Lawrence  Lesch

Lawrence Lesch

1632386640

Programmatically Get Javascript Objects Key Names Using Proxy API

I find myself frequently iterating over some Javascript object keys and then performing specific actions which are relevant for some key. For example, imagine there’s a website which aggregates various ratings of TV shows. The website wants to display all ratings of a particular show but the font size of IMDB rating should be 24 while all other ratings font size should be 16. Let’s say the website uses React.js as their frontend framework

#javascript 

What is GEEK

Buddha Community

Programmatically Get Javascript Objects Key Names Using Proxy API
Sasha  Lee

Sasha Lee

1650636000

Dl4clj: Clojure Wrapper for Deeplearning4j.

dl4clj

Port of deeplearning4j to clojure

Contact info

If you have any questions,

  • my email is will@yetanalytics.com
  • I'm will_hoyt in the clojurians slack
  • twitter is @FeLungz (don't check very often)

TODO

  • update examples dir
  • finish README
    • add in examples using Transfer Learning
  • finish tests
    • eval is missing regression tests, roc tests
    • nn-test is missing regression tests
    • spark tests need to be redone
    • need dl4clj.core tests
  • revist spark for updates
  • write specs for user facing functions
    • this is very important, match isnt strict for maps
    • provides 100% certianty of the input -> output flow
    • check the args as they come in, dispatch once I know its safe, test the pure output
  • collapse overlapping api namespaces
  • add to core use case flows

Features

Stable Features with tests

  • Neural Networks DSL
  • Early Stopping Training
  • Transfer Learning
  • Evaluation
  • Data import

Features being worked on for 0.1.0

  • Clustering (testing in progress)
  • Spark (currently being refactored)
  • Front End (maybe current release, maybe future release. Not sure yet)
  • Version of dl4j is 0.0.8 in this project. Current dl4j version is 0.0.9
  • Parallelism
  • Kafka support
  • Other items mentioned in TODO

Features being worked on for future releases

  • NLP
  • Computational Graphs
  • Reinforement Learning
  • Arbiter

Artifacts

NOT YET RELEASED TO CLOJARS

  • fork or clone to try it out

If using Maven add the following repository definition to your pom.xml:

<repository>
  <id>clojars.org</id>
  <url>http://clojars.org/repo</url>
</repository>

Latest release

With Leiningen:

n/a

With Maven:

n/a

<dependency>
  <groupId>_</groupId>
  <artifactId>_</artifactId>
  <version>_</version>
</dependency>

Usage

Things you need to know

All functions for creating dl4j objects return code by default

  • All of these functions have an option to return the dl4j object
    • :as-code? = false
  • This because all builders require the code representation of dl4j objects
    • this requirement is not going to change
  • INDarray creation fns default to objects, this is for convenience
    • :as-code? is still respected

API functions return code when all args are provided as code

API functions return the value of calling the wrapped method when args are provided as a mixture of objects and code or just objects

The tests are there to help clarify behavior, if you are unsure of how to use a fn, search the tests

  • for questions about spark, refer to the spark section bellow

Example of obj/code duality

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]))

;; as code (the default)

(l/dense-layer-builder
 :activation-fn :relu
 :learning-rate 0.006
 :weight-init :xavier
 :layer-name "example layer"
 :n-in 10
 :n-out 1)

;; =>

(doto
 (org.deeplearning4j.nn.conf.layers.DenseLayer$Builder.)
 (.nOut 1)
 (.activation (dl4clj.constants/value-of {:activation-fn :relu}))
 (.weightInit (dl4clj.constants/value-of {:weight-init :xavier}))
 (.nIn 10)
 (.name "example layer")
 (.learningRate 0.006))

;; as an object

(l/dense-layer-builder
 :activation-fn :relu
 :learning-rate 0.006
 :weight-init :xavier
 :layer-name "example layer"
 :n-in 10
 :n-out 1
 :as-code? false)

;; =>

#object[org.deeplearning4j.nn.conf.layers.DenseLayer 0x69d7d160 "DenseLayer(super=FeedForwardLayer(super=Layer(layerName=example layer, activationFn=relu, weightInit=XAVIER, biasInit=NaN, dist=null, learningRate=0.006, biasLearningRate=NaN, learningRateSchedule=null, momentum=NaN, momentumSchedule=null, l1=NaN, l2=NaN, l1Bias=NaN, l2Bias=NaN, dropOut=NaN, updater=null, rho=NaN, epsilon=NaN, rmsDecay=NaN, adamMeanDecay=NaN, adamVarDecay=NaN, gradientNormalization=null, gradientNormalizationThreshold=NaN), nIn=10, nOut=1))"]

General usage examples

Importing data

Loading data from a file (here its a csv)


(ns my.ns
 (:require [dl4clj.datasets.input-splits :as s]
           [dl4clj.datasets.record-readers :as rr]
           [dl4clj.datasets.api.record-readers :refer :all]
           [dl4clj.datasets.iterators :as ds-iter]
           [dl4clj.datasets.api.iterators :refer :all]
           [dl4clj.helpers :refer [data-from-iter]]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; file splits (convert the data to records)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def poker-path "resources/poker-hand-training.csv")
;; this is not a complete dataset, it is just here to sever as an example

(def file-split (s/new-filesplit :path poker-path))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; record readers, (read the records created by the file split)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def csv-rr (initialize-rr! :rr (rr/new-csv-record-reader :skip-n-lines 0 :delimiter ",")
                                 :input-split file-split))

;; lets look at some data
(println (next-record! :rr csv-rr :as-code? false))
;; => #object[java.util.ArrayList 0x2473e02d [1, 10, 1, 11, 1, 13, 1, 12, 1, 1, 9]]
;; this is our first line from the csv


;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; record readers dataset iterators (turn our writables into a dataset)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def rr-ds-iter (ds-iter/new-record-reader-dataset-iterator
                 :record-reader csv-rr
                 :batch-size 1
                 :label-idx 10
                 :n-possible-labels 10))

;; we use our record reader created above
;; we want to see one example per dataset obj returned (:batch-size = 1)
;; we know our label is at the last index, so :label-idx = 10
;; there are 10 possible types of poker hands so :n-possible-labels = 10
;; you can also set :label-idx to -1 to use the last index no matter the size of the seq

(def other-rr-ds-iter (ds-iter/new-record-reader-dataset-iterator
                       :record-reader csv-rr
                       :batch-size 1
                       :label-idx -1
                       :n-possible-labels 10))

(str (next-example! :iter rr-ds-iter :as-code? false))
;; =>
;;===========INPUT===================
;;[1.00, 10.00, 1.00, 11.00, 1.00, 13.00, 1.00, 12.00, 1.00, 1.00]
;;=================OUTPUT==================
;;[0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 1.00]


;; and to show that :label-idx = -1 gives us the same output

(= (next-example! :iter rr-ds-iter :as-code? false)
   (next-example! :iter other-rr-ds-iter :as-code? false)) ;; => true

INDArrays and Datasets from clojure data structures


(ns my.ns
  (:require [nd4clj.linalg.factory.nd4j :refer [vec->indarray matrix->indarray
                                                indarray-of-zeros indarray-of-ones
                                                indarray-of-rand vec-or-matrix->indarray]]
            [dl4clj.datasets.new-datasets :refer [new-ds]]
            [dl4clj.datasets.api.datasets :refer [as-list]]
            [dl4clj.datasets.iterators :refer [new-existing-dataset-iterator]]
            [dl4clj.datasets.api.iterators :refer :all]
            [dl4clj.datasets.pre-processors :as ds-pp]
            [dl4clj.datasets.api.pre-processors :refer :all]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; INDArray creation
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;;TODO: consider defaulting to code

;; can create from a vector

(vec->indarray [1 2 3 4])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x269df212 [1.00, 2.00, 3.00, 4.00]]

;; or from a matrix

(matrix->indarray [[1 2 3 4] [2 4 6 8]])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x20aa7fe1
;; [[1.00, 2.00, 3.00, 4.00], [2.00, 4.00, 6.00, 8.00]]]


;; will fill in spareness with zeros

(matrix->indarray [[1 2 3 4] [2 4 6 8] [10 12]])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x8b7796c
;;[[1.00, 2.00, 3.00, 4.00],
;; [2.00, 4.00, 6.00, 8.00],
;; [10.00, 12.00, 0.00, 0.00]]]

;; can create an indarray of all zeros with specified shape
;; defaults to :rows = 1 :columns = 1

(indarray-of-zeros :rows 3 :columns 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x6f586a7e
;;[[0.00, 0.00],
;; [0.00, 0.00],
;; [0.00, 0.00]]]

(indarray-of-zeros) ;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0xe59ffec 0.00]

;; and if only one is supplied, will get a vector of specified length

(indarray-of-zeros :rows 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x2899d974 [0.00, 0.00]]

(indarray-of-zeros :columns 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0xa5b9782 [0.00, 0.00]]

;; same considerations/defaults for indarray-of-ones and indarray-of-rand

(indarray-of-ones :rows 2 :columns 3)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x54f08662 [[1.00, 1.00, 1.00], [1.00, 1.00, 1.00]]]

(indarray-of-rand :rows 2 :columns 3)
;; all values are greater than 0 but less than 1
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x2f20293b [[0.85, 0.86, 0.13], [0.94, 0.04, 0.36]]]



;; vec-or-matrix->indarray is built into all functions which require INDArrays
;; so that you can use clojure data structures
;; but you still have the option of passing existing INDArrays

(def example-array (vec-or-matrix->indarray [1 2 3 4]))
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x5c44c71f [1.00, 2.00, 3.00, 4.00]]

(vec-or-matrix->indarray example-array)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x607b03b0 [1.00, 2.00, 3.00, 4.00]]

(vec-or-matrix->indarray (indarray-of-rand :rows 2))
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x49143b08 [0.76, 0.92]]

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; data-set creation
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def ds-with-single-example (new-ds :input [1 2 3 4]
                                    :output [0.0 1.0 0.0]))

(as-list :ds ds-with-single-example :as-code? false)
;; =>
;; #object[java.util.ArrayList 0x5d703d12
;;[===========INPUT===================
;;[1.00, 2.00, 3.00, 4.00]
;;=================OUTPUT==================
;;[0.00, 1.00, 0.00]]]

(def ds-with-multiple-examples (new-ds
                                :input [[1 2 3 4] [2 4 6 8]]
                                :output [[0.0 1.0 0.0] [0.0 0.0 1.0]]))

(as-list :ds ds-with-multiple-examples :as-code? false)
;; =>
;;#object[java.util.ArrayList 0x29c7a9e2
;;[===========INPUT===================
;;[1.00, 2.00, 3.00, 4.00]
;;=================OUTPUT==================
;;[0.00, 1.00, 0.00],
;;===========INPUT===================
;;[2.00, 4.00, 6.00, 8.00]
;;=================OUTPUT==================
;;[0.00, 0.00, 1.00]]]

;; we can create a dataset iterator from the code which creates datasets
;; and set the labels for our outputs (optional)

(def ds-with-multiple-examples
  (new-ds
   :input [[1 2 3 4] [2 4 6 8]]
   :output [[0.0 1.0 0.0] [0.0 0.0 1.0]]))

;; iterator
(def training-rr-ds-iter
  (new-existing-dataset-iterator
   :dataset ds-with-multiple-examples
   :labels ["foo" "baz" "foobaz"]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; data-set normalization
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; this gathers statistics on the dataset and normalizes the data
;; and applies the transformation to all dataset objects in the iterator
(def train-iter-normalized
  (c/normalize-iter! :iter training-rr-ds-iter
                     :normalizer (ds-pp/new-standardize-normalization-ds-preprocessor)
                     :as-code? false))

;; above returns the normalized iterator
;; to get fit normalizer

(def the-normalizer
  (get-pre-processor train-iter-normalized))

