Name and Address Parsing
Recognising Person names and Addresses in a text using NER and NLP modeling techniques
Get data from MongoDB
!pip install pymongo[tls,srv]
import pandas as pd
from tqdm import tqdm
tqdm.pandas()
%reload_ext google.colab.data_table
import pymongo
import pprint
mongo_uri = "mongodb+srv://<userid>:<pass>@<server>.azure.mongodb.net/<db>?retryWrites=true&w=majority"
client = pymongo.MongoClient(mongo_uri)
listdb = client.list_database_names(); listdb
['testdb']
db = client['testdb']
db.list_collection_names()
['entities', 'parcels', 'taxes']
# print(db.command("collstats", "events"))
db.command("dbstats")
{'$clusterTime': {'clusterTime': Timestamp(1598254711, 1),
'signature': {'hash': b'\x19\x92\xd6\x94\x06\xed\xa9\xb5s+\x8e\xcaqn\xb3J\x13_\x1e\xab',
'keyId': 6839689286636273667}},
'avgObjSize': 597.6787450874097,
'collections': 3,
'dataSize': 5494893423.0,
'db': 'testdb',
'fsTotalSize': 34342961152.0,
'fsUsedSize': 11420770304.0,
'indexSize': 378994688.0,
'indexes': 9,
'numExtents': 0,
'objects': 9193724,
'ok': 1.0,
'operationTime': Timestamp(1598254711, 1),
'scaleFactor': 1.0,
'storageSize': 2160832512.0,
'views': 0}
# collection1 = db['entities']
# df1 = pd.DataFrame(list(collection1.find()))
# print(df1.info())
# # print(df1.describe())
# df1.sample(5)
# df1.to_pickle('/content/drive/My Drive/df_entities.p')
# del df1
# import gc
# gc.collect()
258
collection2 = db['parcels']
cursor = collection2.find()
max = 0
maxL = {}
for i in range(cnt):
xx = next(cursor)
if len(xx)>max:
max = len(xx)
maxL = xx
max
46
cnt = collection2.count()
list(maxL)
4597
collection2 = db['parcels']
cursor = collection2.find()
XX = pd.DataFrame(columns=list(maxL))
XX.to_csv('df_parcels.csv', index=False)
for i in tqdm(range((cnt//10000)+1)):
YY = {}
for j in range(10000):
YY[j] = next(cursor)
pd.DataFrame(YY).T.to_csv('df_parcels.csv', mode='a', index=False, header=False)
# pd.DataFrame(next(cursor), index=[j]).to_csv('df_parcels.csv', mode='a', index=False, header=False)
33%|███▎ | 150/460 [04:52<10:09, 1.97s/it]
pd.read_csv('df_parcels.csv')
collection2 = db['parcels']
df2 = pd.DataFrame(list(collection2.find()))
print(df2.info())
# print(df2.describe())
# df2.sample(5)
Data wrangling
!pip install -q probablepeople
!pip install -q usaddress
import os
import re
import numpy as np
import pandas as pd
from tqdm import tqdm
import seaborn as sns
import matplotlib.pyplot as plt
import probablepeople as pp
import usaddress as ua
import nltk
nltk.download('punkt')
nltk.download('averaged_perceptron_tagger')
nltk.download('maxent_ne_chunker')
nltk.download('words')
import warnings
warnings.filterwarnings("ignore")
tqdm.pandas()
%reload_ext autoreload
%autoreload 2
%reload_ext google.colab.data_table
%config InlineBackend.figure_format = 'retina'
plt.style.use('fivethirtyeight')
plt.style.use('seaborn-notebook')
df = pd.read_pickle(os.path.join(path,'sample.p'))
df.info()
data = df.oor.tolist()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 1000 entries, 3605089 to 2516200
Data columns (total 2 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 _id 1000 non-null object
1 oor 1000 non-null object
dtypes: object(2)
memory usage: 23.4+ KB
data[:10]
combos = [' US ',' USA ',' U.S. ',' U.S.A. ']
def patch_country(text):
cflagz = ''
otext = text
text = text[-10:]
text = ' ' + text + ' '
for i in combos:
if i in text:
cflagz = i
text = otext[:-10] + re.sub(cflagz, '', text)
return text, cflagz
def patch_household(xz):
xx = pd.DataFrame(pp.tag(xz)[0], index=[0])
xx['Type'] = 'Person'
if 'And' in xx.columns.tolist():
xx = pd.DataFrame({'Household':xz}, index=[0])
xx['Type'] = 'Household'
return xx
tags = pp.LABELS
tags.extend(ua.LABELS)
tags = list(set(tags))
additionals = ['CountryName', 'Household', 'Type', 'NameConfidence', 'AddrConfidence', 'Text']
tags.extend(additionals)
dfx = pd.DataFrame(columns=tags)
dfx = dfx.loc[:,~dfx.columns.duplicated()]
# def temp(text):
# # replace long country name with short one
# text = re.sub('UNITED STATES OF AMERICA','US',text)
# # country patch
# text, cflag = patch_country(text)
# # try and catch
# try:
# df1 = pd.DataFrame(ua.tag(text)[0], index=[0])
# xz = df1.Recipient.values[0]
# df2 = patch_household(xz)
# xx = pd.concat([df1,df2], axis=1)
# xx['Confidence'] = ua.tag(text)[1]
# except:
# try:
# df2 = patch_household(text)
# xx = df2
# xx['Confidence'] = 'AddressError'
# except:
# xx = pd.DataFrame({'Recipient':text}, index=[0])
# xx['Confidence'] = 'PersonError'
# # add country label
# if cflag!='':
# xx['CountryName'] = cflag
# return xx#.T.to_dict()[0]
def temp(text):
