simplify examples structure

CHANGELOG.md

@@ -8,7 +8,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
-- Added parity test between a  vanilla MNIST model  and  lightning model ([#1284](https://github.com/PyTorchLightning/pytorch-lightning/pull/1284))
+- Added parity test between a vanilla MNIST model and lightning model ([#1284](https://github.com/PyTorchLightning/pytorch-lightning/pull/1284))
+- Added parity test between a vanilla RNN model and lightning model ([#1351](https://github.com/PyTorchLightning/pytorch-lightning/pull/1351))
 - Added Reinforcement Learning - Deep Q-network (DQN) lightning example ([#1232](https://github.com/PyTorchLightning/pytorch-lightning/pull/1232))
 - Added support for hierarchical `dict` ([#1152](https://github.com/PyTorchLightning/pytorch-lightning/pull/1152))
 - Added `TrainsLogger` class ([#1122](https://github.com/PyTorchLightning/pytorch-lightning/pull/1122))
@@ -41,6 +42,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Give warnings for unimplemented required lightning methods ([#1317](https://github.com/PyTorchLightning/pytorch-lightning/pull/1317))
 - Enhanced load_from_checkpoint to also forward params to the model ([#1307](https://github.com/PyTorchLightning/pytorch-lightning/pull/1307))
 - Made `evaluate` method private >> `Trainer._evaluate(...)`. ([#1260](https://github.com/PyTorchLightning/pytorch-lightning/pull/1260))
+- Simplify the PL examples structure (shallower and more readable) ([#1247](https://github.com/PyTorchLightning/pytorch-lightning/pull/1247))
 
 ### Deprecated
 

pl_examples/README.md

@@ -1,14 +1,67 @@
 # Examples   
-This folder has 4 sections:   
+This folder has 3 sections:   
 
-### Basic examples   
-These show the most common use of Lightning for either CPU or GPU training.   
+## Basic Examples   
+Use these examples to test how lightning works.   
 
-### Domain templates   
-These are templates to show common approaches such as GANs and RL.   
+#### Test on CPU  
+```bash
+python cpu_template.py
+```
 
-### Full examples 
-Contains examples demonstrating ImageNet training, Semantic Segmentation, etc.
+---   
+#### Train on a single GPU
+```bash
+python gpu_template.py --gpus 1
+```   
 
-### Multi-node examples   
-These show how to run jobs on a GPU cluster using lightning.
+---    
+#### DataParallel (dp)   
+Train on multiple GPUs using DataParallel.
+
+```bash
+python gpu_template.py --gpus 2 --distributed_backend dp
+```   
+
+---
+#### DistributedDataParallel (ddp)    
+
+Train on multiple GPUs using DistributedDataParallel   
+```bash
+python gpu_template.py --gpus 2 --distributed_backend ddp
+```
+
+---
+#### DistributedDataParallel+DP (ddp2)    
+
+Train on multiple GPUs using DistributedDataParallel + dataparallel.
+On a single node, uses all GPUs for 1 model. Then shares gradient information
+across nodes.   
+```bash
+python gpu_template.py --gpus 2 --distributed_backend ddp2
+```
+
+## Multi-node example   
+
+This demo launches a job using 2 GPUs on 2 different nodes (4 GPUs total).
+To run this demo do the following:
+
+1. Log into the jumphost node of your SLURM-managed cluster.  
+2. Create a conda environment with Lightning and a GPU PyTorch version.   
+3. Choose a script to submit    
+
+### DDP  
+Submit this job to run with DistributedDataParallel (2 nodes, 2 gpus each)
+```bash
+sbatch ddp_job_submit.sh YourEnv
+```
+
+### DDP2  
+Submit this job to run with a different implementation of DistributedDataParallel.
+In this version, each node acts like DataParallel but syncs across nodes like DDP.
+```bash
+sbatch ddp2_job_submit.sh YourEnv
+```
+
+## Domain templates   
+These are templates to show common approaches such as GANs and RL.

pl_examples/__init__.py

@@ -140,7 +140,7 @@ def optimize_on_cluster(hyperparams):
 
 """
 
-from .basic_examples.lightning_module_template import LightningTemplateModel
+from pl_examples.models.lightning_template import LightningTemplateModel
 
 __all__ = [
     'LightningTemplateModel'

pl_examples/basic_examples/README.md

@@ -1,4 +1,4 @@
-# Basic Examples   
+## Basic Examples   
 Use these examples to test how lightning works.   
 