Model configuration

Creating a neural network configuration with singe and multiple layers

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.conf.distributions :as dist]
            [dl4clj.nn.conf.input-pre-processor :as pp]
            [dl4clj.nn.conf.step-fns :as s-fn]))

;; nn/builder has 3 types of args
;; 1) args which set network configuration params
;; 2) args which set default values for layers
;; 3) args which set multi layer network configuration params

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; single layer nn configuration
;; here we are setting network configuration
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(nn/builder :optimization-algo :stochastic-gradient-descent
            :seed 123
            :iterations 1
            :minimize? true
            :use-drop-connect? false
            :lr-score-based-decay-rate 0.002
            :regularization? false
            :step-fn :default-step-fn
            :layers {:dense-layer {:activation-fn :relu
                                   :updater :adam
                                   :adam-mean-decay 0.2
                                   :adam-var-decay 0.1
                                   :learning-rate 0.006
                                   :weight-init :xavier
                                   :layer-name "single layer model example"
                                   :n-in 10
                                   :n-out 20}})

;; there are several options within a nn-conf map which can be configuration maps
;; or calls to fns
;; It doesn't matter which option you choose and you don't have to stay consistent
;; the list of params which can be passed as config maps or fn calls will
;; be enumerated at a later date

(nn/builder :optimization-algo :stochastic-gradient-descent
            :seed 123
            :iterations 1
            :minimize? true
            :use-drop-connect? false
            :lr-score-based-decay-rate 0.002
            :regularization? false
            :step-fn (s-fn/new-default-step-fn)
            :build? true
            ;; dont need to specify layer order, theres only one
            :layers (l/dense-layer-builder
                    :activation-fn :relu
                    :updater :adam
                    :adam-mean-decay 0.2
                    :adam-var-decay 0.1
                    :dist (dist/new-normal-distribution :mean 0 :std 1)
                    :learning-rate 0.006
                    :weight-init :xavier
                    :layer-name "single layer model example"
                    :n-in 10
                    :n-out 20))

;; these configurations are the same

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; multi-layer configuration
;; here we are also setting layer defaults
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; defaults will apply to layers which do not specify those value in their config

(nn/builder
 :optimization-algo :stochastic-gradient-descent
 :seed 123
 :iterations 1
 :minimize? true
 :use-drop-connect? false
 :lr-score-based-decay-rate 0.002
 :regularization? false
 :default-activation-fn :sigmoid
 :default-weight-init :uniform

 ;; we need to specify the layer order
 :layers {0 (l/activation-layer-builder
             :activation-fn :relu
             :updater :adam
             :adam-mean-decay 0.2
             :adam-var-decay 0.1
             :learning-rate 0.006
             :weight-init :xavier
             :layer-name "example first layer"
             :n-in 10
             :n-out 20)
          1 {:output-layer {:n-in 20
                            :n-out 2
                            :loss-fn :mse
                            :layer-name "example output layer"}}})

;; specifying multi-layer config params

(nn/builder
 ;; network args
 :optimization-algo :stochastic-gradient-descent
 :seed 123
 :iterations 1
 :minimize? true
 :use-drop-connect? false
 :lr-score-based-decay-rate 0.002
 :regularization? false

 ;; layer defaults
 :default-activation-fn :sigmoid
 :default-weight-init :uniform

 ;; the layers
 :layers {0 (l/activation-layer-builder
             :activation-fn :relu
             :updater :adam
             :adam-mean-decay 0.2
             :adam-var-decay 0.1
             :learning-rate 0.006
             :weight-init :xavier
             :layer-name "example first layer"
             :n-in 10
             :n-out 20)
          1 {:output-layer {:n-in 20
                            :n-out 2
                            :loss-fn :mse
                            :layer-name "example output layer"}}}
 ;; multi layer network args
 :backprop? true
 :backprop-type :standard
 :pretrain? false
 :input-pre-processors {0 (pp/new-zero-mean-pre-pre-processor)
                        1 {:unit-variance-processor {}}})

Configuration to Trained models

Multi Layer models

(ns my.ns
  (:require [dl4clj.datasets.iterators :as iter]
            [dl4clj.datasets.input-splits :as split]
            [dl4clj.datasets.record-readers :as rr]
            [dl4clj.optimize.listeners :as listener]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.multilayer.multi-layer-network :as mln]
            [dl4clj.nn.api.model :refer [init! set-listeners!]]
            [dl4clj.nn.api.multi-layer-network :refer [evaluate-classification]]
            [dl4clj.datasets.api.record-readers :refer [initialize-rr!]]
            [dl4clj.eval.api.eval :refer [get-stats get-accuracy]]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; nn-conf -> multi-layer-network
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123 :iterations 1 :regularization? true

   ;; setting layer defaults
   :default-activation-fn :relu :default-l2 7.5e-6
   :default-weight-init :xavier :default-learning-rate 0.0015
   :default-updater :nesterovs :default-momentum 0.98

   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}

   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def multi-layer-network (c/model-from-conf nn-conf))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; local cpu training with dl4j pre-built iterators
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; lets use the pre-built Mnist data set iterator

(def train-mnist-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-mnist-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

;; and lets set a listener so we can know how training is going

(def score-listener (listener/new-score-iteration-listener :print-every-n 5))

;; and attach it to our model

;; TODO: listeners are broken, look into log4j warnning
(def mln-with-listener (set-listeners! :model multi-layer-network
                                       :listeners [score-listener]))

(def trained-mln (mln/train-mln-with-ds-iter! :mln mln-with-listener
                                              :iter train-mnist-iter
                                              :n-epochs 15
                                              :as-code? false))

;; training happens because :as-code? = false
;; if it was true, we would still just have a data structure
;; we now have a trained model that has seen the training dataset 15 times
;; time to evaluate our model

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;Create an evaluation object
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def eval-obj (evaluate-classification :mln trained-mln
                                       :iter test-mnist-iter))

;; always remember that these objects are stateful, dont use the same eval-obj
;; to eval two different networks
;; we trained the model on a training dataset.  We evaluate on a test set

(println (get-stats :evaler eval-obj))
;; this will print the stats to standard out for each feature/label pair

;;Examples labeled as 0 classified by model as 0: 968 times
;;Examples labeled as 0 classified by model as 1: 1 times
;;Examples labeled as 0 classified by model as 2: 1 times
;;Examples labeled as 0 classified by model as 3: 1 times
;;Examples labeled as 0 classified by model as 5: 1 times
;;Examples labeled as 0 classified by model as 6: 3 times
;;Examples labeled as 0 classified by model as 7: 1 times
;;Examples labeled as 0 classified by model as 8: 2 times
;;Examples labeled as 0 classified by model as 9: 2 times
;;Examples labeled as 1 classified by model as 1: 1126 times
;;Examples labeled as 1 classified by model as 2: 2 times
;;Examples labeled as 1 classified by model as 3: 1 times
;;Examples labeled as 1 classified by model as 5: 1 times
;;Examples labeled as 1 classified by model as 6: 2 times
;;Examples labeled as 1 classified by model as 7: 1 times
;;Examples labeled as 1 classified by model as 8: 2 times
;;Examples labeled as 2 classified by model as 0: 3 times
;;Examples labeled as 2 classified by model as 1: 2 times
;;Examples labeled as 2 classified by model as 2: 1006 times
;;Examples labeled as 2 classified by model as 3: 2 times
;;Examples labeled as 2 classified by model as 4: 3 times
;;Examples labeled as 2 classified by model as 6: 3 times
;;Examples labeled as 2 classified by model as 7: 7 times
;;Examples labeled as 2 classified by model as 8: 6 times
;;Examples labeled as 3 classified by model as 2: 4 times
;;Examples labeled as 3 classified by model as 3: 990 times
;;Examples labeled as 3 classified by model as 5: 3 times
;;Examples labeled as 3 classified by model as 7: 3 times
;;Examples labeled as 3 classified by model as 8: 3 times
;;Examples labeled as 3 classified by model as 9: 7 times
;;Examples labeled as 4 classified by model as 2: 2 times
;;Examples labeled as 4 classified by model as 3: 1 times
;;Examples labeled as 4 classified by model as 4: 967 times
;;Examples labeled as 4 classified by model as 6: 4 times
;;Examples labeled as 4 classified by model as 7: 1 times
;;Examples labeled as 4 classified by model as 9: 7 times
;;Examples labeled as 5 classified by model as 0: 2 times
;;Examples labeled as 5 classified by model as 3: 6 times
;;Examples labeled as 5 classified by model as 4: 1 times
;;Examples labeled as 5 classified by model as 5: 874 times
;;Examples labeled as 5 classified by model as 6: 3 times
;;Examples labeled as 5 classified by model as 7: 1 times
;;Examples labeled as 5 classified by model as 8: 3 times
;;Examples labeled as 5 classified by model as 9: 2 times
;;Examples labeled as 6 classified by model as 0: 4 times
;;Examples labeled as 6 classified by model as 1: 3 times
;;Examples labeled as 6 classified by model as 3: 2 times
;;Examples labeled as 6 classified by model as 4: 4 times
;;Examples labeled as 6 classified by model as 5: 4 times
;;Examples labeled as 6 classified by model as 6: 939 times
;;Examples labeled as 6 classified by model as 7: 1 times
;;Examples labeled as 6 classified by model as 8: 1 times
;;Examples labeled as 7 classified by model as 1: 7 times
;;Examples labeled as 7 classified by model as 2: 4 times
;;Examples labeled as 7 classified by model as 3: 3 times
;;Examples labeled as 7 classified by model as 7: 1005 times
;;Examples labeled as 7 classified by model as 8: 2 times
;;Examples labeled as 7 classified by model as 9: 7 times
;;Examples labeled as 8 classified by model as 0: 3 times
;;Examples labeled as 8 classified by model as 2: 3 times
;;Examples labeled as 8 classified by model as 3: 2 times
;;Examples labeled as 8 classified by model as 4: 4 times
;;Examples labeled as 8 classified by model as 5: 3 times
;;Examples labeled as 8 classified by model as 6: 2 times
;;Examples labeled as 8 classified by model as 7: 4 times
;;Examples labeled as 8 classified by model as 8: 947 times
;;Examples labeled as 8 classified by model as 9: 6 times
;;Examples labeled as 9 classified by model as 0: 2 times
;;Examples labeled as 9 classified by model as 1: 2 times
;;Examples labeled as 9 classified by model as 3: 4 times
;;Examples labeled as 9 classified by model as 4: 8 times
;;Examples labeled as 9 classified by model as 6: 1 times
;;Examples labeled as 9 classified by model as 7: 4 times
;;Examples labeled as 9 classified by model as 8: 2 times
;;Examples labeled as 9 classified by model as 9: 986 times

;;==========================Scores========================================
;; Accuracy:        0.9808
;; Precision:       0.9808
;; Recall:          0.9807
;; F1 Score:        0.9807
;;========================================================================

;; can get the stats that are printed via fns in the evaluation namespace
;; after running eval-model-whole-ds

(get-accuracy :evaler evaler-with-stats) ;; => 0.9808

Model Tuning

Early Stopping (controlling training)

it is recommened you start here when designing models

using dl4clj.core


(ns my.ns
  (:require [dl4clj.earlystopping.termination-conditions :refer :all]
            [dl4clj.earlystopping.model-saver :refer [new-in-memory-saver]]
            [dl4clj.nn.api.multi-layer-network :refer [evaluate-classification]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :as iter]
            [dl4clj.core :as c]))

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123
   :iterations 1
   :regularization? true

   ;; setting layer defaults
   :default-activation-fn :relu
   :default-l2 7.5e-6
   :default-weight-init :xavier
   :default-learning-rate 0.0015
   :default-updater :nesterovs
   :default-momentum 0.98

   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}

   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def train-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