# country patch
text = re.sub('UNITED STATES OF AMERICA','US',text)
text, cflag = patch_country(text)
# address parsing
try:
df1 = pd.DataFrame(ua.tag(text)[0], index=[0])
df1['AddrConfidence'] = ua.tag(text)[1]
except:
df1 = pd.DataFrame(ua.parse(text)).groupby(1).agg({0: lambda x: ' '.join(x)}).T
df1['AddrConfidence'] = 'Error'
# address to name linking
try:
xz = df1.Recipient.values[0]
except:
xz = text
# name parsing
try:
df2 = pd.DataFrame(pp.tag(xz)[0], index=[0])
df2['NameConfidence'] = pp.tag(xz)[1]
except:
df2 = pd.DataFrame(pp.parse(xz)).groupby(1).agg({0: lambda x: ' '.join(x)}).T
df2['NameConfidence'] = 'Error'
# person name patch
df2['Type'] = 'Person'
if 'MiddleName' in df2.columns.tolist():
gname, mname, sname = df2.MiddleName, df2.Surname, df2.GivenName
df2.MiddleName, df2.Surname, df2.GivenName = mname[0], sname[0], gname[0]
elif 'GivenName' in df2.columns.tolist():
gname, sname = df2.Surname, df2.GivenName
df2.Surname, df2.GivenName = sname[0], gname[0]
# household name patch
if 'And' in df2.columns.tolist():
df2 = pd.DataFrame({'Household':xz}, index=[0])
df2['Type'] = 'Household'
# concatenation
df = pd.concat([df1,df2], axis=1)
df['Text'] = text
df['CountryName'] = cflag if cflag!='' else np.NaN
try:
df.loc[pd.notnull(df['CorporationName']), 'Type'] = 'Corporation'
except:
pass
return df
errs = {}
dfx = pd.DataFrame(columns=tags)
dfx = dfx.loc[:,~dfx.columns.duplicated()]
for idx, text in tqdm(enumerate(data)):
try:
dfx = dfx.append(temp(text))
dfx.loc[pd.notnull(dfx['CorporationName']), 'Type'] = 'Corporation'
except Exception as e:
errs[idx] = e
pass
dfx = dfx.fillna('')
1000it [00:18, 52.66it/s]
err_df = pd.DataFrame(errs, index=[0]).T
err_df = err_df.merge(pd.Series(data, name='text'), left_index=True, right_index=True)
err_df.columns = ['error','text']
err_df = err_df[['text','error']]
err_df.iloc[:]
dfbackup = dfx.copy()
dfx = dfbackup[tags]
dfx.drop(['Recipient', 'CountryName'], axis=1, inplace=True)
dfx = dfx.replace(r'^\s*$', np.nan, regex=True).dropna(axis=1, how='all')
dfx.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 989 entries, 0 to 0
Data columns (total 39 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 USPSBoxType 59 non-null object
1 CorporationCommitteeType 1 non-null object
2 SubaddressType 18 non-null object
3 ZipCode 703 non-null object
4 GivenName 417 non-null object
5 StreetNamePreDirectional 86 non-null object
6 CorporationNameAndCompany 4 non-null object
7 StreetName 906 non-null object
8 CorporationName 381 non-null object
9 PlaceName 888 non-null object
10 StateName 807 non-null object
11 ShortForm 1 non-null object
12 Surname 407 non-null object
13 CorporationLegalType 106 non-null object
14 StreetNamePostType 863 non-null object
15 PrefixOther 7 non-null object
16 CorporationNameBranchType 4 non-null object
17 AddressNumber 902 non-null object
18 USPSBoxID 59 non-null object
19 MiddleInitial 152 non-null object
20 AddressNumberSuffix 1 non-null object
21 CorporationNameBranchIdentifier 9 non-null object
22 SuffixOther 3 non-null object
23 LandmarkName 2 non-null object
24 CorporationNameOrganization 57 non-null object
25 BuildingName 52 non-null object
26 OccupancyIdentifier 227 non-null object
27 MiddleName 33 non-null object
28 StreetNamePreType 16 non-null object
29 StreetNamePostDirectional 15 non-null object
30 OccupancyType 170 non-null object
31 FirstInitial 1 non-null object
32 SubaddressIdentifier 19 non-null object
33 SuffixGenerational 8 non-null object
34 Household 192 non-null object
35 Type 989 non-null object
36 NameConfidence 797 non-null object
37 AddrConfidence 989 non-null object
38 Text 986 non-null object
dtypes: object(39)
memory usage: 309.1+ KB
def func(text):
text = str(text)
xx = temp(text)
xx = xx.drop(['AddrConfidence','NameConfidence','Text'], axis=1)
try:
xx = xx.drop(['Recipient'], axis=1)
except:
pass
xx = xx.T.to_dict()[0]
return str(xx)
# !pip install -q gradio
# import gradio as gr
gr.Interface(fn=func, inputs="text", outputs="text").launch()
Sampling
path = '.'
