Skip to content

PyTorch implementation of Value Iteration Networks (VIN): Clean, Simple and Modular. Visualization in Visdom.

Notifications You must be signed in to change notification settings

tensorpro/VIN_PyTorch_Visdom

 
 

Repository files navigation

This is an implementation of Value Iteration Networks (VIN) in PyTorch to reproduce the results.(TensorFlow version)

Architecture of Value Iteration Network

Key idea

  • A fully differentiable neural network with a 'planning' sub-module.
  • Value Iteration = Conv Layer + Channel-wise Max Pooling
  • Generalize better than reactive policies for new, unseen tasks.

Learned Reward Image and Its Value Images for each VI Iteration

Visualization Grid world Reward Image Value Images
8x8
16x16
28x28

Dependencies

This repository requires following packages:

  • Python >= 3.6
  • Numpy >= 1.12.1
  • PyTorch >= 0.1.10
  • SciPy >= 0.19.0
  • visdom >= 0.1

Datasets

Each data sample consists of (x, y) coordinates of current state in grid world, followed by an obstacle image and a goal image.

Dataset size 8x8 16x16 28x28
Train set 77760 776440 4510695
Test set 12960 129440 751905

Running Experiment: Training

Grid world 8x8

python run.py --datafile data/gridworld_8x8.npz --imsize 8 --lr 0.005 --epochs 30 --k 10 --batch_size 128

Grid world 16x16

python run.py --datafile data/gridworld_16x16.npz --imsize 16 --lr 0.008 --epochs 30 --k 20 --batch_size 128

Grid world 28x28

python run.py --datafile data/gridworld_28x28.npz --imsize 28 --lr 0.003 --epochs 30 --k 36 --batch_size 128

Flags:

  • datafile: The path to the data files.
  • imsize: The size of input images. From: [8, 16, 28]
  • lr: Learning rate with RMSProp optimizer. Recommended: [0.01, 0.005, 0.002, 0.001]
  • epochs: Number of epochs to train. Default: 30
  • k: Number of Value Iterations. Recommended: [10 for 8x8, 20 for 16x16, 36 for 28x28]
  • ch_i: Number of channels in input layer. Default: 2, i.e. obstacles image and goal image.
  • ch_h: Number of channels in first convolutional layer. Default: 150, described in paper.
  • ch_q: Number of channels in q layer (~actions) in VI-module. Default: 10, described in paper.
  • batch_size: Batch size. Default: 128

Visualization with Visdom

We shall visualize the learned reward image and its corresponding value images for each VI iteration by using visdom.

Firstly start the server

python -m visdom.server

Open Visdom in browser in http://localhost:8097

Then run following to visualize learn reward and value images.

python vis.py --datafile learned_rewards_values_28x28.npz

NOTE: If you would like to produce GIF animation of value images on your own, the following command might be useful.

convert -delay 20 -loop 0 *.png value_function.gif

Benchmarks

GPU: TITAN X

Performance: Test Accuracy

NOTE: This is the accuracy on test set. It is different from the table in the paper, which indicates the success rate from rollouts of the learned policy in the environment.

Test Accuracy 8x8 16x16 28x28
PyTorch 99.16% 92.44% 88.20%
TensorFlow 99.03% 90.2% 82%

Speed with GPU

Speed per epoch 8x8 16x16 28x28
PyTorch 3s 15s 100s
TensorFlow 4s 25s 165s

Frequently Asked Questions

  • Q: How to get reward image from observation ?

    • A: Observation image has 2 channels. First channel is obstacle image (0: free, 1: obstacle). Second channel is goal image (0: free, 10: goal). For example, in 8x8 grid world, the shape of an input tensor with batch size 128 is [128, 2, 8, 8]. Then it is fed into a convolutional layer with [3, 3] filter and 150 feature maps, followed by another convolutional layer with [3, 3] filter and 1 feature map. The shape of the output tensor is [128, 1, 8, 8]. This is the reward image.
  • Q: What is exactly transition model, and how to obtain value image by VI-module from reward image ?

    • A: Let us assume batch size is 128 under 8x8 grid world. Once we obtain the reward image with shape [128, 1, 8, 8], we do convolutional layer for q layers in VI module. The [3, 3] filter represents the transition probabilities. There is a set of 10 filters, each for generating a feature map in q layers. Each feature map corresponds to an "action". Note that this is larger than real available actions which is only 8. Then we do a channel-wise Max Pooling to obtain the value image with shape [128, 1, 8, 8]. Finally we stack this value image with reward image for a new VI iteration.

References

Further Readings

About

PyTorch implementation of Value Iteration Networks (VIN): Clean, Simple and Modular. Visualization in Visdom.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%