(def invalid-score-condition (new-invalid-score-iteration-termination-condition))

(def max-score-condition (new-max-score-iteration-termination-condition
                          :max-score 20.0))

(def max-time-condition (new-max-time-iteration-termination-condition
                         :max-time-val 10
                         :max-time-unit :minutes))

(def score-doesnt-improve-condition (new-score-improvement-epoch-termination-condition
                                     :max-n-epoch-no-improve 5))

(def target-score-condition (new-best-score-epoch-termination-condition
                             :best-expected-score 0.009))

(def max-number-epochs-condition (new-max-epochs-termination-condition :max-n 20))

(def in-mem-saver (new-in-memory-saver))

(def trained-mln
;; defaults to returning the model
  (c/train-with-early-stopping
   :nn-conf nn-conf
   :training-iter train-mnist-iter
   :testing-iter test-mnist-iter
   :eval-every-n-epochs 1
   :iteration-termination-conditions [invalid-score-condition
                                      max-score-condition
                                      max-time-condition]
   :epoch-termination-conditions [score-doesnt-improve-condition
                                  target-score-condition
                                  max-number-epochs-condition]
   :save-last-model? true
   :model-saver in-mem-saver
   :as-code? false))

(def model-evaler
  (evaluate-classification :mln trained-mln :iter test-mnist-iter))

(println (get-stats :evaler model-evaler))
  • explicit, step by step way of doing this
(ns my.ns
  (:require [dl4clj.earlystopping.early-stopping-config :refer [new-early-stopping-config]]
            [dl4clj.earlystopping.termination-conditions :refer :all]
            [dl4clj.earlystopping.model-saver :refer [new-in-memory-saver new-local-file-model-saver]]
            [dl4clj.earlystopping.score-calc :refer [new-ds-loss-calculator]]
            [dl4clj.earlystopping.early-stopping-trainer :refer [new-early-stopping-trainer]]
            [dl4clj.earlystopping.api.early-stopping-trainer :refer [fit-trainer!]]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.multilayer.multi-layer-network :as mln]
            [dl4clj.utils :refer [load-model!]]
            [dl4clj.datasets.iterators :as iter]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; start with our network config
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123 :iterations 1 :regularization? true
   ;; setting layer defaults
   :default-activation-fn :relu :default-l2 7.5e-6
   :default-weight-init :xavier :default-learning-rate 0.0015
   :default-updater :nesterovs :default-momentum 0.98
   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}
   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def mln (c/model-from-conf nn-conf))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; the training/testing data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def train-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; we are going to need termination conditions
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; these allow us to control when we exit training

;; this can be based off of iterations or epochs

;; iteration termination conditions

(def invalid-score-condition (new-invalid-score-iteration-termination-condition))

(def max-score-condition (new-max-score-iteration-termination-condition
                          :max-score 20.0))

(def max-time-condition (new-max-time-iteration-termination-condition
                         :max-time-val 10
                         :max-time-unit :minutes))

;; epoch termination conditions

(def score-doesnt-improve-condition (new-score-improvement-epoch-termination-condition
                                     :max-n-epoch-no-improve 5))

(def target-score-condition (new-best-score-epoch-termination-condition :best-expected-score 0.009))

(def max-number-epochs-condition (new-max-epochs-termination-condition :max-n 20))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; we also need a way to save our model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; can be in memory or to a local directory

(def in-mem-saver (new-in-memory-saver))

(def local-file-saver (new-local-file-model-saver :directory "resources/tmp/readme/"))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; set up your score calculator
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def score-calcer (new-ds-loss-calculator :iter test-iter
                                          :average? true))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; create an early stopping configuration
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; termination conditions
;; a way to save our model
;; a way to calculate the score of our model on the dataset

(def early-stopping-conf
  (new-early-stopping-config
   :epoch-termination-conditions [score-doesnt-improve-condition
                                  target-score-condition
                                  max-number-epochs-condition]
   :iteration-termination-conditions [invalid-score-condition
                                      max-score-condition
                                      max-time-condition]
   :eval-every-n-epochs 5
   :model-saver local-file-saver
   :save-last-model? true
   :score-calculator score-calcer))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; create an early stopping trainer from our data, model and early stopping conf
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def es-trainer (new-early-stopping-trainer :early-stopping-conf early-stopping-conf
                                            :mln mln
                                            :iter train-iter))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; fit and use our early stopping trainer
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def es-trainer-fitted (fit-trainer! es-trainer :as-code? false))

;; when the trainer terminates, you will see something like this
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO  Completed training epoch 14
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO  New best model: score = 0.005225599372851298,
;;                                                   epoch = 14 (previous: score = 0.018243224899038346, epoch = 7)
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO Hit epoch termination condition at epoch 14.
;;                                           Details: BestScoreEpochTerminationCondition(0.009)

;; and if we look at the es-trainer-fitted object we see

;;#object[org.deeplearning4j.earlystopping.EarlyStoppingResult 0x5ab74f27 EarlyStoppingResult
;;(terminationReason=EpochTerminationCondition,details=BestScoreEpochTerminationCondition(0.009),
;; bestModelEpoch=14,bestModelScore=0.005225599372851298,totalEpochs=15)]

;; and our model has been saved to /resources/tmp/readme/bestModel.bin
;; there we have our model config, model params and our updater state

;; we can then load this model to use it or continue refining it

(def loaded-model (load-model! :path "resources/tmp/readme/bestModel.bin"
                               :load-updater? true))

Transfer Learning (freezing layers)


;; TODO: need to write up examples

Spark Training

dl4j Spark usage

How it is done in dl4clj

  • Uses dl4clj.core
    • This example uses a fn which takes care of most steps for you
      • allows you to pass args as code bc the fn accounts for the multiple spark contexts issue encountered when everything is just a data structure

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :refer [new-iris-data-set-iterator]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.spark.masters.param-avg :as master]
            [dl4clj.spark.data.java-rdd :refer [new-java-spark-context
                                                java-rdd-from-iter]]
            [dl4clj.spark.api.dl4j-multi-layer :refer [eval-classification-spark-mln
                                                       get-spark-context]]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 1, create your model config
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def mln-conf
  (nn/builder
   :optimization-algo :stochastic-gradient-descent
   :default-learning-rate 0.006
   :layers {0 (l/dense-layer-builder :n-in 4 :n-out 2 :activation-fn :relu)
            1 {:output-layer
               {:loss-fn :negativeloglikelihood
                :n-in 2 :n-out 3
                :activation-fn :soft-max
                :weight-init :xavier}}}
   :backprop? true
   :backprop-type :standard))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 2, training master
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def training-master
  (master/new-parameter-averaging-training-master
   :build? true
   :rdd-n-examples 10
   :n-workers 4
   :averaging-freq 10
   :batch-size-per-worker 2
   :export-dir "resources/spark/master/"
   :rdd-training-approach :direct
   :repartition-data :always
   :repartition-strategy :balanced
   :seed 1234
   :save-updater? true
   :storage-level :none))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 3, spark context
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def your-spark-context
  (new-java-spark-context :app-name "example app"))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 4, training data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def iris-iter
  (new-iris-data-set-iterator
   :batch-size 1
   :n-examples 5))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, spark mln
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def fitted-spark-mln
  (c/train-with-spark :spark-context your-spark-context
                      :mln-conf mln-conf
                      :training-master training-master
                      :iter iris-iter
                      :n-epochs 1
                      :as-code? false))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, use spark context from spark-mln to create rdd
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; TODO: eliminate this step

(def our-rdd
  (let [sc (get-spark-context fitted-spark-mln :as-code? false)]
    (java-rdd-from-iter :spark-context sc
                        :iter iris-iter)))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 6, evaluation model and print stats (poor performance of model expected)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def eval-obj
  (eval-classification-spark-mln
   :spark-mln fitted-spark-mln
   :rdd our-rdd))

(println (get-stats :evaler eval-obj))

  • this example demonstrates the dl4j workflow
    • NOTE: unlike the previous example, this one requires dl4j objects to be used
      • this is becaues spark only wants you to have one spark context at a time
(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :refer [new-iris-data-set-iterator]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.spark.masters.param-avg :as master]
            [dl4clj.spark.data.java-rdd :refer [new-java-spark-context java-rdd-from-iter]]
            [dl4clj.spark.dl4j-multi-layer :as spark-mln]
            [dl4clj.spark.api.dl4j-multi-layer :refer [fit-spark-mln!
                                                       eval-classification-spark-mln]]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 1, create your model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def mln-conf
  (nn/builder
   :optimization-algo :stochastic-gradient-descent
   :default-learning-rate 0.006
   :layers {0 (l/dense-layer-builder :n-in 4 :n-out 2 :activation-fn :relu)
            1 {:output-layer
               {:loss-fn :negativeloglikelihood
                :n-in 2 :n-out 3
                :activation-fn :soft-max
                :weight-init :xavier}}}
   :backprop? true
   :as-code? false
   :backprop-type :standard))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 2, create a training master
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; not all options specified, but most are

(def training-master
  (master/new-parameter-averaging-training-master
   :build? true
   :rdd-n-examples 10
   :n-workers 4
   :averaging-freq 10
   :batch-size-per-worker 2
   :export-dir "resources/spark/master/"
   :rdd-training-approach :direct
   :repartition-data :always
   :repartition-strategy :balanced
   :seed 1234
   :as-code? false
   :save-updater? true
   :storage-level :none))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 3, create a Spark Multi Layer Network
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def your-spark-context
  (new-java-spark-context :app-name "example app" :as-code? false))

;; new-java-spark-context will turn an existing spark-configuration into a java spark context
;; or create a new java spark context with master set to "local[*]" and the app name
;; set to :app-name


(def spark-mln
  (spark-mln/new-spark-multi-layer-network
   :spark-context your-spark-context
   :mln mln-conf
   :training-master training-master
   :as-code? false))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 4, load your data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; one way is via a dataset-iterator
;; can make one directly from a dataset (iterator data-set)
;; see: nd4clj.linalg.dataset.api.data-set and nd4clj.linalg.dataset.data-set
;; we are going to use a pre-built one

(def iris-iter
  (new-iris-data-set-iterator
   :batch-size 1
   :n-examples 5
   :as-code? false))

;; now lets convert the data into a javaRDD

(def our-rdd
  (java-rdd-from-iter :spark-context your-spark-context
                      :iter iris-iter))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, fit and evaluate the model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def fitted-spark-mln
  (fit-spark-mln!
   :spark-mln spark-mln
   :rdd our-rdd
   :n-epochs 1))
;; this fn also has the option to supply :path-to-data instead of :rdd
;; that path should point to a directory containing a number of dataset objects

(def eval-obj
  (eval-classification-spark-mln
   :spark-mln fitted-spark-mln
   :rdd our-rdd))
;; we would want to have different testing and training rdd's but here we are using
;; the data we trained on

;; lets get the stats for how our model performed

(println (get-stats :evaler eval-obj))

Terminology

Coming soon

Packages to come back to:

Implement ComputationGraphs and the classes which use them

NLP

Parallelism

TSNE

UI


Author: yetanalytics
Source Code: https://github.com/yetanalytics/dl4clj
License: BSD-2-Clause License

#machine-learning #deep-learning 

Arvel  Parker

Arvel Parker

1591611780

How to Find Ulimit For user on Linux

How can I find the correct ulimit values for a user account or process on Linux systems?

For proper operation, we must ensure that the correct ulimit values set after installing various software. The Linux system provides means of restricting the number of resources that can be used. Limits set for each Linux user account. However, system limits are applied separately to each process that is running for that user too. For example, if certain thresholds are too low, the system might not be able to server web pages using Nginx/Apache or PHP/Python app. System resource limits viewed or set with the NA command. Let us see how to use the ulimit that provides control over the resources available to the shell and processes.

