release v0.2.5.1

add a hook for on_tng_metrics so that users get access to the grad_no…

README.md

@@ -116,8 +116,8 @@ def validation_end(self, outputs):
     return tqdm_dic
 ```
    
-## TensorboardX    
+## Tensorboard    
-Lightning is fully integrated with tensorboardX.   
+Lightning is fully integrated with tensorboard.   
 
 <p align="center">
   <a href="https://williamfalcon.github.io/pytorch-lightning/">

docs/Trainer/Training Loop.md

@@ -19,7 +19,7 @@ Cut the learning rate by 10 at every epoch listed in this list.
 trainer = Trainer(lr_scheduler_milestones=None)
 
 # cut LR by 10 at 100, 200, and 300 epochs 
-trainer = Trainer(lr_scheduler_milestones=[100, 200, 300])
+trainer = Trainer(lr_scheduler_milestones='100, 200, 300')
 ```
 
 ---

examples/new_project_templates/lightning_module_template.py

@@ -84,9 +84,10 @@ class LightningTemplateModel(LightningModule):
         loss_val = self.loss(y, y_hat)
 
         output = OrderedDict({
-            'loss': loss_val,
+            'loss': loss_val
-            'tqdm_metrics': {}
         })
+
+        # can also return just a scalar instead of a dict (return loss_val)
         return output
 
     def validation_step(self, data_batch, batch_i):
@@ -107,8 +108,10 @@ class LightningTemplateModel(LightningModule):
 
         output = OrderedDict({
             'val_loss': loss_val,
-            'val_acc': torch.tensor(val_acc),
+            'val_acc': torch.tensor(val_acc).cuda(loss_val.device.index),
         })
+
+        # can also return just a scalar instead of a dict (return loss_val)
         return output
 
     def validation_end(self, outputs):
@@ -117,6 +120,10 @@ class LightningTemplateModel(LightningModule):
         :param outputs: list of individual outputs of each validation step
         :return:
         """
+        # if returned a scalar from validation_step, outputs is a list of tensor scalars
+        # we return just the average in this case (if we want)
+        # return torch.stack(outputs).mean()
+
         val_loss_mean = 0
         val_acc_mean = 0
         for output in outputs:

examples/new_project_templates/single_gpu_node_dp_template.py

@@ -0,0 +1,112 @@
+"""
+Runs a model on a single node across N-gpus.
+"""
+import os
+import sys
+import numpy as np
+from time import sleep
+import torch
+
+from test_tube import HyperOptArgumentParser, Experiment, SlurmCluster
+from pytorch_lightning.models.trainer import Trainer
+from pytorch_lightning.utils.arg_parse import add_default_args
+
+from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint
+
+SEED = 2334
+torch.manual_seed(SEED)
+np.random.seed(SEED)
+
+from lightning_module_template import LightningTemplateModel
+
+
+def main(hparams):
+    """
+    Main training routine specific for this project
+    :param hparams:
+    :return:
+    """
+    # ------------------------
+    # 1 INIT LIGHTNING MODEL
+    # ------------------------
+    print('loading model...')
+    model = LightningTemplateModel(hparams)
+    print('model built')
+
+    # ------------------------
+    # 2 INIT TEST TUBE EXP
+    # ------------------------
+
+    # init experiment
+    exp = Experiment(
+        name=hyperparams.experiment_name,
+        save_dir=hyperparams.test_tube_save_path,
+        autosave=False,
+        description='test demo'
+    )
+
+    exp.argparse(hparams)
+    exp.save()
+
+    # ------------------------
+    # 3 DEFINE CALLBACKS
+    # ------------------------
+    model_save_path = '{}/{}/{}'.format(hparams.model_save_path, exp.name, exp.version)
+    early_stop = EarlyStopping(
+        monitor='val_acc',
+        patience=3,
+        verbose=True,
+        mode='max'
+    )
+
+    checkpoint = ModelCheckpoint(
+        filepath=model_save_path,
+        save_best_only=True,
+        verbose=True,
+        monitor='val_loss',
+        mode='min'
+    )
+
+    # ------------------------
+    # 4 INIT TRAINER
+    # ------------------------
+    trainer = Trainer(
+        experiment=exp,
+        checkpoint_callback=checkpoint,
+        early_stop_callback=early_stop,
+        gpus=hparams.gpus,
+    )
+
+    # ------------------------
+    # 5 START TRAINING
+    # ------------------------
+    trainer.fit(model)
+
+
+if __name__ == '__main__':
+
+    # dirs
+    root_dir = os.path.dirname(os.path.realpath(__file__))
+    demo_log_dir = os.path.join(root_dir, 'pt_lightning_demo_logs')
+    checkpoint_dir = os.path.join(demo_log_dir, 'model_weights')
+    test_tube_dir = os.path.join(demo_log_dir, 'test_tube_data')
+
+    # although we user hyperOptParser, we are using it only as argparse right now
+    parent_parser = HyperOptArgumentParser(strategy='grid_search', add_help=False)
+
+    # gpu args
+    parent_parser.add_argument('--gpus', type=str, default='-1', help='how many gpus to use in the node. -1 uses all the gpus on the node')
+    parent_parser.add_argument('--test_tube_save_path', type=str, default=test_tube_dir, help='where to save logs')
+    parent_parser.add_argument('--model_save_path', type=str, default=checkpoint_dir, help='where to save model')
+    parent_parser.add_argument('--experiment_name', type=str, default='pt_lightning_exp_a', help='test tube exp name')
+
+    # allow model to overwrite or extend args
+    parser = LightningTemplateModel.add_model_specific_args(parent_parser, root_dir)
+    hyperparams = parser.parse_args()
+
+    # ---------------------
+    # RUN TRAINING
+    # ---------------------
+    # run on HPC cluster
+    print(f'RUNNING INTERACTIVE MODE ON GPUS. gpu ids: {hyperparams.gpus}')
+    main(hyperparams)