import os
import re
import numpy as np
import pandas as pd
from tqdm import tqdm
import warnings
tqdm.pandas()
warnings.filterwarnings("ignore")
# df = pd.read_pickle(os.path.join(path,'df_entities.p'))
# df.info()
# def process_text(text):
# text = re.sub('"','',text)
# text = re.sub(',','',text)
# text = ' '.join(text.split())
# return text
# sampledf = df[['oor']].sample(5000, random_state=40)
# sampledf['oor'] = sampledf.oor.apply(process_text)
# sampledf = sampledf.replace(r'^\s*$', np.nan, regex=True).dropna()
# msk = np.random.rand(len(sampledf)) < 0.8
# sampledf[msk].to_csv('train_28118_21.csv', index=False, header=None)
# sampledf[~msk].to_csv('test_28118_21.csv', index=False, header=None)
!pip install doccano-transformer
import spacy
import random
from doccano_transformer.datasets import NERDataset
from doccano_transformer.utils import read_jsonl
dataset = read_jsonl(filepath='file.json1', dataset=NERDataset, encoding='utf-8')
dataset.to_spacy(tokenizer=str.split)
TRAIN_DATA = []
for x in dataset:
xx = {}
try:
xx['entities'] = list(x.labels.values())[0]
TRAIN_DATA.append((x.text.lower(), xx))
except:
pass
def train_spacy(data, iterations, model=None):
TRAIN_DATA = data
if model is not None:
nlp = spacy.load(model)
print("Loaded model '%s'" % model)
else:
nlp = spacy.blank("en")
print("Created blank 'en' model")
if 'ner' not in nlp.pipe_names:
ner = nlp.create_pipe('ner')
nlp.add_pipe(ner, last=True)
else:
ner = nlp.get_pipe("ner")
# add labels
for _, annotations in TRAIN_DATA:
for ent in annotations.get('entities'):
ner.add_label(ent[2])
# get names of other pipes to disable them during training
other_pipes = [pipe for pipe in nlp.pipe_names if pipe != 'ner']
with nlp.disable_pipes(*other_pipes): # only train NER
optimizer = nlp.begin_training()
for itn in range(iterations):
print("Statring iteration " + str(itn))
random.shuffle(TRAIN_DATA)
losses = {}
for text, annotations in TRAIN_DATA:
nlp.update(
[text], # batch of texts
[annotations], # batch of annotations
drop=0.1, # dropout - make it harder to memorise data
sgd=optimizer, # callable to update weights
losses=losses)
print(losses)
return nlp
prdnlp = train_spacy(TRAIN_DATA, 200)
Created blank 'en' model
Statring iteration 0
{'ner': 3007.4073503919476}
Statring iteration 1
{'ner': 1640.818005657745}
Statring iteration 2
{'ner': 1289.2707772499139}
Statring iteration 3
{'ner': 1141.5257107774235}
Statring iteration 4
{'ner': 915.537918427113}
Statring iteration 5
{'ner': 880.0738022538334}
Statring iteration 6
{'ner': 697.9718340145506}
Statring iteration 7
{'ner': 477.89584190603824}
...
Statring iteration 192
{'ner': 81.44864770046895}
Statring iteration 193
{'ner': 78.21613277818398}
Statring iteration 194
{'ner': 62.061153309127164}
Statring iteration 195
{'ner': 62.953992718801345}
Statring iteration 196
{'ner': 53.10038627353098}
Statring iteration 197
{'ner': 98.58842041503256}
Statring iteration 198
{'ner': 71.83606691682377}
Statring iteration 199
{'ner': 72.84951195810225}
# Save our trained Model
# modelfile = input("Enter your Model Name: ")
# prdnlp.to_disk(modelfile)
prdnlp.to_disk('ner233')
import shutil
shutil.make_archive(os.path.join(path,'ner233'),'zip','/content/ner233')
# test = pd.read_csv('/content/test_28118_21.csv', header=None)
# test.head()
# from spacy.gold import GoldParse
# from spacy.scorer import Scorer
# def evaluate(ner_model, examples):
# scorer = Scorer()
# for input_, annot in examples:
# doc_gold_text = ner_model.make_doc(input_)
# gold = GoldParse(doc_gold_text, entities=annot)
# pred_value = ner_model(input_)
# scorer.score(pred_value, gold)
# return scorer.scores
# examples = [
# ('Who is Shaka Khan?',
# [(7, 17, 'PERSON')]),
# ('I like London and Berlin.',
# [(7, 13, 'LOC'), (18, 24, 'LOC')])
# ]
# results = evaluate(prdnlp, examples)
# def report_scores(scores, e):
# """
# prints precision recall and f_measure
# :param scores:
# :return:
# """
# precision = '%.2f' % scores['ents_p']
# recall = '%.2f' % scores['ents_r']
# f_measure = '%.2f' % scores['ents_f']
# print('%-25s %-10s %-10s %-10s' % (e, precision, recall, f_measure))
# report_scores(results, 'x')
x 0.00 0.00 0.00
# Household -> Pass again for name tags
tags = ['EntityType','Recipient','Address',
'GivenName','MiddleName','SurName','Household','Corporation',
'StreetAddress','City','State','Zipcode','Country']
PAD_TAIL = '<redacted>'
def household_patch(row):
if row.EntityType=='Household':
text = row.Household.split('&')[0] if '&' in row.Household else ' '.join(row.Household.split()[:2])
text = parseit(text+PAD_TAIL)
row['GivenName'] = text.loc[0,'GivenName']
row['SurName'] = text.loc[0,'SurName']
row['MiddleName'] = text.loc[0,'MiddleName']
return row
def add_on(xx):
# Address is the combination of street, city, state, zip and country
xx['Address'] = xx['StreetAddress'] +' '+ xx['City'] +' '+ xx['State'] +' '+ xx['Zipcode'] +' '+ xx['Country']
# default is person, if household column is not empty, then household, same for corporation
xx['EntityType'] = 'Person'
xx.loc[xx.Household!='','EntityType'] = 'Household'
xx.loc[xx.Corporation!='','EntityType'] = 'Corporation'