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

MEAN Stack Tutorial MongoDB ExpressJS AngularJS NodeJS

We are going to build a full stack Todo App using the MEAN (MongoDB, ExpressJS, AngularJS and NodeJS). This is the last part of three-post series tutorial.

MEAN Stack tutorial series:

AngularJS tutorial for beginners (Part I)
Creating RESTful APIs with NodeJS and MongoDB Tutorial (Part II)
MEAN Stack Tutorial: MongoDB, ExpressJS, AngularJS and NodeJS (Part III) 👈 you are here
Before completing the app, let’s cover some background about the this stack. If you rather jump to the hands-on part click here to get started.

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

Creating RESTful APIs with NodeJS and MongoDB Tutorial

Welcome to this tutorial about RESTful API using Node.js (Express.js) and MongoDB (mongoose)! We are going to learn how to install and use each component individually and then proceed to create a RESTful API.

MEAN Stack tutorial series:

AngularJS tutorial for beginners (Part I)
Creating RESTful APIs with NodeJS and MongoDB Tutorial (Part II) 👈 you are here
MEAN Stack Tutorial: MongoDB, ExpressJS, AngularJS and NodeJS (Part III)

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

Everything You Need to Know About Instagram Bot with Python

How to build an Instagram bot using Python

Instagram is the fastest-growing social network, with 1 billion monthly users. It also has the highest engagement rate. To gain followers on Instagram, you’d have to upload engaging content, follow users, like posts, comment on user posts and a whole lot. This can be time-consuming and daunting. But there is hope, you can automate all of these tasks. In this course, we’re going to build an Instagram bot using Python to automate tasks on Instagram.

What you’ll learn:

  • Instagram Automation
  • Build a Bot with Python

Increase your Instagram followers with a simple Python bot

I got around 500 real followers in 4 days!

Growing an audience is an expensive and painful task. And if you’d like to build an audience that’s relevant to you, and shares common interests, that’s even more difficult. I always saw Instagram has a great way to promote my photos, but I never had more than 380 followers… Every once in a while, I decide to start posting my photos on Instagram again, and I manage to keep posting regularly for a while, but it never lasts more than a couple of months, and I don’t have many followers to keep me motivated and engaged.

The objective of this project is to build a bigger audience and as a plus, maybe drive some traffic to my website where I sell my photos!

A year ago, on my last Instagram run, I got one of those apps that lets you track who unfollowed you. I was curious because in a few occasions my number of followers dropped for no apparent reason. After some research, I realized how some users basically crawl for followers. They comment, like and follow people — looking for a follow back. Only to unfollow them again in the next days.

I can’t say this was a surprise to me, that there were bots in Instagram… It just made me want to build one myself!

And that is why we’re here, so let’s get to it! I came up with a simple bot in Python, while I was messing around with Selenium and trying to figure out some project to use it. Simply put, Selenium is like a browser you can interact with very easily in Python.

Ideally, increasing my Instagram audience will keep me motivated to post regularly. As an extra, I included my website in my profile bio, where people can buy some photos. I think it is a bit of a stretch, but who knows?! My sales are basically zero so far, so it should be easy to track that conversion!

Just what the world needed! Another Instagram bot…

After giving this project some thought, my objective was to increase my audience with relevant people. I want to get followers that actually want to follow me and see more of my work. It’s very easy to come across weird content in the most used hashtags, so I’ve planed this bot to lookup specific hashtags and interact with the photos there. This way, I can be very specific about what kind of interests I want my audience to have. For instance, I really like long exposures, so I can target people who use that hashtag and build an audience around this kind of content. Simple and efficient!

My gallery is a mix of different subjects and styles, from street photography to aerial photography, and some travel photos too. Since it’s my hometown, I also have lots of Lisbon images there. These will be the main topics I’ll use in the hashtags I want to target.

This is not a “get 1000 followers in 24 hours” kind of bot!

So what kind of numbers are we talking about?

I ran the bot a few times in a few different hashtags like “travelblogger”, “travelgram”, “lisbon”, “dronephotography”. In the course of three days I went from 380 to 800 followers. Lots of likes, comments and even some organic growth (people that followed me but were not followed by the bot).

To be clear, I’m not using this bot intensively, as Instagram will stop responding if you run it too fast. It needs to have some sleep commands in between the actions, because after some comments and follows in a short period of time, Instagram stops responding and the bot crashes.

You will be logged into your account, so I’m almost sure that Instagram can know you’re doing something weird if you speed up the process. And most importantly, after doing this for a dozen hashtags, it just gets harder to find new users in the same hashtags. You will need to give it a few days to refresh the user base there.

But I don’t want to follow so many people in the process…

The most efficient way to get followers in Instagram (apart from posting great photos!) is to follow people. And this bot worked really well for me because I don’t care if I follow 2000 people to get 400 followers.

The bot saves a list with all the users that were followed while it was running, so someday I may actually do something with this list. For instance, I can visit each user profile, evaluate how many followers or posts they have, and decide if I want to keep following them. Or I can get the first picture in their gallery and check its date to see if they are active users.

If we remove the follow action from the bot, I can assure you the growth rate will suffer, as people are less inclined to follow based on a single like or comment.

Why will you share your code?!

That’s the debate I had with myself. Even though I truly believe in giving back to the community (I still learn a lot from it too!), there are several paid platforms that do more or less the same as this project. Some are shady, some are used by celebrities. The possibility of starting a similar platform myself, is not off the table yet, so why make the code available?

With that in mind, I decided to add an extra level of difficulty to the process, so I was going to post the code below as an image. I wrote “was”, because meanwhile, I’ve realized the image I’m getting is low quality. Which in turn made me reconsider and post the gist. I’m that nice! The idea behind the image was that if you really wanted to use it, you would have to type the code yourself. And that was my way of limiting the use of this tool to people that actually go through the whole process to create it and maybe even improve it.

I learn a lot more when I type the code myself, instead of copy/pasting scripts. I hope you feel the same way!

The script isn’t as sophisticated as it could be, and I know there’s lots of room to improve it. But hey… it works! I have other projects I want to add to my portfolio, so my time to develop it further is rather limited. Nevertheless, I will try to update this article if I dig deeper.

This is the last subtitle!

You’ll need Python (I’m using Python 3.7), Selenium, a browser (in my case I’ll be using Chrome) and… obviously, an Instagram account! Quick overview regarding what the bot will do:

  • Open a browser and login with your credentials
  • For every hashtag in the hashtag list, it will open the page and click the first picture to open it
  • It will then like, follow, comment and move to the next picture, in a 200 iterations loop (number can be adjusted)
  • Saves a list with all the users you followed using the bot

If you reached this paragraph, thank you! You totally deserve to collect your reward! If you find this useful for your profile/brand in any way, do share your experience below :)

from selenium import webdriver
from selenium.webdriver.common.keys import Keys
from time import sleep, strftime
from random import randint
import pandas as pd

chromedriver_path = 'C:/Users/User/Downloads/chromedriver_win32/chromedriver.exe' # Change this to your own chromedriver path!
webdriver = webdriver.Chrome(executable_path=chromedriver_path)
sleep(2)
webdriver.get('https://www.instagram.com/accounts/login/?source=auth_switcher')
sleep(3)

username = webdriver.find_element_by_name('username')
username.send_keys('your_username')
password = webdriver.find_element_by_name('password')
password.send_keys('your_password')

button_login = webdriver.find_element_by_css_selector('#react-root > section > main > div > article > div > div:nth-child(1) > div > form > div:nth-child(3) > button')
button_login.click()
sleep(3)

notnow = webdriver.find_element_by_css_selector('body > div:nth-child(13) > div > div > div > div.mt3GC > button.aOOlW.HoLwm')
notnow.click() #comment these last 2 lines out, if you don't get a pop up asking about notifications

In order to use chrome with Selenium, you need to install chromedriver. It’s a fairly simple process and I had no issues with it. Simply install and replace the path above. Once you do that, our variable webdriver will be our Chrome tab.

In cell number 3 you should replace the strings with your own username and the respective password. This is for the bot to type it in the fields displayed. You might have already noticed that when running cell number 2, Chrome opened a new tab. After the password, I’ll define the login button as an object, and in the following line, I click it.

Once you get in inspect mode find the bit of html code that corresponds to what you want to map. Right click it and hover over Copy. You will see that you have some options regarding how you want it to be copied. I used a mix of XPath and css selectors throughout the code (it’s visible in the find_element_ method). It took me a while to get all the references to run smoothly. At points, the css or the xpath directions would fail, but as I adjusted the sleep times, everything started running smoothly.

In this case, I selected “copy selector” and pasted it inside a find_element_ method (cell number 3). It will get you the first result it finds. If it was find_elements_, all elements would be retrieved and you could specify which to get.

Once you get that done, time for the loop. You can add more hashtags in the hashtag_list. If you run it for the first time, you still don’t have a file with the users you followed, so you can simply create prev_user_list as an empty list.

Once you run it once, it will save a csv file with a timestamp with the users it followed. That file will serve as the prev_user_list on your second run. Simple and easy to keep track of what the bot does.

Update with the latest timestamp on the following runs and you get yourself a series of csv backlogs for every run of the bot.

Instagram bot with Python

The code is really simple. If you have some basic notions of Python you can probably pick it up quickly. I’m no Python ninja and I was able to build it, so I guess that if you read this far, you are good to go!

hashtag_list = ['travelblog', 'travelblogger', 'traveler']

# prev_user_list = [] - if it's the first time you run it, use this line and comment the two below
prev_user_list = pd.read_csv('20181203-224633_users_followed_list.csv', delimiter=',').iloc[:,1:2] # useful to build a user log
prev_user_list = list(prev_user_list['0'])

new_followed = []
tag = -1
followed = 0
likes = 0
comments = 0

for hashtag in hashtag_list:
    tag += 1
    webdriver.get('https://www.instagram.com/explore/tags/'+ hashtag_list[tag] + '/')
    sleep(5)
    first_thumbnail = webdriver.find_element_by_xpath('//*[@id="react-root"]/section/main/article/div[1]/div/div/div[1]/div[1]/a/div')
    
    first_thumbnail.click()
    sleep(randint(1,2))    
    try:        
        for x in range(1,200):
            username = webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/header/div[2]/div[1]/div[1]/h2/a').text
            
            if username not in prev_user_list:
                # If we already follow, do not unfollow
                if webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/header/div[2]/div[1]/div[2]/button').text == 'Follow':
                    
                    webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/header/div[2]/div[1]/div[2]/button').click()
                    
                    new_followed.append(username)
                    followed += 1

                    # Liking the picture
                    button_like = webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/div[2]/section[1]/span[1]/button/span')
                    
                    button_like.click()
                    likes += 1
                    sleep(randint(18,25))

                    # Comments and tracker
                    comm_prob = randint(1,10)
                    print('{}_{}: {}'.format(hashtag, x,comm_prob))
                    if comm_prob > 7:
                        comments += 1
                        webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/div[2]/section[1]/span[2]/button/span').click()
                        comment_box = webdriver.find_element_by_xpath('/html/body/div[3]/div/div[2]/div/article/div[2]/section[3]/div/form/textarea')

                        if (comm_prob < 7):
                            comment_box.send_keys('Really cool!')
                            sleep(1)
                        elif (comm_prob > 6) and (comm_prob < 9):
                            comment_box.send_keys('Nice work :)')
                            sleep(1)
                        elif comm_prob == 9:
                            comment_box.send_keys('Nice gallery!!')
                            sleep(1)
                        elif comm_prob == 10:
                            comment_box.send_keys('So cool! :)')
                            sleep(1)
                        # Enter to post comment
                        comment_box.send_keys(Keys.ENTER)
                        sleep(randint(22,28))

                # Next picture
                webdriver.find_element_by_link_text('Next').click()
                sleep(randint(25,29))
            else:
                webdriver.find_element_by_link_text('Next').click()
                sleep(randint(20,26))
    # some hashtag stops refreshing photos (it may happen sometimes), it continues to the next
    except:
        continue

for n in range(0,len(new_followed)):
    prev_user_list.append(new_followed[n])
    
updated_user_df = pd.DataFrame(prev_user_list)
updated_user_df.to_csv('{}_users_followed_list.csv'.format(strftime("%Y%m%d-%H%M%S")))
print('Liked {} photos.'.format(likes))
print('Commented {} photos.'.format(comments))
print('Followed {} new people.'.format(followed))

Instagram bot with Python

The print statement inside the loop is the way I found to be able to have a tracker that lets me know at what iteration the bot is all the time. It will print the hashtag it’s in, the number of the iteration, and the random number generated for the comment action. I decided not to post comments in every page, so I added three different comments and a random number between 1 and 10 that would define if there was any comment at all, or one of the three. The loop ends, we append the new_followed users to the previous users “database” and saves the new file with the timestamp. You should also get a small report.