examples/new_project_templates/single_gpu_template.py

@@ -0,0 +1,112 @@
+"""
+Runs a model on a single node across N-gpus.
+"""
+import os
+import sys
+import numpy as np
+from time import sleep
+import torch
+
+from test_tube import HyperOptArgumentParser, Experiment, SlurmCluster
+from pytorch_lightning.models.trainer import Trainer
+from pytorch_lightning.utils.arg_parse import add_default_args
+
+from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint
+
+SEED = 2334
+torch.manual_seed(SEED)
+np.random.seed(SEED)
+
+from lightning_module_template import LightningTemplateModel
+
+
+def main(hparams):
+    """
+    Main training routine specific for this project
+    :param hparams:
+    :return:
+    """
+    # ------------------------
+    # 1 INIT LIGHTNING MODEL
+    # ------------------------
+    print('loading model...')
+    model = LightningTemplateModel(hparams)
+    print('model built')
+
+    # ------------------------
+    # 2 INIT TEST TUBE EXP
+    # ------------------------
+
+    # init experiment
+    exp = Experiment(
+        name=hyperparams.experiment_name,
+        save_dir=hyperparams.test_tube_save_path,
+        autosave=False,
+        description='test demo'
+    )
+
+    exp.argparse(hparams)
+    exp.save()
+
+    # ------------------------
+    # 3 DEFINE CALLBACKS
+    # ------------------------
+    model_save_path = '{}/{}/{}'.format(hparams.model_save_path, exp.name, exp.version)
+    early_stop = EarlyStopping(
+        monitor='val_acc',
+        patience=3,
+        verbose=True,
+        mode='max'
+    )
+
+    checkpoint = ModelCheckpoint(
+        filepath=model_save_path,
+        save_best_only=True,
+        verbose=True,
+        monitor='val_loss',
+        mode='min'
+    )
+
+    # ------------------------
+    # 4 INIT TRAINER
+    # ------------------------
+    trainer = Trainer(
+        experiment=exp,
+        checkpoint_callback=checkpoint,
+        early_stop_callback=early_stop,
+        gpus=hparams.gpus,
+    )
+
+    # ------------------------
+    # 5 START TRAINING
+    # ------------------------
+    trainer.fit(model)
+
+
+if __name__ == '__main__':
+
+    # dirs
+    root_dir = os.path.dirname(os.path.realpath(__file__))
+    demo_log_dir = os.path.join(root_dir, 'pt_lightning_demo_logs')
+    checkpoint_dir = os.path.join(demo_log_dir, 'model_weights')
+    test_tube_dir = os.path.join(demo_log_dir, 'test_tube_data')
+
+    # although we user hyperOptParser, we are using it only as argparse right now
+    parent_parser = HyperOptArgumentParser(strategy='grid_search', add_help=False)
+
+    # gpu args
+    parent_parser.add_argument('--gpus', type=str, default='0', help='how many gpus to use in the node. -1 uses all the gpus on the node')
+    parent_parser.add_argument('--test_tube_save_path', type=str, default=test_tube_dir, help='where to save logs')
+    parent_parser.add_argument('--model_save_path', type=str, default=checkpoint_dir, help='where to save model')
+    parent_parser.add_argument('--experiment_name', type=str, default='pt_lightning_exp_a', help='test tube exp name')
+
+    # allow model to overwrite or extend args
+    parser = LightningTemplateModel.add_model_specific_args(parent_parser, root_dir)
+    hyperparams = parser.parse_args()
+
+    # ---------------------
+    # RUN TRAINING
+    # ---------------------
+    # run on HPC cluster
+    print(f'RUNNING INTERACTIVE MODE ON GPUS. gpu ids: {hyperparams.gpus}')
+    main(hyperparams)

pytorch_lightning/models/trainer.py

@@ -6,6 +6,7 @@ import subprocess
 import traceback
 import warnings
 import os
+import pdb
 
 import torch
 from torch.utils.data.distributed import DistributedSampler
@@ -17,7 +18,7 @@ import tqdm
 
 from pytorch_lightning.root_module.memory import get_gpu_memory_map
 from pytorch_lightning.root_module.model_saving import TrainerIO
-from pytorch_lightning.pt_overrides.override_data_parallel import LightningDistributedDataParallel
+from pytorch_lightning.pt_overrides.override_data_parallel import LightningDistributedDataParallel, LightningDataParallel
 
 
 try:
@@ -27,11 +28,33 @@ except ModuleNotFoundError:
     APEX_AVAILABLE = False
 
 
+def reduce_distributed_output(output, nb_gpus):
+    if nb_gpus <= 1:
+        return output
+
+    # when using DP, we get one output per gpu
+    # average outputs and return
+    if type(output) is torch.Tensor:
+        return output.mean()
+
+    for k, v in output.items():
+        # recurse on nested dics
+        if isinstance(output[k], dict):
+            output[k] = reduce_distributed_output(output[k], nb_gpus)
+
+        # reduce only metrics that have the same nb of gpus
+        elif output[k].size(0) == nb_gpus:
+            reduced = torch.mean(output[k])
+            output[k] = reduced
+    return output
+
+
 class Trainer(TrainerIO):
 