# default is full name of person, if entity is corporation ,then corporation name, same for household
xx['Recipient'] = xx['SurName'] +' '+ xx['GivenName'] +' '+ xx['MiddleName']
xx['Recipient'] = xx.apply(lambda row: row.Corporation if row.EntityType=='Corporation' else row.Recipient, axis=1)
xx['Recipient'] = xx.apply(lambda row: row.Household if row.EntityType=='Household' else row.Recipient, axis=1)
# adding household to name field patch
xx = xx.apply(household_patch, axis=1)
# converting to dictionary format
xx = xx.replace(r'^\s*$', np.nan, regex=True).dropna(axis=1, how='any')
xx = xx.T.to_dict()[0]
# return the processed data
return xx
def parseit(text):
text = str(text).upper()
text = re.sub('"','',text)
text = re.sub(',','',text)
text = ' '.join(text.split())
output = {}
doc = prdnlp(text)
for ent in doc.ents:
output[ent.label_] = ent.text
df = pd.DataFrame(output, index=[0])
dfx = pd.DataFrame(columns=tags)
dfx = dfx.append(df).fillna('')
return dfx
def func(text):
X = parseit(text)
X = add_on(X)
X = str(X)
return X
# text = test[0].sample().tolist()[0]
text = '<redacted>'
# xx = parseit(text); xx
# xx = pd.DataFrame(xx, index=[0]).T.reset_index(); xx.columns = ['Tag','Entity']; xx
func('<redacted>')
# !pip install -q gradio
# import gradio as gr
# inputs = gr.inputs.Textbox(lines=3, label='Input')
# outputs = gr.outputs.Textbox(label='Output')
gr.Interface(fn=func, inputs=inputs, outputs=outputs).launch()
Flair NER Model
!pip install doccano-transformer
from doccano_transformer.datasets import NERDataset
from doccano_transformer.utils import read_jsonl
dataset = read_jsonl(filepath='file_2.json1', dataset=NERDataset, encoding='utf-8')
dataset.to_spacy(tokenizer=str.split)
TRAIN_DATA = []
for x in dataset:
xx = {}
try:
xx['entities'] = list(x.labels.values())[0]
TRAIN_DATA.append((x.text, xx))
except:
pass
import spacy
from spacy.gold import biluo_tags_from_offsets
nlp = spacy.load('en_core_web_sm')
docs = []
for text, annot in TRAIN_DATA:
doc = nlp(text)
tags = biluo_tags_from_offsets(doc, annot['entities'])
tags = [x.replace('U-','B-') for x in tags]
tags = [x.replace('L-','I-') for x in tags]
tmpdocs = [(x,y) for x,y in zip(text.split(),tags)]
docs.append(tmpdocs)
import random
import pandas as pd
X = pd.DataFrame(docs[0])
X['break'] = 'train'
X = X.append(pd.Series(dtype='str'), ignore_index=True)
for x in docs[1:]:
Xt = pd.DataFrame(x)
Xt = Xt.append(pd.Series(dtype='str'), ignore_index=True)
Xt['break'] = random.choices(['train','val','test'], weights=(90,5,5), k=1)[0]
X = X.append(Xt)
X = X.fillna('')
X = X.reset_index(drop=True)
train = X.loc[X['break']=='train',[0,1]]; train.to_csv('train.txt', sep=' ', header=None, index=False)
val = X.loc[X['break']=='val',[0,1]]; val.to_csv('val.txt', sep=' ', header=None, index=False)
test = X.loc[X['break']=='test',[0,1]]; test.to_csv('test.txt', sep=' ', header=None, index=False)
import re
import numpy as np
!pip install -q flair
from flair.data import Corpus
from flair.datasets import ColumnCorpus
columns = {0:'text', 1:'ner'}
data_folder = '/content'
corpus: Corpus = ColumnCorpus(data_folder, columns,
train_file = 'train.txt',
test_file = 'test.txt',
dev_file = 'val.txt')
tag_type = 'ner'
tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
from flair.embeddings import WordEmbeddings, StackedEmbeddings, FlairEmbeddings, CharacterEmbeddings
embedding_types = [
WordEmbeddings('glove'),
# comment in this line to use character embeddings
# CharacterEmbeddings(),
# comment in these lines to use flair embeddings
# FlairEmbeddings('news-forward'),
# FlairEmbeddings('news-backward'),
]
embeddings : StackedEmbeddings = StackedEmbeddings(embeddings=embedding_types)
from flair.models import SequenceTagger
tagger : SequenceTagger = SequenceTagger(hidden_size=256,
embeddings=embeddings,
tag_dictionary=tag_dictionary,
tag_type=tag_type,
use_crf=True)
from flair.trainers import ModelTrainer
trainer : ModelTrainer = ModelTrainer(tagger, corpus)
trainer.train('/content',
learning_rate=0.5,
mini_batch_size=32,
max_epochs=150,
)
2020-09-11 09:30:11,352 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,354 Model: "SequenceTagger(
(embeddings): StackedEmbeddings(
(list_embedding_0): WordEmbeddings('glove')
)
(word_dropout): WordDropout(p=0.05)
(locked_dropout): LockedDropout(p=0.5)
(embedding2nn): Linear(in_features=100, out_features=100, bias=True)
(rnn): LSTM(100, 256, batch_first=True, bidirectional=True)
(linear): Linear(in_features=512, out_features=24, bias=True)
(beta): 1.0
(weights): None
(weight_tensor) None
)"
2020-09-11 09:30:11,357 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,359 Corpus: "Corpus: 526 train + 36 dev + 31 test sentences"
2020-09-11 09:30:11,360 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,364 Parameters:
2020-09-11 09:30:11,365 - learning_rate: "0.5"
2020-09-11 09:30:11,368 - mini_batch_size: "32"
2020-09-11 09:30:11,373 - patience: "3"
2020-09-11 09:30:11,374 - anneal_factor: "0.5"
2020-09-11 09:30:11,376 - max_epochs: "150"
2020-09-11 09:30:11,377 - shuffle: "True"
2020-09-11 09:30:11,378 - train_with_dev: "False"
2020-09-11 09:30:11,380 - batch_growth_annealing: "False"
