зеркало из https://github.com/microsoft/pai.git
45 строки
1.9 KiB
YAML
45 строки
1.9 KiB
YAML
protocolVersion: 2
|
|
name: chainer_cifar
|
|
type: job
|
|
version: 1.0
|
|
contributor: OpenPAI
|
|
description: |
|
|
# Chainer CIFAR Image Classification Example
|
|
This is an example of a convolutional neural network (convnet) applied to an image classification task using the CIFAR-10 or CIFAR-100 dataset on OpenPAI.
|
|
The CIFAR datasets can be a good choice for initial experiments with convnets because the size and
|
|
number of images is small enough to allow typical models to be trained in a reasonable amount of time.
|
|
However, the classification task is still challenging because natural images are used.
|
|
|
|
Specifically, there are 50000 color training images of size 32x32 pixels with either 10 class labels (for CIFAR-10) or 100 class labels (for CIFAR-100).
|
|
For CIFAR-10, state of the art methods without data augmentation can achieve similar to human-level classification accuracy of around 94%.
|
|
For CIFAR-100, state of the art without data augmentation is around 20% (DenseNet).
|
|
|
|
If you want to run this example on the N-th GPU, pass `--gpu=N` to the script. To run on CPU, pass `--gpu=-1`.
|
|
For example, to run the default model, which uses CIFAR-10 and GPU 0, `python train_cifar.py`;
|
|
to run the CIFAR-100 dataset on GPU 1, `python train_cifar.py --gpu=1 --dataset='cifar100'`.
|
|
|
|
Reference, https://github.com/chainer/chainer/tree/master/examples/cifar
|
|
|
|
prerequisites:
|
|
- protocolVersion: 2
|
|
name: chainer_example
|
|
type: dockerimage
|
|
version: 1.0
|
|
contributor : OpenPAI
|
|
description: |
|
|
This is an [example chainer Docker image on OpenPAI](https://github.com/Microsoft/pai/tree/master/examples/chainer).
|
|
uri : openpai/pai.example.chainer
|
|
|
|
taskRoles:
|
|
train:
|
|
instances: 1
|
|
completion:
|
|
minSucceededInstances: 1
|
|
dockerImage: chainer_example
|
|
resourcePerInstance:
|
|
cpu: 4
|
|
memoryMB: 8192
|
|
gpu: 1
|
|
commands:
|
|
- python ./chainer/examples/cifar/train_cifar.py
|