     def __init__(self,
                  experiment,
-                 checkpoint_callback, early_stop_callback,
+                 early_stop_callback=None,
+                 checkpoint_callback=None,
                  gradient_clip=0,
                  cluster=None,
                  process_position=0,
@@ -44,20 +67,57 @@ class Trainer(TrainerIO):
                  check_val_every_n_epoch=1,
                  fast_dev_run=False,
                  accumulate_grad_batches=1,
-                 enable_early_stop=True, max_nb_epochs=1000, min_nb_epochs=1,
+                 max_nb_epochs=1000, min_nb_epochs=1,
-                 train_percent_check=1.0, val_percent_check=1.0, test_percent_check=1.0, val_check_interval=0.95,
+                 train_percent_check=1.0, val_percent_check=1.0, test_percent_check=1.0,
+                 val_check_interval=0.95,
                  log_save_interval=100, add_log_row_interval=10,
                  lr_scheduler_milestones=None,
+                 distributed_backend='dp',
                  use_amp=False,
                  print_nan_grads=False,
+                 print_weights_summary=True,
                  amp_level='O2',
                  nb_sanity_val_steps=5):
 
+        """
+
+        :param experiment: Test-tube experiment
+        :param early_stop_callback: from pytorch_lightning import EarlyStopping
+        :param checkpoint_callback: from pytorch_lightning import Checkpoint
+        :param gradient_clip:
+        :param cluster:
+        :param process_position:
+        :param current_gpu_name:
+        :param nb_gpu_nodes:
+        :param gpus:
+        :param progress_bar:
+        :param overfit_pct:
+        :param track_grad_norm:
+        :param check_val_every_n_epoch:
+        :param fast_dev_run:
+        :param accumulate_grad_batches:
+        :param max_nb_epochs:
+        :param min_nb_epochs:
+        :param train_percent_check:
+        :param val_percent_check:
+        :param test_percent_check:
+        :param val_check_interval:
+        :param log_save_interval:
+        :param add_log_row_interval:
+        :param lr_scheduler_milestones:
+        :param distributed_backend: 'np' to use DistributedParallel, 'ddp' to use DistributedDataParallel
+        :param use_amp:
+        :param print_nan_grads:
+        :param print_weights_summary:
+        :param amp_level:
+        :param nb_sanity_val_steps:
+        """
+
         # Transfer params
         self.nb_gpu_nodes = nb_gpu_nodes
         self.gradient_clip = gradient_clip
         self.check_val_every_n_epoch = check_val_every_n_epoch
-        self.enable_early_stop = enable_early_stop
+        self.enable_early_stop = early_stop_callback is not None
         self.track_grad_norm = track_grad_norm
         self.fast_dev_run = fast_dev_run
         self.on_gpu = gpus is not None and torch.cuda.is_available()
@@ -67,8 +127,12 @@ class Trainer(TrainerIO):
         self.cluster = cluster
         self.process_position = process_position
         self.current_gpu_name = current_gpu_name
+        self.print_weights_summary = print_weights_summary
         self.checkpoint_callback = checkpoint_callback
-        self.checkpoint_callback.save_function = self.save_checkpoint
+
+        if self.checkpoint_callback is not None:
+            self.checkpoint_callback.save_function = self.save_checkpoint
+
         self.early_stop = early_stop_callback
         self.model = None
         self.max_nb_epochs = max_nb_epochs
@@ -82,6 +146,8 @@ class Trainer(TrainerIO):
         self.print_nan_grads = print_nan_grads
         self.data_parallel_device_ids = None
         self.world_size = 1
+        self.use_ddp = False
+        self.use_dp = False
 
         # gpus come in as a string.
         # if gpus = -1 then use all available devices
@@ -95,8 +161,14 @@ class Trainer(TrainerIO):
             # set the correct cuda visible devices (using pci order)
             os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
             os.environ["CUDA_VISIBLE_DEVICES"] = ','.join([str(x) for x in self.data_parallel_device_ids])
+            print(f'VISIBLE GPUS: {os.environ["CUDA_VISIBLE_DEVICES"]}')
 
-        self.data_parallel = self.data_parallel_device_ids is not None and len(self.data_parallel_device_ids) > 1
+        # make DP and DDP mutually exclusive
+        # single GPU will also use DP with devices=[0]
+        have_gpus = self.data_parallel_device_ids is not None and len(self.data_parallel_device_ids) > 0
+        if have_gpus:
+            self.use_dp = distributed_backend == 'dp'
+            self.use_ddp = distributed_backend == 'ddp'
 