2020-09-11 09:30:11,381 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,382 Model training base path: "/content"
2020-09-11 09:30:11,384 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,385 Device: cuda:0
2020-09-11 09:30:11,387 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,388 Embeddings storage mode: cpu
2020-09-11 09:30:11,390 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:11,535 epoch 1 - iter 1/17 - loss 3.43299055 - samples/sec: 224.37 - lr: 0.500000
2020-09-11 09:30:11,674 epoch 1 - iter 2/17 - loss 16.62798858 - samples/sec: 233.49 - lr: 0.500000
2020-09-11 09:30:11,773 epoch 1 - iter 3/17 - loss 12.71970431 - samples/sec: 327.61 - lr: 0.500000
2020-09-11 09:30:11,886 epoch 1 - iter 4/17 - loss 12.07628489 - samples/sec: 288.50 - lr: 0.500000
2020-09-11 09:30:12,011 epoch 1 - iter 5/17 - loss 11.39370499 - samples/sec: 258.89 - lr: 0.500000
2020-09-11 09:30:12,120 epoch 1 - iter 6/17 - loss 10.47585432 - samples/sec: 297.29 - lr: 0.500000
2020-09-11 09:30:12,226 epoch 1 - iter 7/17 - loss 10.23672819 - samples/sec: 315.00 - lr: 0.500000
2020-09-11 09:30:12,327 epoch 1 - iter 8/17 - loss 10.13630027 - samples/sec: 326.75 - lr: 0.500000
2020-09-11 09:30:12,438 epoch 1 - iter 9/17 - loss 9.72696336 - samples/sec: 292.91 - lr: 0.500000
2020-09-11 09:30:12,542 epoch 1 - iter 10/17 - loss 9.41571531 - samples/sec: 317.52 - lr: 0.500000
2020-09-11 09:30:12,654 epoch 1 - iter 11/17 - loss 9.45854959 - samples/sec: 288.96 - lr: 0.500000
2020-09-11 09:30:12,769 epoch 1 - iter 12/17 - loss 9.27839812 - samples/sec: 282.59 - lr: 0.500000
2020-09-11 09:30:12,887 epoch 1 - iter 13/17 - loss 9.43319035 - samples/sec: 280.07 - lr: 0.500000
2020-09-11 09:30:13,009 epoch 1 - iter 14/17 - loss 9.24219717 - samples/sec: 264.80 - lr: 0.500000
2020-09-11 09:30:13,120 epoch 1 - iter 15/17 - loss 9.08057699 - samples/sec: 291.56 - lr: 0.500000
2020-09-11 09:30:13,239 epoch 1 - iter 16/17 - loss 9.04910478 - samples/sec: 272.95 - lr: 0.500000
2020-09-11 09:30:13,321 epoch 1 - iter 17/17 - loss 9.02534998 - samples/sec: 401.20 - lr: 0.500000
2020-09-11 09:30:13,324 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:13,325 EPOCH 1 done: loss 9.0253 - lr 0.5000000
2020-09-11 09:30:13,419 DEV : loss 5.717854976654053 - score 0.8698
2020-09-11 09:30:13,423 BAD EPOCHS (no improvement): 0
saving best model
2020-09-11 09:30:16,358 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:16,472 epoch 2 - iter 1/17 - loss 5.83918762 - samples/sec: 296.08 - lr: 0.500000
2020-09-11 09:30:16,588 epoch 2 - iter 2/17 - loss 6.36326265 - samples/sec: 277.67 - lr: 0.500000
2020-09-11 09:30:16,713 epoch 2 - iter 3/17 - loss 6.86526155 - samples/sec: 258.96 - lr: 0.500000
2020-09-11 09:30:16,825 epoch 2 - iter 4/17 - loss 6.52817869 - samples/sec: 295.13 - lr: 0.500000
2020-09-11 09:30:16,952 epoch 2 - iter 5/17 - loss 6.21855764 - samples/sec: 253.81 - lr: 0.500000
2020-09-11 09:30:17,070 epoch 2 - iter 6/17 - loss 6.00696262 - samples/sec: 275.53 - lr: 0.500000
2020-09-11 09:30:17,186 epoch 2 - iter 7/17 - loss 5.75874615 - samples/sec: 277.33 - lr: 0.500000
2020-09-11 09:30:17,302 epoch 2 - iter 8/17 - loss 6.11816967 - samples/sec: 278.78 - lr: 0.500000
2020-09-11 09:30:17,415 epoch 2 - iter 9/17 - loss 6.44984976 - samples/sec: 288.02 - lr: 0.500000
2020-09-11 09:30:17,530 epoch 2 - iter 10/17 - loss 6.60953989 - samples/sec: 279.91 - lr: 0.500000
2020-09-11 09:30:17,642 epoch 2 - iter 11/17 - loss 6.44363026 - samples/sec: 291.03 - lr: 0.500000
2020-09-11 09:30:17,764 epoch 2 - iter 12/17 - loss 6.50487467 - samples/sec: 263.27 - lr: 0.500000
2020-09-11 09:30:17,875 epoch 2 - iter 13/17 - loss 6.54225309 - samples/sec: 291.79 - lr: 0.500000
2020-09-11 09:30:18,001 epoch 2 - iter 14/17 - loss 6.54856760 - samples/sec: 257.35 - lr: 0.500000
2020-09-11 09:30:18,115 epoch 2 - iter 15/17 - loss 6.56940902 - samples/sec: 282.18 - lr: 0.500000
2020-09-11 09:30:18,222 epoch 2 - iter 16/17 - loss 6.58133185 - samples/sec: 303.36 - lr: 0.500000
2020-09-11 09:30:18,317 epoch 2 - iter 17/17 - loss 6.60969802 - samples/sec: 342.00 - lr: 0.500000
2020-09-11 09:30:18,319 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:18,325 EPOCH 2 done: loss 6.6097 - lr 0.5000000
2020-09-11 09:30:18,418 DEV : loss 7.014471054077148 - score 0.7978
2020-09-11 09:30:18,421 BAD EPOCHS (no improvement): 1
2020-09-11 09:30:18,427 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:18,541 epoch 3 - iter 1/17 - loss 4.26388168 - samples/sec: 286.30 - lr: 0.500000
2020-09-11 09:30:18,669 epoch 3 - iter 2/17 - loss 4.22117805 - samples/sec: 251.47 - lr: 0.500000
2020-09-11 09:30:18,789 epoch 3 - iter 3/17 - loss 4.93522231 - samples/sec: 269.09 - lr: 0.500000
2020-09-11 09:30:18,908 epoch 3 - iter 4/17 - loss 5.27011728 - samples/sec: 272.53 - lr: 0.500000
2020-09-11 09:30:19,023 epoch 3 - iter 5/17 - loss 5.39386015 - samples/sec: 282.34 - lr: 0.500000
2020-09-11 09:30:19,147 epoch 3 - iter 6/17 - loss 5.57506188 - samples/sec: 260.17 - lr: 0.500000
2020-09-11 09:30:19,263 epoch 3 - iter 7/17 - loss 5.80165243 - samples/sec: 280.20 - lr: 0.500000
2020-09-11 09:30:19,377 epoch 3 - iter 8/17 - loss 5.74144500 - samples/sec: 284.49 - lr: 0.500000