Instagram bot with Python

And that’s it!

After a few hours without checking the phone, these were the numbers I was getting. I definitely did not expect it to do so well! In about 4 days since I’ve started testing it, I had around 500 new followers, which means I have doubled my audience in a matter of days. I’m curious to see how many of these new followers I will lose in the next days, to see if the growth can be sustainable. I also had a lot more “likes” in my latest photos, but I guess that’s even more expected than the follow backs.

Instagram bot with Python

It would be nice to get this bot running in a server, but I have other projects I want to explore, and configuring a server is not one of them! Feel free to leave a comment below, and I’ll do my best to answer your questions.

I’m actually curious to see how long will I keep posting regularly! If you feel like this article was helpful for you, consider thanking me by buying one of my photos.

Instagram bot with Python



How to Make an Instagram Bot With Python and InstaPy

Instagram bot with Python

What do SocialCaptain, Kicksta, Instavast, and many other companies have in common? They all help you reach a greater audience, gain more followers, and get more likes on Instagram while you hardly lift a finger. They do it all through automation, and people pay them a good deal of money for it. But you can do the same thing—for free—using InstaPy!

In this tutorial, you’ll learn how to build a bot with Python and InstaPy, which automates your Instagram activities so that you gain more followers and likes with minimal manual input. Along the way, you’ll learn about browser automation with Selenium and the Page Object Pattern, which together serve as the basis for InstaPy.

In this tutorial, you’ll learn:

  • How Instagram bots work
  • How to automate a browser with Selenium
  • How to use the Page Object Pattern for better readability and testability
  • How to build an Instagram bot with InstaPy

You’ll begin by learning how Instagram bots work before you build one.

Table of Contents

  • How Instagram Bots Work
  • How to Automate a Browser
  • How to Use the Page Object Pattern
  • How to Build an Instagram Bot With InstaPy
    • Essential Features
    • Additional Features in InstaPy
  • Conclusion

Important: Make sure you check Instagram’s Terms of Use before implementing any kind of automation or scraping techniques.

How Instagram Bots Work

How can an automation script gain you more followers and likes? Before answering this question, think about how an actual person gains more followers and likes.

They do it by being consistently active on the platform. They post often, follow other people, and like and leave comments on other people’s posts. Bots work exactly the same way: They follow, like, and comment on a consistent basis according to the criteria you set.

The better the criteria you set, the better your results will be. You want to make sure you’re targeting the right groups because the people your bot interacts with on Instagram will be more likely to interact with your content.

For example, if you’re selling women’s clothing on Instagram, then you can instruct your bot to like, comment on, and follow mostly women or profiles whose posts include hashtags such as #beauty, #fashion, or #clothes. This makes it more likely that your target audience will notice your profile, follow you back, and start interacting with your posts.

How does it work on the technical side, though? You can’t use the Instagram Developer API since it is fairly limited for this purpose. Enter browser automation. It works in the following way:

  1. You serve it your credentials.
  2. You set the criteria for who to follow, what comments to leave, and which type of posts to like.
  3. Your bot opens a browser, types in https://instagram.com on the address bar, logs in with your credentials, and starts doing the things you instructed it to do.

Next, you’ll build the initial version of your Instagram bot, which will automatically log in to your profile. Note that you won’t use InstaPy just yet.

How to Automate a Browser

For this version of your Instagram bot, you’ll be using Selenium, which is the tool that InstaPy uses under the hood.

First, install Selenium. During installation, make sure you also install the Firefox WebDriver since the latest version of InstaPy dropped support for Chrome. This also means that you need the Firefox browser installed on your computer.

Now, create a Python file and write the following code in it:

from time import sleep

from selenium import webdriver


browser = webdriver.Firefox()


browser.get('https://www.instagram.com/')


sleep(5)


browser.close()

Run the code and you’ll see that a Firefox browser opens and directs you to the Instagram login page. Here’s a line-by-line breakdown of the code:

  • Lines 1 and 2 import sleep and webdriver.
  • Line 4 initializes the Firefox driver and sets it to browser.
  • Line 6 types https://www.instagram.com/ on the address bar and hits Enter.
  • Line 8 waits for five seconds so you can see the result. Otherwise, it would close the browser instantly.
  • Line 10 closes the browser.

This is the Selenium version of Hello, World. Now you’re ready to add the code that logs in to your Instagram profile. But first, think about how you would log in to your profile manually. You would do the following:

  1. Go to https://www.instagram.com/.
  2. Click the login link.
  3. Enter your credentials.
  4. Hit the login button.

The first step is already done by the code above. Now change it so that it clicks on the login link on the Instagram home page:

from time import sleep

from selenium import webdriver


browser = webdriver.Firefox()

browser.implicitly_wait(5)


browser.get('https://www.instagram.com/')


login_link = browser.find_element_by_xpath("//a[text()='Log in']")

login_link.click()


sleep(5)


browser.close()

Note the highlighted lines:

  • Line 5 sets five seconds of waiting time. If Selenium can’t find an element, then it waits for five seconds to allow everything to load and tries again.
  • Line 9 finds the element <a> whose text is equal to Log in. It does this using XPath, but there are a few other methods you could use.
  • Line 10 clicks on the found element <a> for the login link.

Run the script and you’ll see your script in action. It will open the browser, go to Instagram, and click on the login link to go to the login page.

On the login page, there are three important elements:

  1. The username input
  2. The password input
  3. The login button

Next, change the script so that it finds those elements, enters your credentials, and clicks on the login button:

from time import sleep

from selenium import webdriver


browser = webdriver.Firefox()

browser.implicitly_wait(5)


browser.get('https://www.instagram.com/')


login_link = browser.find_element_by_xpath("//a[text()='Log in']")

login_link.click()


sleep(2)


username_input = browser.find_element_by_css_selector("input[name='username']")

password_input = browser.find_element_by_css_selector("input[name='password']")


username_input.send_keys("<your username>")

password_input.send_keys("<your password>")


login_button = browser.find_element_by_xpath("//button[@type='submit']")

login_button.click()


sleep(5)


browser.close()

Here’s a breakdown of the changes:

  1. Line 12 sleeps for two seconds to allow the page to load.
  2. Lines 14 and 15 find username and password inputs by CSS. You could use any other method that you prefer.
  3. Lines 17 and 18 type your username and password in their respective inputs. Don’t forget to fill in <your username> and <your password>!
  4. Line 20 finds the login button by XPath.
  5. Line 21 clicks on the login button.

Run the script and you’ll be automatically logged in to to your Instagram profile.

You’re off to a good start with your Instagram bot. If you were to continue writing this script, then the rest would look very similar. You would find the posts that you like by scrolling down your feed, find the like button by CSS, click on it, find the comments section, leave a comment, and continue.

The good news is that all of those steps can be handled by InstaPy. But before you jump into using Instapy, there is one other thing that you should know about to better understand how InstaPy works: the Page Object Pattern.

How to Use the Page Object Pattern

Now that you’ve written the login code, how would you write a test for it? It would look something like the following:

def test_login_page(browser):
    browser.get('https://www.instagram.com/accounts/login/')
    username_input = browser.find_element_by_css_selector("input[name='username']")
    password_input = browser.find_element_by_css_selector("input[name='password']")
    username_input.send_keys("<your username>")
    password_input.send_keys("<your password>")
    login_button = browser.find_element_by_xpath("//button[@type='submit']")
    login_button.click()

    errors = browser.find_elements_by_css_selector('#error_message')
    assert len(errors) == 0

Can you see what’s wrong with this code? It doesn’t follow the DRY principle. That is, the code is duplicated in both the application and the test code.

Duplicating code is especially bad in this context because Selenium code is dependent on UI elements, and UI elements tend to change. When they do change, you want to update your code in one place. That’s where the Page Object Pattern comes in.

With this pattern, you create page object classes for the most important pages or fragments that provide interfaces that are straightforward to program to and that hide the underlying widgetry in the window. With this in mind, you can rewrite the code above and create a HomePage class and a LoginPage class:

from time import sleep

class LoginPage:
    def __init__(self, browser):
        self.browser = browser

    def login(self, username, password):
        username_input = self.browser.find_element_by_css_selector("input[name='username']")
        password_input = self.browser.find_element_by_css_selector("input[name='password']")
        username_input.send_keys(username)
        password_input.send_keys(password)
        login_button = browser.find_element_by_xpath("//button[@type='submit']")
        login_button.click()
        sleep(5)

class HomePage:
    def __init__(self, browser):
        self.browser = browser
        self.browser.get('https://www.instagram.com/')

    def go_to_login_page(self):
        self.browser.find_element_by_xpath("//a[text()='Log in']").click()
        sleep(2)
        return LoginPage(self.browser)

The code is the same except that the home page and the login page are represented as classes. The classes encapsulate the mechanics required to find and manipulate the data in the UI. That is, there are methods and accessors that allow the software to do anything a human can.

One other thing to note is that when you navigate to another page using a page object, it returns a page object for the new page. Note the returned value of go_to_log_in_page(). If you had another class called FeedPage, then login() of the LoginPage class would return an instance of that: return FeedPage().

Here’s how you can put the Page Object Pattern to use:

from selenium import webdriver

browser = webdriver.Firefox()
browser.implicitly_wait(5)

home_page = HomePage(browser)
login_page = home_page.go_to_login_page()
login_page.login("<your username>", "<your password>")

browser.close()

It looks much better, and the test above can now be rewritten to look like this:

def test_login_page(browser):
    home_page = HomePage(browser)
    login_page = home_page.go_to_login_page()
    login_page.login("<your username>", "<your password>")

    errors = browser.find_elements_by_css_selector('#error_message')
    assert len(errors) == 0

With these changes, you won’t have to touch your tests if something changes in the UI.

For more information on the Page Object Pattern, refer to the official documentation and to Martin Fowler’s article.

Now that you’re familiar with both Selenium and the Page Object Pattern, you’ll feel right at home with InstaPy. You’ll build a basic bot with it next.

Note: Both Selenium and the Page Object Pattern are widely used for other websites, not just for Instagram.

How to Build an Instagram Bot With InstaPy

In this section, you’ll use InstaPy to build an Instagram bot that will automatically like, follow, and comment on different posts. First, you’ll need to install InstaPy:

$ python3 -m pip install instapy

This will install instapy in your system.

Essential Features

Now you can rewrite the code above with InstaPy so that you can compare the two options. First, create another Python file and put the following code in it:

from instapy import InstaPy

InstaPy(username="<your_username>", password="<your_password>").login()

Replace the username and password with yours, run the script, and voilà! With just one line of code, you achieved the same result.