         # process info
         self.proc_rank = 0
@@ -137,6 +209,10 @@ class Trainer(TrainerIO):
             '''
             warnings.warn(msg)
 
+    @property
+    def data_parallel(self):
+        return self.use_dp or self.use_ddp
+
     def __determine_data_use_amount(self, train_percent_check, val_percent_check, test_percent_check, overfit_pct):
         """
         Use less data for debugging purposes
@@ -212,9 +288,6 @@ class Trainer(TrainerIO):
         :param max_batches: Scalar
         :return:
         """
-        if self.proc_rank == 0:
-            print('validating...')
-
         # enable eval mode
         model.zero_grad()
         model.eval()
@@ -238,8 +311,12 @@ class Trainer(TrainerIO):
             # -----------------
             # RUN VALIDATION STEP
             # -----------------
-            if self.data_parallel:
+            if self.use_ddp:
                 output = model(data_batch, batch_i)
+            elif self.use_dp:
+                output = model(data_batch, batch_i)
+                output = reduce_distributed_output(output, len(self.data_parallel_device_ids))
+
             else:
                 output = model.validation_step(data_batch, batch_i)
 
@@ -273,7 +350,7 @@ class Trainer(TrainerIO):
         self.test_dataloader = model.test_dataloader
         self.val_dataloader = model.val_dataloader
 
-        if self.data_parallel and type(self.tng_dataloader.sampler) is not DistributedSampler:
+        if self.use_ddp and not isinstance(self.tng_dataloader.sampler, DistributedSampler):
             msg = '''
             when using multiple gpus and multiple nodes you must pass a DistributedSampler to DataLoader(sampler).
             
@@ -293,16 +370,34 @@ class Trainer(TrainerIO):
     # -----------------------------
     def fit(self, model):
 
-        # when using gpus, first thing we do is spawn a new process between each worker
+        # when using multi-node or DDP within a node start each module in a separate process
-        #  multi-gpu and multi-nodes
+        if self.use_ddp:
-        if self.data_parallel:
+            # must copy only the meta of the exp so it survives pickle/unpickle when going to new process
             self.experiment = self.experiment.get_meta_copy()
-            mp.spawn(self.dp_train, nprocs=len(self.data_parallel_device_ids), args=(model, ))
 
-        # treat 1 gpu as a different case to avoid nccl bugs
+            # whenever we have the correct number of tasks, we let slurm manage processes
-        elif len(self.data_parallel_device_ids) == 1:
+            # otherwise we launch the required number of processes
-            self.single_gpu_train(model)
+            nb_slurm_tasks = int(os.environ['SLURM_NTASKS'])
+            nb_requested_gpus = len(self.data_parallel_device_ids)
+            is_slurm_managing_tasks = nb_slurm_tasks == nb_requested_gpus
+            if is_slurm_managing_tasks:
+                task = int(os.environ['SLURM_LOCALID'])
+                self.ddp_train(task, model)
+            else:
+                msg = f"""
+                You requested {nb_requested_gpus} GPUs but launched {nb_slurm_tasks} slurm tasks. 
+                We will launch {nb_requested_gpus} processes for you. 
+                We recommend you let slurm manage the processes by setting: --ntasks-per-node={nb_requested_gpus}
+                """
+                warnings.warn(msg)
+                mp.spawn(self.ddp_train, nprocs=len(self.data_parallel_device_ids), args=(model, ))
 
+        # 1 gpu or dp option triggers training using DP module
+        # easier to avoid NCCL issues
+        elif self.use_dp:
+            self.dp_train(model)
+
+        # ON CPU
         else:
             # CHOOSE OPTIMIZER
             # filter out the weights that were done on gpu so we can load on good old cpus
@@ -318,14 +413,15 @@ class Trainer(TrainerIO):
 
             self.__run_pretrain_routine(model)
 
-    def single_gpu_train(self, model):
+    def dp_train(self, model):
-        # torch.cuda.set_device(0)
-        model.cuda(0)
 
         # CHOOSE OPTIMIZER
         # filter out the weights that were done on gpu so we can load on good old cpus
         self.optimizers = model.configure_optimizers()
 
+        model.cuda(self.data_parallel_device_ids[0])
+        model = LightningDataParallel(model, device_ids=self.data_parallel_device_ids)
+
         # run through amp wrapper
         if self.use_amp:
             # An example
@@ -336,7 +432,7 @@ class Trainer(TrainerIO):
 
         self.__run_pretrain_routine(model)
 
-    def dp_train(self, gpu_nb, model):
+    def ddp_train(self, gpu_nb, model):
         """
         Entry point into a DP thread
         :param gpu_nb:
@@ -443,7 +539,7 @@ class Trainer(TrainerIO):
                 self.lr_schedulers.append(scheduler)
 
         # print model summary
-        if self.proc_rank == 0:
+        if self.proc_rank == 0 and self.print_weights_summary:
             ref_model.summarize()
 
         # give model convenience properties
@@ -538,9 +634,11 @@ class Trainer(TrainerIO):
                     if self.track_grad_norm > 0:
                         model = self.__get_model()
                         grad_norm_dic = model.grad_norm(self.track_grad_norm)
-
                         metrics.update(grad_norm_dic)
 
+                    if self.__is_function_implemented('on_tng_metrics'):
+                        model.on_tng_metrics(metrics)
+
                     # log metrics
                     scalar_metrics = self.__metrics_to_scalars(metrics, blacklist=self.__log_vals_blacklist())
                     if self.proc_rank == 0:
@@ -562,9 +660,9 @@ class Trainer(TrainerIO):
                 model.on_epoch_end()
 
             # early stopping
-            if self.enable_early_stop:
+            met_min_epochs = epoch_nb > self.min_nb_epochs
+            if self.enable_early_stop and met_min_epochs:
                 should_stop = self.early_stop_callback.on_epoch_end(epoch=epoch_nb, logs=self.__tng_tqdm_dic)
-                met_min_epochs = epoch_nb > self.min_nb_epochs
 
                 # stop training
                 stop = should_stop and met_min_epochs
@@ -587,7 +685,7 @@ class Trainer(TrainerIO):
 
     def __log_vals_blacklist(self):
         """avoid logging some vals lightning uses to maintain state"""
-        blacklist = {'batch_nb', 'v_nb', 'epoch', 'gpu'}
+        blacklist = {'batch_nb', 'v_nb', 'gpu'}
         return blacklist
 
     def __run_tng_batch(self, data_batch, batch_nb):
@@ -596,8 +694,8 @@ class Trainer(TrainerIO):
 
         # hook
         if self.__is_function_implemented('on_batch_start'):
-            model = self.__get_model()
+            model_ref = self.__get_model()
-            response = model.on_batch_start(data_batch)
+            response = model_ref.on_batch_start(data_batch)
 
             if response == -1:
                 return -1
@@ -607,8 +705,11 @@ class Trainer(TrainerIO):
 
         # forward pass
         # return a scalar value and a dic with tqdm metrics
-        if self.data_parallel:
+        if self.use_ddp:
             output = self.model(data_batch, batch_nb)
+        elif self.use_dp:
+            output = self.model(data_batch, batch_nb)
+            output = reduce_distributed_output(output, len(self.data_parallel_device_ids))
         else:
             output = self.model.training_step(data_batch, batch_nb)
 
@@ -719,6 +820,6 @@ class Trainer(TrainerIO):
             self.prog_bar.set_postfix(**tqdm_metrics)
 