2020-09-11 09:30:19,489 epoch 3 - iter 9/17 - loss 5.83528980 - samples/sec: 289.27 - lr: 0.500000
2020-09-11 09:30:19,602 epoch 3 - iter 10/17 - loss 5.94376082 - samples/sec: 285.24 - lr: 0.500000
2020-09-11 09:30:19,719 epoch 3 - iter 11/17 - loss 6.10046205 - samples/sec: 276.22 - lr: 0.500000
2020-09-11 09:30:19,834 epoch 3 - iter 12/17 - loss 6.31112369 - samples/sec: 281.37 - lr: 0.500000
2020-09-11 09:30:19,943 epoch 3 - iter 13/17 - loss 6.59776673 - samples/sec: 296.36 - lr: 0.500000
2020-09-11 09:30:20,049 epoch 3 - iter 14/17 - loss 6.57175589 - samples/sec: 306.47 - lr: 0.500000
2020-09-11 09:30:20,175 epoch 3 - iter 15/17 - loss 6.53515657 - samples/sec: 256.03 - lr: 0.500000
2020-09-11 09:30:20,280 epoch 3 - iter 16/17 - loss 6.45572126 - samples/sec: 308.22 - lr: 0.500000
2020-09-11 09:30:20,363 epoch 3 - iter 17/17 - loss 6.33213214 - samples/sec: 394.86 - lr: 0.500000
2020-09-11 09:30:20,364 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:20,369 EPOCH 3 done: loss 6.3321 - lr 0.5000000
2020-09-11 09:30:20,447 DEV : loss 6.267121315002441 - score 0.7705
2020-09-11 09:30:20,451 BAD EPOCHS (no improvement): 2
2020-09-11 09:30:20,453 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:20,567 epoch 4 - iter 1/17 - loss 5.04142189 - samples/sec: 288.98 - lr: 0.500000
2020-09-11 09:30:20,708 epoch 4 - iter 2/17 - loss 6.17017388 - samples/sec: 229.78 - lr: 0.500000
2020-09-11 09:30:20,842 epoch 4 - iter 3/17 - loss 5.64043872 - samples/sec: 241.44 - lr: 0.500000
2020-09-11 09:30:20,949 epoch 4 - iter 4/17 - loss 5.29508078 - samples/sec: 311.53 - lr: 0.500000
2020-09-11 09:30:21,053 epoch 4 - iter 5/17 - loss 4.82136688 - samples/sec: 311.77 - lr: 0.500000
2020-09-11 09:30:21,174 epoch 4 - iter 6/17 - loss 4.83669527 - samples/sec: 269.39 - lr: 0.500000
2020-09-11 09:30:21,286 epoch 4 - iter 7/17 - loss 4.87344299 - samples/sec: 293.77 - lr: 0.500000
2020-09-11 09:30:21,396 epoch 4 - iter 8/17 - loss 5.19339198 - samples/sec: 292.26 - lr: 0.500000
2020-09-11 09:30:21,504 epoch 4 - iter 9/17 - loss 5.36148289 - samples/sec: 303.14 - lr: 0.500000
2020-09-11 09:30:21,612 epoch 4 - iter 10/17 - loss 5.43170042 - samples/sec: 298.35 - lr: 0.500000
2020-09-11 09:30:21,724 epoch 4 - iter 11/17 - loss 5.40744721 - samples/sec: 290.04 - lr: 0.500000
2020-09-11 09:30:21,838 epoch 4 - iter 12/17 - loss 5.46409893 - samples/sec: 284.06 - lr: 0.500000
2020-09-11 09:30:21,947 epoch 4 - iter 13/17 - loss 5.48474825 - samples/sec: 295.24 - lr: 0.500000
2020-09-11 09:30:22,054 epoch 4 - iter 14/17 - loss 5.34932395 - samples/sec: 303.95 - lr: 0.500000
2020-09-11 09:30:22,167 epoch 4 - iter 15/17 - loss 5.32448457 - samples/sec: 284.46 - lr: 0.500000
2020-09-11 09:30:22,292 epoch 4 - iter 16/17 - loss 5.21404728 - samples/sec: 260.52 - lr: 0.500000
2020-09-11 09:30:22,374 epoch 4 - iter 17/17 - loss 5.10825051 - samples/sec: 394.18 - lr: 0.500000
2020-09-11 09:30:22,375 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:22,380 EPOCH 4 done: loss 5.1083 - lr 0.5000000
2020-09-11 09:30:22,456 DEV : loss 4.367828369140625 - score 0.8491
2020-09-11 09:30:22,459 BAD EPOCHS (no improvement): 3
2020-09-11 09:30:22,461 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:22,570 epoch 5 - iter 1/17 - loss 6.80663157 - samples/sec: 303.24 - lr: 0.500000
2020-09-11 09:30:22,666 epoch 5 - iter 2/17 - loss 5.35868144 - samples/sec: 341.59 - lr: 0.500000
2020-09-11 09:30:22,779 epoch 5 - iter 3/17 - loss 5.17130645 - samples/sec: 288.28 - lr: 0.500000
2020-09-11 09:30:22,892 epoch 5 - iter 4/17 - loss 4.78146195 - samples/sec: 284.10 - lr: 0.500000
2020-09-11 09:30:23,007 epoch 5 - iter 5/17 - loss 4.49372272 - samples/sec: 283.10 - lr: 0.500000
2020-09-11 09:30:23,116 epoch 5 - iter 6/17 - loss 4.40794814 - samples/sec: 296.27 - lr: 0.500000
2020-09-11 09:30:23,233 epoch 5 - iter 7/17 - loss 5.19045050 - samples/sec: 277.11 - lr: 0.500000
2020-09-11 09:30:23,341 epoch 5 - iter 8/17 - loss 5.04928717 - samples/sec: 300.85 - lr: 0.500000
2020-09-11 09:30:23,461 epoch 5 - iter 9/17 - loss 4.83003471 - samples/sec: 268.51 - lr: 0.500000
2020-09-11 09:30:23,569 epoch 5 - iter 10/17 - loss 5.06927426 - samples/sec: 301.46 - lr: 0.500000
2020-09-11 09:30:23,696 epoch 5 - iter 11/17 - loss 4.97240255 - samples/sec: 255.22 - lr: 0.500000
2020-09-11 09:30:23,806 epoch 5 - iter 12/17 - loss 5.07106882 - samples/sec: 294.08 - lr: 0.500000
2020-09-11 09:30:23,929 epoch 5 - iter 13/17 - loss 5.17605252 - samples/sec: 262.74 - lr: 0.500000
2020-09-11 09:30:24,038 epoch 5 - iter 14/17 - loss 5.14826872 - samples/sec: 295.90 - lr: 0.500000
2020-09-11 09:30:24,144 epoch 5 - iter 15/17 - loss 5.33293700 - samples/sec: 305.56 - lr: 0.500000
2020-09-11 09:30:24,257 epoch 5 - iter 16/17 - loss 5.31216620 - samples/sec: 286.96 - lr: 0.500000
2020-09-11 09:30:24,342 epoch 5 - iter 17/17 - loss 5.50642596 - samples/sec: 380.72 - lr: 0.500000
2020-09-11 09:30:24,347 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:24,348 EPOCH 5 done: loss 5.5064 - lr 0.5000000
2020-09-11 09:30:24,428 DEV : loss 7.306588172912598 - score 0.8609
Epoch 5: reducing learning rate of group 0 to 2.5000e-01.
