Merge pull request #9 from benoitmartin88/rc

Hotfix sigint (#8)

.gitignore

@@ -1,3 +1,5 @@
+.DS_Store
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]

CHANGELOG.md

@@ -8,7 +8,12 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 # Unreleased
 
 
-# # [0.2.0] - 2019-09-20
+# [0.2.1] - 2019-10-10
+## Bugfix
+- Fix ignored SIGINT
+
+
+# [0.2.0] - 2019-09-20
 ## New
 - Add `ModuleTrainer.evaluate` method
 - Add CsvWriter to evaluate method

pytorchtrainer/__init__.py

@@ -1,5 +1,5 @@
 
-__version__ = '0.2.0'
+__version__ = '0.2.1'
 
 
 from .trainer import create_default_trainer, ModuleTrainer, State

pytorchtrainer/trainer.py

@@ -80,6 +80,7 @@ def __init__(self, model: nn.Module, optimizer: optim.Optimizer, train_function,
         signal.signal(signal.SIGINT, self.__graceful_exit)
 
         self.state = State()
+        self.stop_condition = None
         self.model = model
         self.optimizer = optimizer
         self.train_function = train_function
@@ -112,10 +113,11 @@ def train(self, train_dataloader: torch.utils.data.DataLoader, max_epochs=100, s
         if not callable(stop_condition):
             raise TypeError("Argument stop_condition should be a function.")
 
+        self.stop_condition = stop_condition
         self.model.train()  # set the module to training mode
 
         train_start = time()
-        while self.state.current_epoch < max_epochs and not stop_condition(self.state):
+        while self.state.current_epoch < max_epochs and not self.stop_condition(self.state):
             self.model.zero_grad()
 
             for self.state.current_iteration, batch in enumerate(train_dataloader):

test/test_trainer.py

@@ -188,3 +188,31 @@ def test_add_progressbar_metric_errors(self):
         self.assertRaises(TypeError, trainer.add_progressbar_metric, None, [])
         self.assertRaises(TypeError, trainer.add_progressbar_metric, "", [None])
         self.assertRaises(RuntimeError, trainer.add_progressbar_metric, "", [CsvWriter()])
+
+    def test_sigint(self):
+        max_epoch = 100
+
+        def _run(q):
+            print("_run from sub-process")
+            trainer = create_default_trainer(self.model, self.optimizer, self.criterion)
+            trainer.train(self.train_loader, max_epochs=max_epoch)
+            q.put(trainer.state.current_epoch)
+
+        import multiprocessing as mp
+        import os
+        import signal
+        import time
+
+        q = mp.Queue()
+        p = mp.Process(target=_run, args=(q,))
+        p.start()
+
+        # wait 1 second before sending sigint
+        time.sleep(1)
+        os.kill(p.pid, signal.SIGINT)
+
+        last_epoch = q.get()
+
+        p.join()
+
+        self.assertLess(last_epoch, max_epoch)