Even though your results are the same, you can see that the behavior isn’t exactly the same. In addition to simply logging in to your profile, InstaPy does some other things, such as checking your internet connection and the status of the Instagram servers. This can be observed directly on the browser or in the logs:

INFO [2019-12-17 22:03:19] [username]  -- Connection Checklist [1/3] (Internet Connection Status)
INFO [2019-12-17 22:03:20] [username]  - Internet Connection Status: ok
INFO [2019-12-17 22:03:20] [username]  - Current IP is "17.283.46.379" and it's from "Germany/DE"
INFO [2019-12-17 22:03:20] [username]  -- Connection Checklist [2/3] (Instagram Server Status)
INFO [2019-12-17 22:03:26] [username]  - Instagram WebSite Status: Currently Up

Pretty good for one line of code, isn’t it? Now it’s time to make the script do more interesting things than just logging in.

For the purpose of this example, assume that your profile is all about cars, and that your bot is intended to interact with the profiles of people who are also interested in cars.

First, you can like some posts that are tagged #bmw or #mercedes using like_by_tags():

from instapy import InstaPy


session = InstaPy(username="<your_username>", password="<your_password>")

session.login()

session.like_by_tags(["bmw", "mercedes"], amount=5)

Here, you gave the method a list of tags to like and the number of posts to like for each given tag. In this case, you instructed it to like ten posts, five for each of the two tags. But take a look at what happens after you run the script:

INFO [2019-12-17 22:15:58] [username]  Tag [1/2]
INFO [2019-12-17 22:15:58] [username]  --> b'bmw'
INFO [2019-12-17 22:16:07] [username]  desired amount: 14  |  top posts [disabled]: 9  |  possible posts: 43726739
INFO [2019-12-17 22:16:13] [username]  Like# [1/14]
INFO [2019-12-17 22:16:13] [username]  https://www.instagram.com/p/B6MCcGcC3tU/
INFO [2019-12-17 22:16:15] [username]  Image from: b'mattyproduction'
INFO [2019-12-17 22:16:15] [username]  Link: b'https://www.instagram.com/p/B6MCcGcC3tU/'
INFO [2019-12-17 22:16:15] [username]  Description: b'Mal etwas anderes \xf0\x9f\x91\x80\xe2\x98\xba\xef\xb8\x8f Bald ist das komplette Video auf YouTube zu finden (n\xc3\xa4here Infos werden folgen). Vielen Dank an @patrick_jwki @thehuthlife  und @christic_  f\xc3\xbcr das bereitstellen der Autos \xf0\x9f\x94\xa5\xf0\x9f\x98\x8d#carporn#cars#tuning#bagged#bmw#m2#m2competition#focusrs#ford#mk3#e92#m3#panasonic#cinematic#gh5s#dji#roninm#adobe#videography#music#bimmer#fordperformance#night#shooting#'
INFO [2019-12-17 22:16:15] [username]  Location: b'K\xc3\xb6ln, Germany'
INFO [2019-12-17 22:16:51] [username]  --> Image Liked!
INFO [2019-12-17 22:16:56] [username]  --> Not commented
INFO [2019-12-17 22:16:57] [username]  --> Not following
INFO [2019-12-17 22:16:58] [username]  Like# [2/14]
INFO [2019-12-17 22:16:58] [username]  https://www.instagram.com/p/B6MDK1wJ-Kb/
INFO [2019-12-17 22:17:01] [username]  Image from: b'davs0'
INFO [2019-12-17 22:17:01] [username]  Link: b'https://www.instagram.com/p/B6MDK1wJ-Kb/'
INFO [2019-12-17 22:17:01] [username]  Description: b'Someone said cloud? \xf0\x9f\xa4\x94\xf0\x9f\xa4\xad\xf0\x9f\x98\x88 \xe2\x80\xa2\n\xe2\x80\xa2\n\xe2\x80\xa2\n\xe2\x80\xa2\n#bmw #bmwrepost #bmwm4 #bmwm4gts #f82 #bmwmrepost #bmwmsport #bmwmperformance #bmwmpower #bmwm4cs #austinyellow #davs0 #mpower_official #bmw_world_ua #bimmerworld #bmwfans #bmwfamily #bimmers #bmwpost #ultimatedrivingmachine #bmwgang #m3f80 #m5f90 #m4f82 #bmwmafia #bmwcrew #bmwlifestyle'
INFO [2019-12-17 22:17:34] [username]  --> Image Liked!
INFO [2019-12-17 22:17:37] [username]  --> Not commented
INFO [2019-12-17 22:17:38] [username]  --> Not following

By default, InstaPy will like the first nine top posts in addition to your amount value. In this case, that brings the total number of likes per tag to fourteen (nine top posts plus the five you specified in amount).

Also note that InstaPy logs every action it takes. As you can see above, it mentions which post it liked as well as its link, description, location, and whether the bot commented on the post or followed the author.

You may have noticed that there are delays after almost every action. That’s by design. It prevents your profile from getting banned on Instagram.

Now, you probably don’t want your bot liking inappropriate posts. To prevent that from happening, you can use set_dont_like():

from instapy import InstaPy

session = InstaPy(username="<your_username>", password="<your_password>")
session.login()
session.like_by_tags(["bmw", "mercedes"], amount=5)
session.set_dont_like(["naked", "nsfw"])

With this change, posts that have the words naked or nsfw in their descriptions won’t be liked. You can flag any other words that you want your bot to avoid.

Next, you can tell the bot to not only like the posts but also to follow some of the authors of those posts. You can do that with set_do_follow():

from instapy import InstaPy

session = InstaPy(username="<your_username>", password="<your_password>")
session.login()
session.like_by_tags(["bmw", "mercedes"], amount=5)
session.set_dont_like(["naked", "nsfw"])
session.set_do_follow(True, percentage=50)

If you run the script now, then the bot will follow fifty percent of the users whose posts it liked. As usual, every action will be logged.

You can also leave some comments on the posts. There are two things that you need to do. First, enable commenting with set_do_comment():

from instapy import InstaPy

session = InstaPy(username="<your_username>", password="<your_password>")
session.login()
session.like_by_tags(["bmw", "mercedes"], amount=5)
session.set_dont_like(["naked", "nsfw"])
session.set_do_follow(True, percentage=50)
session.set_do_comment(True, percentage=50)

Next, tell the bot what comments to leave with set_comments():

from instapy import InstaPy

session = InstaPy(username="<your_username>", password="<your_password>")
session.login()
session.like_by_tags(["bmw", "mercedes"], amount=5)
session.set_dont_like(["naked", "nsfw"])
session.set_do_follow(True, percentage=50)
session.set_do_comment(True, percentage=50)
session.set_comments(["Nice!", "Sweet!", "Beautiful :heart_eyes:"])

Run the script and the bot will leave one of those three comments on half the posts that it interacts with.

Now that you’re done with the basic settings, it’s a good idea to end the session with end():

from instapy import InstaPy

session = InstaPy(username="<your_username>", password="<your_password>")
session.login()
session.like_by_tags(["bmw", "mercedes"], amount=5)
session.set_dont_like(["naked", "nsfw"])
session.set_do_follow(True, percentage=50)
session.set_do_comment(True, percentage=50)
session.set_comments(["Nice!", "Sweet!", "Beautiful :heart_eyes:"])
session.end()

This will close the browser, save the logs, and prepare a report that you can see in the console output.

Additional Features in InstaPy

InstaPy is a sizable project that has a lot of thoroughly documented features. The good news is that if you’re feeling comfortable with the features you used above, then the rest should feel pretty similar. This section will outline some of the more useful features of InstaPy.

Quota Supervisor

You can’t scrape Instagram all day, every day. The service will quickly notice that you’re running a bot and will ban some of its actions. That’s why it’s a good idea to set quotas on some of your bot’s actions. Take the following for example:

session.set_quota_supervisor(enabled=True, peak_comments_daily=240, peak_comments_hourly=21)

The bot will keep commenting until it reaches its hourly and daily limits. It will resume commenting after the quota period has passed.

Headless Browser

This feature allows you to run your bot without the GUI of the browser. This is super useful if you want to deploy your bot to a server where you may not have or need the graphical interface. It’s also less CPU intensive, so it improves performance. You can use it like so:

session = InstaPy(username='test', password='test', headless_browser=True)

Note that you set this flag when you initialize the InstaPy object.

Using AI to Analyze Posts

Earlier you saw how to ignore posts that contain inappropriate words in their descriptions. What if the description is good but the image itself is inappropriate? You can integrate your InstaPy bot with ClarifAI, which offers image and video recognition services:

session.set_use_clarifai(enabled=True, api_key='<your_api_key>')
session.clarifai_check_img_for(['nsfw'])

Now your bot won’t like or comment on any image that ClarifAI considers NSFW. You get 5,000 free API-calls per month.

Relationship Bounds

It’s often a waste of time to interact with posts by people who have a lot of followers. In such cases, it’s a good idea to set some relationship bounds so that your bot doesn’t waste your precious computing resources:

session.set_relationship_bounds(enabled=True, max_followers=8500)

With this, your bot won’t interact with posts by users who have more than 8,500 followers.

For many more features and configurations in InstaPy, check out the documentation.

Conclusion

InstaPy allows you to automate your Instagram activities with minimal fuss and effort. It’s a very flexible tool with a lot of useful features.

In this tutorial, you learned:

  • How Instagram bots work
  • How to automate a browser with Selenium
  • How to use the Page Object Pattern to make your code more maintainable and testable
  • How to use InstaPy to build a basic Instagram bot

Read the InstaPy documentation and experiment with your bot a little bit. Soon you’ll start getting new followers and likes with a minimal amount of effort. I gained a few new followers myself while writing this tutorial.


Automating Instagram API with Python

Instagram bot with Python

Gain active followers - Algorithm

Maybe some of you do not agree it is a good way to grow your IG page by using follow for follow method but after a lot of researching I found the proper way to use this method.

I have done and used this strategy for a while and my page visits also followers started growing.

The majority of people failing because they randomly targeting the followers and as a result, they are not coming back to your page. So, the key is to find people those have same interests with you.

If you have a programming page go and search for IG pages which have big programming community and once you find one, don’t send follow requests to followers of this page. Because some of them are not active even maybe fake accounts. So, in order to gain active followers, go the last post of this page and find people who liked the post.

Unofficial Instagram API

In order to query data from Instagram I am going to use the very cool, yet unofficial, Instagram API written by Pasha Lev.

**Note:**Before you test it make sure you verified your phone number in your IG account.

The program works pretty well so far but in case of any problems I have to put disclaimer statement here:

Disclaimer: This post published educational purposes only as well as to give general information about Instagram API. I am not responsible for any actions and you are taking your own risk.

Let’s start by installing and then logging in with API.

pip install InstagramApi

from InstagramAPI import InstagramAPI

api = InstagramAPI("username", "password")
api.login()

Once you run the program you will see “Login success!” in your console.

Get users from liked list

We are going to search for some username (your target page) then get most recent post from this user. Then, get users who liked this post. Unfortunately, I can’t find solution how to paginate users so right now it gets about last 500 user.

users_list = []

def get_likes_list(username):
    api.login()
    api.searchUsername(username)
    result = api.LastJson
    username_id = result['user']['pk'] # Get user ID
    user_posts = api.getUserFeed(username_id) # Get user feed
    result = api.LastJson
    media_id = result['items'][0]['id'] # Get most recent post
    api.getMediaLikers(media_id) # Get users who liked
    users = api.LastJson['users']
    for user in users: # Push users to list
        users_list.append({'pk':user['pk'], 'username':user['username']})

Follow Users

Once we get the users list, it is time to follow these users.

IMPORTANT NOTE: set time limit as much as you can to avoid automation detection.

from time import sleep

following_users = []

def follow_users(users_list):
    api.login()
    api.getSelfUsersFollowing() # Get users which you are following
    result = api.LastJson
    for user in result['users']:
        following_users.append(user['pk'])
    for user in users_list:
        if not user['pk'] in following_users: # if new user is not in your following users                   
            print('Following @' + user['username'])
            api.follow(user['pk'])
            # after first test set this really long to avoid from suspension
            sleep(20)
        else:
            print('Already following @' + user['username'])
            sleep(10)

Unfollow Users

This function will look users which you are following then it will check if this user follows you as well. If user not following you then you are unfollowing as well.

follower_users = []

def unfollow_users():
    api.login()
    api.getSelfUserFollowers() # Get your followers
    result = api.LastJson
    for user in result['users']:
        follower_users.append({'pk':user['pk'], 'username':user['username']})

    api.getSelfUsersFollowing() # Get users which you are following
    result = api.LastJson
    for user in result['users']:
        following_users.append({'pk':user['pk'],'username':user['username']})
    for user in following_users:
        if not user['pk'] in follower_users: # if the user not follows you
            print('Unfollowing @' + user['username'])
            api.unfollow(user['pk'])
            # set this really long to avoid from suspension
            sleep(20) 

Full Code with extra functions

Here is the full code of this automation

import pprint
from time import sleep
from InstagramAPI import InstagramAPI
import pandas as pd

users_list = []
following_users = []
follower_users = []

class InstaBot:

    def __init__(self):
        self.api = InstagramAPI("your_username", "your_password")

    def get_likes_list(self,username):
        api = self.api
        api.login()
        api.searchUsername(username) #Gets most recent post from user
        result = api.LastJson
        username_id = result['user']['pk']
        user_posts = api.getUserFeed(username_id)
        result = api.LastJson
        media_id = result['items'][0]['id']

        api.getMediaLikers(media_id)
        users = api.LastJson['users']
        for user in users:
            users_list.append({'pk':user['pk'], 'username':user['username']})
        bot.follow_users(users_list)

    def follow_users(self,users_list):
        api = self.api
        api.login()
        api.getSelfUsersFollowing()
        result = api.LastJson
        for user in result['users']:
            following_users.append(user['pk'])
        for user in users_list:
            if not user['pk'] in following_users:
                print('Following @' + user['username'])
                api.follow(user['pk'])
                # set this really long to avoid from suspension
                sleep(20)
            else:
                print('Already following @' + user['username'])
                sleep(10)

     def unfollow_users(self):
        api = self.api
        api.login()
        api.getSelfUserFollowers()
        result = api.LastJson
        for user in result['users']:
            follower_users.append({'pk':user['pk'], 'username':user['username']})

        api.getSelfUsersFollowing()
        result = api.LastJson
        for user in result['users']:
            following_users.append({'pk':user['pk'],'username':user['username']})

        for user in following_users:
            if not user['pk'] in [user['pk'] for user in follower_users]:
                print('Unfollowing @' + user['username'])
                api.unfollow(user['pk'])
                # set this really long to avoid from suspension
                sleep(20) 

bot =  InstaBot()
# To follow users run the function below
# change the username ('instagram') to your target username
bot.get_likes_list('instagram')

# To unfollow users uncomment and run the function below
# bot.unfollow_users()

it will look like this:

Reverse Python

some extra functions to play with API:

def get_my_profile_details():
    api.login() 
    api.getSelfUsernameInfo()
    result = api.LastJson
    username = result['user']['username']
    full_name = result['user']['full_name']
    profile_pic_url = result['user']['profile_pic_url']
    followers = result['user']['follower_count']
    following = result['user']['following_count']
    media_count = result['user']['media_count']
    df_profile = pd.DataFrame(
        {'username':username,
        'full name': full_name,
        'profile picture URL':profile_pic_url,
        'followers':followers,
        'following':following,
        'media count': media_count,
        }, index=[0])
    df_profile.to_csv('profile.csv', sep='\t', encoding='utf-8')

def get_my_feed():
    image_urls = []
    api.login()
    api.getSelfUserFeed()
    result = api.LastJson
    # formatted_json_str = pprint.pformat(result)
    # print(formatted_json_str)
    if 'items' in result.keys():
        for item in result['items'][0:5]:
            if 'image_versions2' in item.keys():
                image_url = item['image_versions2']['candidates'][1]['url']
                image_urls.append(image_url)

    df_feed = pd.DataFrame({
                'image URL':image_urls
            })
    df_feed.to_csv('feed.csv', sep='\t', encoding='utf-8')


Building an Instagram Bot with Python and Selenium to Gain More Followers

This is image title

Let’s build an Instagram bot to gain more followers! — I know, I know. That doesn’t sound very ethical, does it? But it’s all justified for educational purposes.

Coding is a super power — we can all agree. That’s why I’ll leave it up to you to not abuse this power. And I trust you’re here to learn how it works. Otherwise, you’d be on GitHub cloning one of the countless Instagram bots there, right?

You’re convinced? — Alright, now let’s go back to unethical practices.

The Plan

So here’s the deal, we want to build a bot in Python and Selenium that goes on the hashtags we specify, likes random posts, then follows the posters. It does that enough — we get follow backs. Simple as that.

Here’s a pretty twisted detail though: we want to keep track of the users we follow so the bot can unfollow them after the number of days we specify.

Setup

So first things first, I want to use a database to keep track of the username and the date added. You might as well save/load from/to a file, but we want this to be ready for more features in case we felt inspired in the future.

So make sure you create a database (I named mine instabot — but you can name it anything you like) and create a table called followed_users within the database with two fields (username, date_added)

Remember the installation path. You’ll need it.

You’ll also need the following python packages:

  • selenium
  • mysql-connector

Getting down to it

Alright, so first thing we’ll be doing is creating settings.json. Simply a .json file that will hold all of our settings so we don’t have to dive into the code every time we want to change something.

Settings

settings.json:

{
  "db": {
    "host": "localhost",
    "user": "root",
    "pass": "",
    "database": "instabot"
  },
  "instagram": {
    "user": "",
    "pass": ""
  },
  "config": {
    "days_to_unfollow": 1,
    "likes_over": 150,
    "check_followers_every": 3600,
    "hashtags": []
  }
}

As you can see, under “db”, we specify the database information. As I mentioned, I used “instabot”, but feel free to use whatever name you want.

You’ll also need to fill Instagram info under “instagram” so the bot can login into your account.

“config” is for our bot’s settings. Here’s what the fields mean:

days_to_unfollow: number of days before unfollowing users

likes_over: ignore posts if the number of likes is above this number

check_followers_every: number of seconds before checking if it’s time to unfollow any of the users

hashtags: a list of strings with the hashtag names the bot should be active on

Constants

Now, we want to take these settings and have them inside our code as constants.

Create Constants.py:

import json
INST_USER= INST_PASS= USER= PASS= HOST= DATABASE= POST_COMMENTS= ''
LIKES_LIMIT= DAYS_TO_UNFOLLOW= CHECK_FOLLOWERS_EVERY= 0
HASHTAGS= []

def init():
    global INST_USER, INST_PASS, USER, PASS, HOST, DATABASE, LIKES_LIMIT, DAYS_TO_UNFOLLOW, CHECK_FOLLOWERS_EVERY, HASHTAGS
    # read file
    data = None
    with open('settings.json', 'r') as myfile:
        data = myfile.read()
    obj = json.loads(data)
    INST_USER = obj['instagram']['user']
    INST_PASS = obj['instagram']['pass']
    USER = obj['db']['user']
    HOST = obj['db']['host']
    PASS = obj['db']['pass']
    DATABASE = obj['db']['database']
    LIKES_LIMIT = obj['config']['likes_over']
    CHECK_FOLLOWERS_EVERY = obj['config']['check_followers_every']
    HASHTAGS = obj['config']['hashtags']
    DAYS_TO_UNFOLLOW = obj['config']['days_to_unfollow']

the init() function we created reads the data from settings.json and feeds them into the constants we declared.

Engine

Alright, time for some architecture. Our bot will mainly operate from a python script with an init and update methods. Create BotEngine.py:

import Constants


def init(webdriver):
    return


def update(webdriver):
    return

We’ll be back later to put the logic here, but for now, we need an entry point.

Entry Point

Create our entry point, InstaBot.py:

from selenium import webdriver
import BotEngine

chromedriver_path = 'YOUR CHROMEDRIVER PATH' 
webdriver = webdriver.Chrome(executable_path=chromedriver_path)

BotEngine.init(webdriver)
BotEngine.update(webdriver)

webdriver.close()

chromedriver_path = ‘YOUR CHROMEDRIVER PATH’ webdriver = webdriver.Chrome(executable_path=chromedriver_path)

BotEngine.init(webdriver)
BotEngine.update(webdriver)

webdriver.close()

Of course, you’ll need to swap “YOUR CHROMEDRIVER PATH” with your actual ChromeDriver path.

Time Helper

We need to create a helper script that will help us calculate elapsed days since a certain date (so we know if we should unfollow user)

Create TimeHelper.py:

import datetime


def days_since_date(n):
    diff = datetime.datetime.now().date() - n
    return diff.days

Database

Create DBHandler.py. It’ll contain a class that handles connecting to the Database for us.

import mysql.connector
import Constants
class DBHandler:
    def __init__(self):
        DBHandler.HOST = Constants.HOST
        DBHandler.USER = Constants.USER
        DBHandler.DBNAME = Constants.DATABASE
        DBHandler.PASSWORD = Constants.PASS
    HOST = Constants.HOST
    USER = Constants.USER
    DBNAME = Constants.DATABASE
    PASSWORD = Constants.PASS
    @staticmethod
    def get_mydb():
        if DBHandler.DBNAME == '':
            Constants.init()
        db = DBHandler()
        mydb = db.connect()
        return mydb

    def connect(self):
        mydb = mysql.connector.connect(
            host=DBHandler.HOST,
            user=DBHandler.USER,
            passwd=DBHandler.PASSWORD,
            database = DBHandler.DBNAME
        )
        return mydb

As you can see, we’re using the constants we defined.

The class contains a static method get_mydb() that returns a database connection we can use.

Now, let’s define a DB user script that contains the DB operations we need to perform on the user.

Create DBUsers.py:

import datetime, TimeHelper
from DBHandler import *
import Constants

#delete user by username
def delete_user(username):
    mydb = DBHandler.get_mydb()
    cursor = mydb.cursor()
    sql = "DELETE FROM followed_users WHERE username = '{0}'".format(username)
    cursor.execute(sql)
    mydb.commit()


#add new username
def add_user(username):
    mydb = DBHandler.get_mydb()
    cursor = mydb.cursor()
    now = datetime.datetime.now().date()
    cursor.execute("INSERT INTO followed_users(username, date_added) VALUES(%s,%s)",(username, now))
    mydb.commit()


#check if any user qualifies to be unfollowed
def check_unfollow_list():
    mydb = DBHandler.get_mydb()
    cursor = mydb.cursor()
    cursor.execute("SELECT * FROM followed_users")
    results = cursor.fetchall()
    users_to_unfollow = []
    for r in results:
        d = TimeHelper.days_since_date(r[1])
        if d > Constants.DAYS_TO_UNFOLLOW:
            users_to_unfollow.append(r[0])
    return users_to_unfollow


#get all followed users
def get_followed_users():
    users = []
    mydb = DBHandler.get_mydb()
    cursor = mydb.cursor()
    cursor.execute("SELECT * FROM followed_users")
    results = cursor.fetchall()
    for r in results:
        users.append(r[0])

    return users

Account Agent

Alright, we’re about to start our bot. We’re creating a script called AccountAgent.py that will contain the agent behavior.

Import some modules, some of which we need for later and write a login function that will make use of our webdriver.

Notice that we have to keep calling the sleep function between actions. If we send too many requests quickly, the Instagram servers will be alarmed and will deny any requests you send.

from time import sleep
import datetime
import DBUsers, Constants
import traceback
import random

def login(webdriver):
    #Open the instagram login page
    webdriver.get('https://www.instagram.com/accounts/login/?source=auth_switcher')
    #sleep for 3 seconds to prevent issues with the server
    sleep(3)
    #Find username and password fields and set their input using our constants
    username = webdriver.find_element_by_name('username')
    username.send_keys(Constants.INST_USER)
    password = webdriver.find_element_by_name('password')
    password.send_keys(Constants.INST_PASS)
    #Get the login button
    try:
        button_login = webdriver.find_element_by_xpath(
            '//*[@id="react-root"]/section/main/div/article/div/div[1]/div/form/div[4]/button')
    except:
        button_login = webdriver.find_element_by_xpath(
            '//*[@id="react-root"]/section/main/div/article/div/div[1]/div/form/div[6]/button/div')
    #sleep again
    sleep(2)
    #click login
    button_login.click()
    sleep(3)
    #In case you get a popup after logging in, press not now.
    #If not, then just return
    try:
        notnow = webdriver.find_element_by_css_selector(
            'body > div.RnEpo.Yx5HN > div > div > div.mt3GC > button.aOOlW.HoLwm')
        notnow.click()
    except:
        return

Also note how we’re getting elements with their xpath. To do so, right click on the element, click “Inspect”, then right click on the element again inside the inspector, and choose Copy->Copy XPath.

Another important thing to be aware of is that element hierarchy change with the page’s layout when you resize or stretch the window. That’s why we’re checking for two different xpaths for the login button.

Now go back to BotEngine.py, we’re ready to login.

Add more imports that we’ll need later and fill in the init function

import AccountAgent, DBUsers
import Constants
import datetime


def init(webdriver):
    Constants.init()
    AccountAgent.login(webdriver)


def update(webdriver):
    return

If you run our entry script now (InstaBot.py) you’ll see the bot logging in.

Perfect, now let’s add a method that will allow us to follow people to AccountAgent.py:

def follow_people(webdriver):
    #all the followed user
    prev_user_list = DBUsers.get_followed_users()
    #a list to store newly followed users
    new_followed = []
    #counters
    followed = 0
    likes = 0
    #Iterate theough all the hashtags from the constants
    for hashtag in Constants.HASHTAGS:
        #Visit the hashtag
        webdriver.get('https://www.instagram.com/explore/tags/' + hashtag+ '/')
        sleep(5)

        #Get the first post thumbnail and click on it
        first_thumbnail = webdriver.find_element_by_xpath(
            '//*[@id="react-root"]/section/main/article/div[1]/div/div/div[1]/div[1]/a/div')

        first_thumbnail.click()
        sleep(random.randint(1,3))

        try:
            #iterate over the first 200 posts in the hashtag
            for x in range(1,200):
                t_start = datetime.datetime.now()
                #Get the poster's username
                username = webdriver.find_element_by_xpath('/html/body/div[3]/div[2]/div/article/header/div[2]/div[1]/div[1]/h2/a').text
                likes_over_limit = False
                try:
                    #get number of likes and compare it to the maximum number of likes to ignore post
                    likes = int(webdriver.find_element_by_xpath(
                        '/html/body/div[3]/div[2]/div/article/div[2]/section[2]/div/div/button/span').text)
                    if likes > Constants.LIKES_LIMIT:
                        print("likes over {0}".format(Constants.LIKES_LIMIT))
                        likes_over_limit = True


                    print("Detected: {0}".format(username))
                    #If username isn't stored in the database and the likes are in the acceptable range
                    if username not in prev_user_list and not likes_over_limit:
                        #Don't press the button if the text doesn't say follow
                        if webdriver.find_element_by_xpath('/html/body/div[3]/div[2]/div/article/header/div[2]/div[1]/div[2]/button').text == 'Follow':
                            #Use DBUsers to add the new user to the database
                            DBUsers.add_user(username)
                            #Click follow
                            webdriver.find_element_by_xpath('/html/body/div[3]/div[2]/div/article/header/div[2]/div[1]/div[2]/button').click()
                            followed += 1
                            print("Followed: {0}, #{1}".format(username, followed))
                            new_followed.append(username)


                        # Liking the picture
                        button_like = webdriver.find_element_by_xpath(
                            '/html/body/div[3]/div[2]/div/article/div[2]/section[1]/span[1]/button')

                        button_like.click()
                        likes += 1
                        print("Liked {0}'s post, #{1}".format(username, likes))
                        sleep(random.randint(5, 18))


                    # Next picture
                    webdriver.find_element_by_link_text('Next').click()
                    sleep(random.randint(20, 30))
                    
                except:
                    traceback.print_exc()
                    continue
                t_end = datetime.datetime.now()

                #calculate elapsed time
                t_elapsed = t_end - t_start
                print("This post took {0} seconds".format(t_elapsed.total_seconds()))


        except:
            traceback.print_exc()
            continue

        #add new list to old list
        for n in range(0, len(new_followed)):
            prev_user_list.append(new_followed[n])
        print('Liked {} photos.'.format(likes))
        print('Followed {} new people.'.format(followed))

It’s pretty long, but generally here’s the steps of the algorithm:

For every hashtag in the hashtag constant list:

  • Visit the hashtag link
  • Open the first thumbnail
  • Now, execute the following code 200 times (first 200 posts in the hashtag)
  • Get poster’s username, check if not already following, follow, like the post, then click next
  • If already following just click next quickly

Now we might as well implement the unfollow method, hopefully the engine will be feeding us the usernames to unfollow in a list:

def unfollow_people(webdriver, people):
    #if only one user, append in a list
    if not isinstance(people, (list,)):
        p = people
        people = []
        people.append(p)

    for user in people:
        try:
            webdriver.get('https://www.instagram.com/' + user + '/')
            sleep(5)
            unfollow_xpath = '//*[@id="react-root"]/section/main/div/header/section/div[1]/div[1]/span/span[1]/button'

            unfollow_confirm_xpath = '/html/body/div[3]/div/div/div[3]/button[1]'

            if webdriver.find_element_by_xpath(unfollow_xpath).text == "Following":
                sleep(random.randint(4, 15))
                webdriver.find_element_by_xpath(unfollow_xpath).click()
                sleep(2)
                webdriver.find_element_by_xpath(unfollow_confirm_xpath).click()
                sleep(4)
            DBUsers.delete_user(user)

        except Exception:
            traceback.print_exc()
            continue

Now we can finally go back and finish the bot by implementing the rest of BotEngine.py:

import AccountAgent, DBUsers
import Constants
import datetime


def init(webdriver):
    Constants.init()
    AccountAgent.login(webdriver)


def update(webdriver):
    #Get start of time to calculate elapsed time later
    start = datetime.datetime.now()
    #Before the loop, check if should unfollow anyone
    _check_follow_list(webdriver)
    while True:
        #Start following operation
        AccountAgent.follow_people(webdriver)
        #Get the time at the end
        end = datetime.datetime.now()
        #How much time has passed?
        elapsed = end - start
        #If greater than our constant to check on
        #followers, check on followers
        if elapsed.total_seconds() >= Constants.CHECK_FOLLOWERS_EVERY:
            #reset the start variable to now
            start = datetime.datetime.now()
            #check on followers
            _check_follow_list(webdriver)


def _check_follow_list(webdriver):
    print("Checking for users to unfollow")
    #get the unfollow list
    users = DBUsers.check_unfollow_list()
    #if there's anyone in the list, start unfollowing operation
    if len(users) > 0:
        AccountAgent.unfollow_people(webdriver, users)

Conclusion

And that’s it — now you have yourself a fully functional Instagram bot built with Python and Selenium. There are many possibilities for you to explore now, so make sure you’re using this newly gained skill to solve real life problems!

You can get the source code for the whole project from this GitHub repository.


Building a simple Instagram bot with Python tutorial

Here we build a simple bot using some simple Python which beginner to intermediate coders can follow.

Here’s the code on GitHub
https://github.com/aj-4/ig-followers


Build A (Full-Featured) Instagram Bot With Python

Source Code: https://github.com/jg-fisher/instagram-bot 


How to Get Instagram Followers/Likes Using Python

In this video I show you how to program your own Instagram Bot using Python and Selenium.

https://www.youtube.com/watch?v=BGU2X5lrz9M 

Code Link:

from selenium import webdriver
from selenium.webdriver.common.keys import Keys
import time
import random
import sys


def print_same_line(text):
    sys.stdout.write('\r')
    sys.stdout.flush()
    sys.stdout.write(text)
    sys.stdout.flush()


class InstagramBot:

    def __init__(self, username, password):
        self.username = username
        self.password = password
        self.driver = webdriver.Chrome()

    def closeBrowser(self):
        self.driver.close()

    def login(self):
        driver = self.driver
        driver.get("https://www.instagram.com/")
        time.sleep(2)
        login_button = driver.find_element_by_xpath("//a[@href='/accounts/login/?source=auth_switcher']")
        login_button.click()
        time.sleep(2)
        user_name_elem = driver.find_element_by_xpath("//input[@name='username']")
        user_name_elem.clear()
        user_name_elem.send_keys(self.username)
        passworword_elem = driver.find_element_by_xpath("//input[@name='password']")
        passworword_elem.clear()
        passworword_elem.send_keys(self.password)
        passworword_elem.send_keys(Keys.RETURN)
        time.sleep(2)


    def like_photo(self, hashtag):
        driver = self.driver
        driver.get("https://www.instagram.com/explore/tags/" + hashtag + "/")
        time.sleep(2)

        # gathering photos
        pic_hrefs = []
        for i in range(1, 7):
            try:
                driver.execute_script("window.scrollTo(0, document.body.scrollHeight);")
                time.sleep(2)
                # get tags
                hrefs_in_view = driver.find_elements_by_tag_name('a')
                # finding relevant hrefs
                hrefs_in_view = [elem.get_attribute('href') for elem in hrefs_in_view
                                 if '.com/p/' in elem.get_attribute('href')]
                # building list of unique photos
                [pic_hrefs.append(href) for href in hrefs_in_view if href not in pic_hrefs]
                # print("Check: pic href length " + str(len(pic_hrefs)))
            except Exception:
                continue

        # Liking photos
        unique_photos = len(pic_hrefs)
        for pic_href in pic_hrefs:
            driver.get(pic_href)
            time.sleep(2)
            driver.execute_script("window.scrollTo(0, document.body.scrollHeight);")
            try:
                time.sleep(random.randint(2, 4))
                like_button = lambda: driver.find_element_by_xpath('//span[@aria-label="Like"]').click()
                like_button().click()
                for second in reversed(range(0, random.randint(18, 28))):
                    print_same_line("#" + hashtag + ': unique photos left: ' + str(unique_photos)
                                    + " | Sleeping " + str(second))
                    time.sleep(1)
            except Exception as e:
                time.sleep(2)
            unique_photos -= 1

if __name__ == "__main__":

    username = "USERNAME"
    password = "PASSWORD"

    ig = InstagramBot(username, password)
    ig.login()

    hashtags = ['amazing', 'beautiful', 'adventure', 'photography', 'nofilter',
                'newyork', 'artsy', 'alumni', 'lion', 'best', 'fun', 'happy',
                'art', 'funny', 'me', 'followme', 'follow', 'cinematography', 'cinema',
                'love', 'instagood', 'instagood', 'followme', 'fashion', 'sun', 'scruffy',
                'street', 'canon', 'beauty', 'studio', 'pretty', 'vintage', 'fierce']

    while True:
        try:
            # Choose a random tag from the list of tags
            tag = random.choice(hashtags)
            ig.like_photo(tag)
        except Exception:
            ig.closeBrowser()
            time.sleep(60)
            ig = InstagramBot(username, password)
            ig.login()

Build An INSTAGRAM Bot With Python That Gets You Followers


Instagram Automation Using Python


How to Create an Instagram Bot | Get More Followers


Building a simple Instagram Influencer Bot with Python tutorial

#python #chatbot #web-development