2020-09-11 09:30:24,430 BAD EPOCHS (no improvement): 4
2020-09-11 09:30:24,434 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:24,544 epoch 6 - iter 1/17 - loss 6.96324205 - samples/sec: 296.94 - lr: 0.250000
2020-09-11 09:30:24,658 epoch 6 - iter 2/17 - loss 5.62635350 - samples/sec: 284.77 - lr: 0.250000
2020-09-11 09:30:24,779 epoch 6 - iter 3/17 - loss 5.42040205 - samples/sec: 266.03 - lr: 0.250000
2020-09-11 09:30:24,922 epoch 6 - iter 4/17 - loss 5.56092465 - samples/sec: 226.18 - lr: 0.250000
2020-09-11 09:30:25,033 epoch 6 - iter 5/17 - loss 5.47617340 - samples/sec: 291.45 - lr: 0.250000
2020-09-11 09:30:25,141 epoch 6 - iter 6/17 - loss 5.06147250 - samples/sec: 300.30 - lr: 0.250000
2020-09-11 09:30:25,253 epoch 6 - iter 7/17 - loss 4.85828168 - samples/sec: 288.46 - lr: 0.250000
2020-09-11 09:30:25,371 epoch 6 - iter 8/17 - loss 4.62894660 - samples/sec: 273.66 - lr: 0.250000
2020-09-11 09:30:25,485 epoch 6 - iter 9/17 - loss 4.33706284 - samples/sec: 283.62 - lr: 0.250000
2020-09-11 09:30:25,613 epoch 6 - iter 10/17 - loss 4.21852846 - samples/sec: 253.76 - lr: 0.250000
2020-09-11 09:30:25,719 epoch 6 - iter 11/17 - loss 4.07177886 - samples/sec: 305.92 - lr: 0.250000
2020-09-11 09:30:25,833 epoch 6 - iter 12/17 - loss 3.97449460 - samples/sec: 283.66 - lr: 0.250000
2020-09-11 09:30:25,952 epoch 6 - iter 13/17 - loss 3.94040249 - samples/sec: 272.32 - lr: 0.250000
2020-09-11 09:30:26,055 epoch 6 - iter 14/17 - loss 3.86654231 - samples/sec: 315.98 - lr: 0.250000
2020-09-11 09:30:26,171 epoch 6 - iter 15/17 - loss 3.94532137 - samples/sec: 277.03 - lr: 0.250000
2020-09-11 09:30:26,283 epoch 6 - iter 16/17 - loss 3.87008142 - samples/sec: 290.29 - lr: 0.250000
2020-09-11 09:30:26,349 epoch 6 - iter 17/17 - loss 3.82113767 - samples/sec: 506.56 - lr: 0.250000
2020-09-11 09:30:26,350 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:26,355 EPOCH 6 done: loss 3.8211 - lr 0.2500000
2020-09-11 09:30:26,436 DEV : loss 3.1234371662139893 - score 0.8808
2020-09-11 09:30:26,438 BAD EPOCHS (no improvement): 0
saving best model
2020-09-11 09:30:29,221 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:29,343 epoch 7 - iter 1/17 - loss 3.70950317 - samples/sec: 268.26 - lr: 0.250000
2020-09-11 09:30:29,446 epoch 7 - iter 2/17 - loss 3.15139985 - samples/sec: 313.16 - lr: 0.250000
2020-09-11 09:30:29,557 epoch 7 - iter 3/17 - loss 2.90648588 - samples/sec: 292.79 - lr: 0.250000
2020-09-11 09:30:29,691 epoch 7 - iter 4/17 - loss 2.84409702 - samples/sec: 240.22 - lr: 0.250000
2020-09-11 09:30:29,792 epoch 7 - iter 5/17 - loss 2.73230619 - samples/sec: 322.98 - lr: 0.250000
2020-09-11 09:30:29,917 epoch 7 - iter 6/17 - loss 2.97766495 - samples/sec: 258.82 - lr: 0.250000
2020-09-11 09:30:30,024 epoch 7 - iter 7/17 - loss 3.07457617 - samples/sec: 301.19 - lr: 0.250000
2020-09-11 09:30:30,129 epoch 7 - iter 8/17 - loss 3.16595078 - samples/sec: 311.51 - lr: 0.250000
2020-09-11 09:30:30,226 epoch 7 - iter 9/17 - loss 3.29891888 - samples/sec: 332.91 - lr: 0.250000
2020-09-11 09:30:30,337 epoch 7 - iter 10/17 - loss 3.39236808 - samples/sec: 291.15 - lr: 0.250000
2020-09-11 09:30:30,441 epoch 7 - iter 11/17 - loss 3.41726154 - samples/sec: 312.11 - lr: 0.250000
2020-09-11 09:30:30,549 epoch 7 - iter 12/17 - loss 3.41482002 - samples/sec: 299.15 - lr: 0.250000
2020-09-11 09:30:30,648 epoch 7 - iter 13/17 - loss 3.32942341 - samples/sec: 326.18 - lr: 0.250000
2020-09-11 09:30:30,793 epoch 7 - iter 14/17 - loss 3.42877962 - samples/sec: 223.33 - lr: 0.250000
2020-09-11 09:30:30,905 epoch 7 - iter 15/17 - loss 3.37255146 - samples/sec: 287.33 - lr: 0.250000
