piyushghai commented on a change in pull request #14462: [MXNET-1358][Fit API] Fit api tutorial URL: https://github.com/apache/incubator-mxnet/pull/14462#discussion_r272693646
########## File path: docs/tutorials/gluon/fit_api_tutorial.md ########## @@ -0,0 +1,208 @@ +<!--- Licensed to the Apache Software Foundation (ASF) under one --> +<!--- or more contributor license agreements. See the NOTICE file --> +<!--- distributed with this work for additional information --> +<!--- regarding copyright ownership. The ASF licenses this file --> +<!--- to you under the Apache License, Version 2.0 (the --> +<!--- "License"); you may not use this file except in compliance --> +<!--- with the License. You may obtain a copy of the License at --> + +<!--- http://www.apache.org/licenses/LICENSE-2.0 --> + +<!--- Unless required by applicable law or agreed to in writing, --> +<!--- software distributed under the License is distributed on an --> +<!--- "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY --> +<!--- KIND, either express or implied. See the License for the --> +<!--- specific language governing permissions and limitations --> +<!--- under the License. --> + + +# Gluon Fit API + +In this tutorial, we will see how to use the [Gluon Fit API](https://cwiki.apache.org/confluence/display/MXNET/Gluon+Fit+API+-+Tech+Design) which is a simple and flexible way to train deep learning models using the [Gluon APIs](http://mxnet.incubator.apache.org/versions/master/gluon/index.html) in Apache MXNet. + +Prior to Fit API, training using Gluon required one to write a custom ["Gluon training loop"](https://mxnet.incubator.apache.org/versions/master/tutorials/gluon/logistic_regression_explained.html#defining-and-training-the-model). Fit API reduces the complexity and amount of boiler plate code required to train a model, provides an easy to use and a powerful API. + +To demonstrate the Fit API, this tutorial will train an Image Classification model using the [ResNet-18](https://arxiv.org/abs/1512.03385) architecture for the neural network. The model will be trained using the [Fashion-MNIST dataset](https://research.zalando.com/welcome/mission/research-projects/fashion-mnist/). + + +## Prerequisites + +To complete this tutorial, you will need: + +- [MXNet](https://mxnet.incubator.apache.org/install/#overview) (The version of MXNet will be >= 1.5.0) +- [Jupyter Notebook](https://jupyter.org/index.html) (For interactively running the provided .ipynb file) + + + + +```python +import mxnet as mx +from mxnet import gluon +from mxnet.gluon.model_zoo import vision +from mxnet.gluon.estimator import estimator, event_handler + +ctx = mx.gpu() if mx.context.num_gpus() > 0 else mx.cpu() +mx.random.seed(7) # Set a fixed seed +``` + +## Dataset + +[Fashion-MNIST](https://research.zalando.com/welcome/mission/research-projects/fashion-mnist/) dataset consists of fashion items divided into ten categories : t-shirt/top, trouser, pullover, dress, coat, sandal, shirt, sneaker, bag and ankle boot. + +- It has 60,000 gray scale images of size 28 * 28 for training. +- It has 10,000 gray scale images os size 28 * 28 for testing/validation. + +We will use ```gluon.data.vision``` package to directly import the Fashion-MNIST dataset and perform pre-processing on it. + + +```python +# Get the training data +fashion_mnist_train = gluon.data.vision.FashionMNIST(train=True) + +# Get the validation data +fashion_mnist_val = gluon.data.vision.FashionMNIST(train=False) +``` + + +```python +transforms = [gluon.data.vision.transforms.Resize(224), # We pick 224 as the model we use takes an input of size 224. + gluon.data.vision.transforms.ToTensor()] + +# Now we will stack all these together. +transforms = gluon.data.vision.transforms.Compose(transforms) +``` + + +```python +# Apply the transformations +fashion_mnist_train = fashion_mnist_train.transform_first(transforms) +fashion_mnist_val = fashion_mnist_val.transform_first(transforms) +``` + + +```python +batch_size = 256 # Batch size of the images +num_workers = 4 # The number of parallel workers for loading the data using Data Loaders. + +train_data_loader = gluon.data.DataLoader(fashion_mnist_train, batch_size=batch_size, shuffle=True, num_workers=num_workers) +val_data_loader = gluon.data.DataLoader(fashion_mnist_val, batch_size=batch_size, shuffle=False, num_workers=num_workers) +``` + +## Model and Optimizers + +Let's load the resnet-18 model architecture from [Gluon Model Zoo](http://mxnet.apache.org/api/python/gluon/model_zoo.html) and initialize it's parameters. + + +```python +resnet_18_v1 = vision.resnet18_v1(pretrained=False, classes = 10, ctx=ctx) +resnet_18_v1.initialize(force_reinit=True, init = mx.init.Xavier(), ctx=ctx) +``` + +After defining the model, let's setup the trainer object for training. Review comment: Moved the line about the trainer object to below. I'm not sure I quite understand your first point here about avoiding to create a trainer. There's a minimal bunch of parameters I passed to the trainer, thus keeping it barebones. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services