HeenaR17

diameter(g1, directed=F, weights=NA)
edge_density(g1,loops=F)
reciprocity(g1)
closeness(g1, mode="all", weights=NA)
betweenness(g1, directed=T, weight=NA)
edge_betweenness(g1)

HeenaR17

degree(g1, mode="all")
degree(g1, mode="in")
degree(g1, mode="out")

HeenaR17

#Network measures
degree(g1)

HeenaR17

g1 <- graph(c("Heena","Tina","Tina","Disha","Disha","Heena","Heena","Disha","Li","Disha"))
plot(g1,
     vertex.color="green", 
     vertex.size=40, 
     edge.color="red")
g1

HeenaR17

library(igraph)
g <- graph(c(1,2,2,3,3,4,4,1))
plot(g, 
     vertex.color="green", 
     vertex.size=40, 
     edge.color="red")
g[]

HeenaR17

model = T5Model("t5-small", args=model_args)
model.train_model(paraphrase_train_df, eval_data=paraphrase_dev_df)

HeenaR17

model_args = {
    "reprocess_input_data": True,
    "overwrite_output_dir": True,
    "max_seq_length": 128,
    "train_batch_size": 16,
    "num_train_epochs": 25,
    "num_beams": None,
    "do_sample": True,
    "max_length": 20,
    "top_k": 50,

HeenaR17

files.upload()
pd.set_option('display.max_colwidth', None)
df = pd.read_csv('train.tsv',sep='\t')
df.head(5)
df.describe()
paraphrase_train_df = df[df['label']==1]
paraphrase_train_df.head(5)
paraphrase_train_df.describe()
paraphrase_train_df["prefix"] = "generate paraphrase"
paraphrase_train_df = paraphrase_train_df.rename(columns={"sentence1": "input_text", "sentence2": "target_text"})

HeenaR17

import pandas as pd
from google.colab import files
from simpletransformers.t5 import T5Model
from pprint import pprint
import logging
logging.basicConfig(level=logging.ERROR)
transformers_logger = logging.getLogger("transformers")
transformers_logger.setLevel(logging.ERROR)
%env WANDB_DISABLED=True

HeenaR17

german_to_english = torch.hub.load('pytorch/fairseq', 'transformer.wmt19.de-en.single_model', tokenizer='moses', bpe='fastbpe')
data = ["back translation is one of the best data augmentation techniques"]

def augment_data(data, x_to_y, y_to_x, n):
  augmented_data = dict()
  for d in data:
    augmented_data[d] = list()
    y_result = x_to_y.generate(x_to_y.encode(d), beam=n)
    for y in y_result:
      x_result = y_to_x.generate(y_to_x.encode(x_to_y.decode(y['tokens'])), beam=n)