 #### Test on CPU  
@@ -36,4 +36,27 @@ On a single node, uses all GPUs for 1 model. Then shares gradient information
 across nodes.   
 ```bash
 python gpu_template.py --gpus 2 --distributed_backend ddp2
-```
+```
+
+
+# Multi-node example   
+
+This demo launches a job using 2 GPUs on 2 different nodes (4 GPUs total).
+To run this demo do the following:
+
+1. Log into the jumphost node of your SLURM-managed cluster.  
+2. Create a conda environment with Lightning and a GPU PyTorch version.   
+3. Choose a script to submit    
+
+#### DDP  
+Submit this job to run with DistributedDataParallel (2 nodes, 2 gpus each)
+```bash
+sbatch ddp_job_submit.sh YourEnv
+```
+
+#### DDP2  
+Submit this job to run with a different implementation of DistributedDataParallel.
+In this version, each node acts like DataParallel but syncs across nodes like DDP.
+```bash
+sbatch ddp2_job_submit.sh YourEnv
+```

pl_examples/basic_examples/cpu_template.py

@@ -8,7 +8,7 @@
 import torch
 
 import pytorch_lightning as pl
-from pl_examples.basic_examples.lightning_module_template import LightningTemplateModel
+from pl_examples.models.lightning_template import LightningTemplateModel
 
 SEED = 2334
 torch.manual_seed(SEED)

pl_examples/basic_examples/gpu_template.py

@@ -8,7 +8,7 @@
 import torch
 
 import pytorch_lightning as pl
-from pl_examples.basic_examples.lightning_module_template import LightningTemplateModel
+from pl_examples.models.lightning_template import LightningTemplateModel
 
 SEED = 2334
 torch.manual_seed(SEED)

pl_examples/multi_node_examples/multi_node_ddp2_demo.py → pl_examples/basic_examples/multi_node_ddp2_demo.py

@@ -8,7 +8,7 @@
 import torch
 
 import pytorch_lightning as pl
-from pl_examples.basic_examples.lightning_module_template import LightningTemplateModel
+from pl_examples.models.lightning_template import LightningTemplateModel
 
 SEED = 2334
 torch.manual_seed(SEED)

pl_examples/multi_node_examples/multi_node_ddp_demo.py → pl_examples/basic_examples/multi_node_ddp_demo.py

@@ -8,7 +8,7 @@
 import torch
 
 import pytorch_lightning as pl
-from pl_examples.basic_examples.lightning_module_template import LightningTemplateModel
+from pl_examples.models.lightning_template import LightningTemplateModel
 
 SEED = 2334
 torch.manual_seed(SEED)

pl_examples/domain_templates/gan.py → pl_examples/domain_templates/generative_adversarial_net.py

@@ -1,6 +1,6 @@
 """
 To run this template just do:
-python gan.py
+python generative_adversarial_net.py
 
 After a few epochs, launch TensorBoard to see the images being generated at every batch:
 

pl_examples/full_examples/semantic_segmentation/semseg.py → pl_examples/domain_templates/semantic_segmentation.py

@@ -6,10 +6,10 @@
 import torch.nn.functional as F
 import torchvision.transforms as transforms
 from PIL import Image
-from models.unet.model import UNet
 from torch.utils.data import DataLoader, Dataset
 
 import pytorch_lightning as pl
+from pl_examples.models.unet import UNet
 
 
 class KITTI(Dataset):

pl_examples/full_examples/semantic_segmentation/models/unet/parts.py → pl_examples/models/unet.py

@@ -3,6 +3,47 @@
 import torch.nn.functional as F
 
 
+class UNet(nn.Module):
+    """
+    Architecture based on U-Net: Convolutional Networks for Biomedical Image Segmentation
+    Link - https://arxiv.org/abs/1505.04597
+
+    Parameters:
+        num_classes (int): Number of output classes required (default 19 for KITTI dataset)
+        bilinear (bool): Whether to use bilinear interpolation or transposed
+            convolutions for upsampling.
+    """
+
+    def __init__(self, num_classes=19, bilinear=False):
+        super().__init__()
+        self.layer1 = DoubleConv(3, 64)
+        self.layer2 = Down(64, 128)
+        self.layer3 = Down(128, 256)
+        self.layer4 = Down(256, 512)
+        self.layer5 = Down(512, 1024)
+
+        self.layer6 = Up(1024, 512, bilinear=bilinear)
+        self.layer7 = Up(512, 256, bilinear=bilinear)
+        self.layer8 = Up(256, 128, bilinear=bilinear)
+        self.layer9 = Up(128, 64, bilinear=bilinear)
+
+        self.layer10 = nn.Conv2d(64, num_classes, kernel_size=1)
+
+    def forward(self, x):
+        x1 = self.layer1(x)
+        x2 = self.layer2(x1)
+        x3 = self.layer3(x2)
+        x4 = self.layer4(x3)
+        x5 = self.layer5(x4)
+
+        x6 = self.layer6(x5, x4)
+        x6 = self.layer7(x6, x3)
+        x6 = self.layer8(x6, x2)
+        x6 = self.layer9(x6, x1)
+
+        return self.layer10(x6)
+
+
 class DoubleConv(nn.Module):
     """
     Double Convolution and BN and ReLU