader]):
-        self.dataloader = dataloader
-
-    def __call__(self) -> Union[List[DataLoader], DataLoader]:
-        return self.dataloader
-
-
 def _set_dataloader(model, dataloader, attribute):
     r'''
     Check dataloaders passed to .fit() method if they are pytorch DataLoader

pytorch_lightning/trainer/training_io.py

@@ -6,6 +6,7 @@ import warnings
 from abc import ABC
 from subprocess import call
 from typing import Union
+from copy import deepcopy
 
 import torch
 import torch.distributed as dist
@@ -233,7 +234,9 @@ class TrainerIOMixin(ABC):
 
         # add the hparams and state_dict from the model
         model = self.get_model()
+
         checkpoint['state_dict'] = model.state_dict()
+
         if hasattr(model, "hparams"):
             checkpoint['hparams'] = vars(model.hparams)
         else: