auto_code_complete is a auto word-completetion program which allows you to customize it on your need

Last update: Feb 22, 2022

Overview

auto_code_complete v1.3

purpose and usage

auto_code_complete is a auto word-completetion program which allows you to customize it on your needs. the model for this program is a combined model of a deep-learning NLP(Natural Language Process) model structure called 'GRU(gated recurrent unit)' and 'LSTM(Long Short Term Memory)'.

the model for this program is one of the deep-learning NLP(Natural Language Process) model structure called 'GRU(gated recurrent unit)'.

data preprocessing

model structure

how to use (terminal)

first, download the repository on your local environment.
install the neccessary libraries on your dependent environment.

pip install -r requirements.txt

change your working directory to auto-complete/ and execute the line below

python -m auto_complete_model

it will require for you to enter the data you want to train with the model

ENTER THE CODE YOU WANT TO TRAIN IN YOUR MODEL : tensorflow tf.keras tf.keras.layers LSTM
==== TRAINING START ====
2022-01-08 18:24:14.308919: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz
Epoch 1/100
3/3 [==============================] - 1s 59ms/step - loss: 4.7865 - acc: 0.0532
Epoch 2/100
3/3 [==============================] - 0s 62ms/step - loss: 3.9297 - acc: 0.2872
Epoch 3/100
3/3 [==============================] - 0s 58ms/step - loss: 2.9941 - acc: 0.5532
...
Epoch 31/100
3/3 [==============================] - 0s 75ms/step - loss: 0.2747 - acc: 0.8617
Epoch 32/100
3/3 [==============================] - 0s 65ms/step - loss: 0.2700 - acc: 0.8298
==== TRAINING DONE ====
Now, Load the best weights on your model.

if you input your dataset successfully, it will ask for any uncompleted word to be entered.

ENTER THE UNCOMPLETED CODE YOU WANT TO COMPLETE : t tf te l la li k ke tf.kera tf.keras.l
t  - best recommendation : tensorflow
		 - all recommendations :  ['tensorflow']
tf  - best recommendation : tf.keras
		 - all recommendations :  ['tfkeras', 'tf.keras']
te  - best recommendation : tensorflow
		 - all recommendations :  ['tensorflow']
l  - best recommendation : list
		 - all recommendations :  ['list', 'layers']
la  - best recommendation : lange
		 - all recommendations :  ['layers', 'lange']
li  - best recommendation : list
		 - all recommendations :  ['list']
k  - best recommendation : keras
		 - all recommendations :  ['keras']
ke  - best recommendation : keras
		 - all recommendations :  ['keras']
tf.kera  - best recommendation : tf.keras
		 - all recommendations :  []
tf.keras.l  - best recommendation : tf.keras.layers
		 - all recommendations :  ['tf.keras.layers']

it will return the best matched word to complete and other recommendations

Do you want to check only the recommendations? (y/n) : y
['tensorflow'], 
['tfkeras', 'tf.keras'], 
['tensorflow'], 
['list', 'layers'], 
['layers', 'lange'], 
['list'], 
['keras'], 
['keras'], 
[], 
['tf.keras.layers']

version update & issues

v1.2 update

2022.01.08

change deep-learning model from GRU to GRU+LSTM to improve the performance

By adding the same structrue of new LSTM layers to concatenate before the output layer to an existing model, it shows faster learning and better accuracies in predicting matched recommendations for given incomplete words.

v1.3.1 update

2022.01.09

fix the glitches in data preprocessing

We solved the problem that it wouldn't add a new dataset on an existing dataset.

add plot_history function in a model class

v1.3.2 update

2022.01.10

add model_save,model_load mode in order that users can save and load their model while training a customized model
add data_split mode so that the big data can be trained seperately.

samp_model = auto_coding(new_code=samp_text,
                      # verbose=0,
                       batch_size=100,
                       epochs=200,
                       patience=10,
                       model_summary=True,
                       model_save=True,
                       model_name='samp_test', # samp_test/samp_test.h5
                       model_load=True,
                       data_split=True,
                       data_split_num=3 # the number into which users want to split the data
                      )

v1.3.3 update

2022.01.11

add new metrics Accuracy for Recommendations to evaluate the model's instant performance when predicting the recommendation list for words.

t  - best match : tf
	 - all recommendations :  ['tensorflow', 'tf']
tup  - best match : tuple
	 - all recommendations :  []
p  - best match : pd
	 - all recommendations :  ['plt', 'pd', 'pandas']
li  - best match : list
	 - all recommendations :  []
d  - best match : dataset
	 - all recommendations :  ['dic', 'dataset']
I  - best match : Import
	 - all recommendations :  []
so  - best match : sort
	 - all recommendations :  ['sort']
m  - best match : matplotlib.pyplot
	 - all recommendations :  []
Accuracy for Best:  0.875
Accuracy for Recommendations :  1.0

auto_code_complete is a auto word-completetion program which allows you to customize it on your need

Related tags

Overview

auto_code_complete v1.3

purpose and usage

data preprocessing

model structure

how to use (terminal)

version update & issues

v1.2 update

v1.3.1 update

v1.3.2 update

v1.3.3 update

Owner

RUO

Ongoing research training transformer language models at scale, including: BERT & GPT-2

Code Implementation of "Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction".

Code for our paper "Mask-Align: Self-Supervised Neural Word Alignment" in ACL 2021

code for modular summarization work published in ACL2021 by Krishna et al

Repositório da disciplina no semestre 2021-2

Converts text into a PDF of handwritten notes

Bnagla hand written document digiiztion

The swas programming language

A fast and easy implementation of Transformer with PyTorch.

Count the frequency of letters or words in a text file and show a graph.

Line as a Visual Sentence: Context-aware Line Descriptor for Visual Localization

A raytrace framework using taichi language

xFormers is a modular and field agnostic library to flexibly generate transformer architectures by interoperable and optimized building blocks.

An open source library for deep learning end-to-end dialog systems and chatbots.

Transformer-based Text Auto-encoder (T-TA) using TensorFlow 2.

Source code of the "Graph-Bert: Only Attention is Needed for Learning Graph Representations" paper

Named Entity Recognition API used by TEI Publisher

Universal Adversarial Triggers for Attacking and Analyzing NLP (EMNLP 2019)

In this Notebook I've build some machine-learning and deep-learning to classify corona virus tweets, in both multi class classification and binary classification.

Kestrel Threat Hunting Language