         # model checkpointing
-        if self.proc_rank == 0:
+        if self.proc_rank == 0 and self.checkpoint_callback:
             print('save callback...')
             self.checkpoint_callback.on_epoch_end(epoch=self.current_epoch, logs=self.__tng_tqdm_dic)

pytorch_lightning/pt_overrides/override_data_parallel.py

@@ -1,6 +1,7 @@
 from torch.nn import DataParallel
 from torch.nn.parallel import DistributedDataParallel
 import itertools
+from itertools import chain
 
 import threading
 import torch
@@ -42,6 +43,29 @@ class LightningDataParallel(DataParallel):
     Override the forward call in lightning so it goes to training and validation step respectively
     """
 
+    def forward(self, *inputs, **kwargs):
+        if not self.device_ids:
+            return self.module(*inputs, **kwargs)
+
+        for t in chain(self.module.parameters(), self.module.buffers()):
+            if t.device != self.src_device_obj:
+                raise RuntimeError("module must have its parameters and buffers "
+                                   "on device {} (device_ids[0]) but found one of "
+                                   "them on device: {}".format(self.src_device_obj, t.device))
+
+        inputs, kwargs = self.scatter(inputs, kwargs, self.device_ids)
+        if len(self.device_ids) == 1:
+            # lightning
+            if self.module.training:
+                return self.module.training_step(*inputs[0], **kwargs[0])
+            else:
+                return self.module.validation_step(*inputs[0], **kwargs[0])
+
+        replicas = self.replicate(self.module, self.device_ids[:len(inputs)])
+        outputs = self.parallel_apply(replicas, inputs, kwargs)
+        return self.gather(outputs, self.output_device)
+
+
     def parallel_apply(self, replicas, inputs, kwargs):
         return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
 

pytorch_lightning/root_module/hooks.py

@@ -19,3 +19,6 @@ class ModelHooks(torch.nn.Module):
     def on_post_performance_check(self):
         pass
 
+    def on_tng_metrics(self, metrics):
+        pass
+

pytorch_lightning/root_module/model_saving.py

@@ -2,7 +2,7 @@ import torch
 import os
 import re
 import pdb
-from pytorch_lightning.pt_overrides.override_data_parallel import LightningDistributedDataParallel
+from pytorch_lightning.pt_overrides.override_data_parallel import LightningDistributedDataParallel, LightningDataParallel
 
 class ModelIO(object):
 
@@ -66,7 +66,8 @@ class TrainerIO(object):
         checkpoint['optimizer_states'] = optimizer_states
 
         # request what to save from the model
-        model = self.model.module if type(self.model) is LightningDistributedDataParallel else self.model
+        is_dp_module = type(self.model) is LightningDistributedDataParallel or type(self.model) is LightningDataParallel
+        model = self.model.module if is_dp_module else self.model
         checkpoint_dict = model.get_save_dict()
 
         # merge trainer and model saving items

pytorch_lightning/root_module/root_module.py

@@ -78,7 +78,7 @@ class LightningModule(GradInformation, ModelIO, OptimizerConfig, ModelHooks):
         :param logs:
         :return:
         """
-        raise NotImplementedError
+        return logs
 
     def loss(self, *args, **kwargs):
         """

setup.py

@@ -7,7 +7,7 @@ from setuptools import setup, find_packages
 # http://blog.ionelmc.ro/2014/05/25/python-packaging/
 setup(
     name="pytorch-lightning",
-    version='0.2.3',
+    version='0.2.5.1',
     description="The Keras for ML researchers using PyTorch",
     author="William Falcon",
     author_email="waf2107@columbia.edu",
@@ -19,8 +19,7 @@ setup(
     install_requires=[
         "torch>=1.1.0",
         "tqdm",
-        "test-tube>=0.6.6",
+        "test-tube>=0.6.7.1",
-        "tensorflow>=1.14.0"
     ],
     packages=find_packages(),
     long_description=open("README.md", encoding="utf-8").read(),