Update model_manager.py - trainer config change - Issue drivendataorg…

…#240

zamba/models/model_manager.py

@@ -280,19 +280,17 @@ def train_model(
 
     accelerator, devices = configure_accelerator_and_devices_from_gpus(train_config.gpus)
 
+
+    multiprocessing_strategy = getattr(train_config,"multiprocessing_strategy",None)
+
     trainer = pl.Trainer(
         accelerator=accelerator,
         devices=devices,
         max_epochs=train_config.max_epochs,
         logger=tensorboard_logger,
-        callbacks=callbacks,
-        fast_dev_run=train_config.dry_run,
-        strategy=(
-            DDPStrategy(find_unused_parameters=False)
-            if (data_module.multiprocessing_context is not None) and (train_config.gpus > 1)
-            else "auto"
-        ),
+        strategy = multiprocessing_strategy,     
     )
+    #Set the strategy within trainer to reflect changes
 
     if video_loader_config.cache_dir is None:
         logger.info("No cache dir is specified. Videos will not be cached.")