2020-09-11 09:30:31,012 epoch 7 - iter 16/17 - loss 3.29338527 - samples/sec: 302.59 - lr: 0.250000
2020-09-11 09:30:31,097 epoch 7 - iter 17/17 - loss 3.36279516 - samples/sec: 382.70 - lr: 0.250000
2020-09-11 09:30:31,098 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:31,104 EPOCH 7 done: loss 3.3628 - lr 0.2500000
2020-09-11 09:30:31,181 DEV : loss 2.352734327316284 - score 0.8164
2020-09-11 09:30:31,183 BAD EPOCHS (no improvement): 1
2020-09-11 09:30:31,187 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:31,298 epoch 8 - iter 1/17 - loss 3.31515622 - samples/sec: 294.10 - lr: 0.250000
2020-09-11 09:30:31,430 epoch 8 - iter 2/17 - loss 3.51384747 - samples/sec: 245.22 - lr: 0.250000
2020-09-11 09:30:31,541 epoch 8 - iter 3/17 - loss 3.30893199 - samples/sec: 290.08 - lr: 0.250000
2020-09-11 09:30:31,656 epoch 8 - iter 4/17 - loss 3.24576128 - samples/sec: 283.74 - lr: 0.250000
2020-09-11 09:30:31,758 epoch 8 - iter 5/17 - loss 3.12689247 - samples/sec: 315.29 - lr: 0.250000
2020-09-11 09:30:31,862 epoch 8 - iter 6/17 - loss 2.94166811 - samples/sec: 313.46 - lr: 0.250000
2020-09-11 09:30:31,992 epoch 8 - iter 7/17 - loss 3.11964280 - samples/sec: 251.93 - lr: 0.250000
2020-09-11 09:30:32,098 epoch 8 - iter 8/17 - loss 3.04969150 - samples/sec: 304.60 - lr: 0.250000
2020-09-11 09:30:32,205 epoch 8 - iter 9/17 - loss 3.06001208 - samples/sec: 302.37 - lr: 0.250000
2020-09-11 09:30:32,313 epoch 8 - iter 10/17 - loss 3.13563135 - samples/sec: 298.31 - lr: 0.250000
2020-09-11 09:30:32,428 epoch 8 - iter 11/17 - loss 3.17867511 - samples/sec: 281.42 - lr: 0.250000
2020-09-11 09:30:32,542 epoch 8 - iter 12/17 - loss 3.17592837 - samples/sec: 287.75 - lr: 0.250000
2020-09-11 09:30:32,643 epoch 8 - iter 13/17 - loss 3.17387684 - samples/sec: 321.38 - lr: 0.250000
2020-09-11 09:30:32,752 epoch 8 - iter 14/17 - loss 3.30220359 - samples/sec: 296.03 - lr: 0.250000
2020-09-11 09:30:32,864 epoch 8 - iter 15/17 - loss 3.28310798 - samples/sec: 288.32 - lr: 0.250000
2020-09-11 09:30:32,984 epoch 8 - iter 16/17 - loss 3.27035704 - samples/sec: 270.40 - lr: 0.250000
2020-09-11 09:30:33,059 epoch 8 - iter 17/17 - loss 3.25974651 - samples/sec: 429.82 - lr: 0.250000
2020-09-11 09:30:33,060 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:33,065 EPOCH 8 done: loss 3.2597 - lr 0.2500000
2020-09-11 09:30:33,142 DEV : loss 2.3950960636138916 - score 0.8911
2020-09-11 09:30:33,145 BAD EPOCHS (no improvement): 0
saving best model
2020-09-11 09:30:35,928 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:36,063 epoch 9 - iter 1/17 - loss 4.13516760 - samples/sec: 241.45 - lr: 0.250000
2020-09-11 09:30:36,177 epoch 9 - iter 2/17 - loss 4.22095203 - samples/sec: 283.96 - lr: 0.250000
2020-09-11 09:30:36,282 epoch 9 - iter 3/17 - loss 3.74937018 - samples/sec: 307.81 - lr: 0.250000
2020-09-11 09:30:36,384 epoch 9 - iter 4/17 - loss 3.56653398 - samples/sec: 316.58 - lr: 0.250000
2020-09-11 09:30:36,505 epoch 9 - iter 5/17 - loss 3.27518516 - samples/sec: 267.32 - lr: 0.250000
2020-09-11 09:30:36,619 epoch 9 - iter 6/17 - loss 3.02496213 - samples/sec: 285.21 - lr: 0.250000
2020-09-11 09:30:36,746 epoch 9 - iter 7/17 - loss 3.03646287 - samples/sec: 253.72 - lr: 0.250000
2020-09-11 09:30:36,862 epoch 9 - iter 8/17 - loss 3.04716738 - samples/sec: 281.19 - lr: 0.250000
2020-09-11 09:30:36,966 epoch 9 - iter 9/17 - loss 3.13016204 - samples/sec: 310.68 - lr: 0.250000
2020-09-11 09:30:37,094 epoch 9 - iter 10/17 - loss 3.07540933 - samples/sec: 250.78 - lr: 0.250000
2020-09-11 09:30:37,203 epoch 9 - iter 11/17 - loss 2.99208668 - samples/sec: 298.74 - lr: 0.250000
2020-09-11 09:30:37,326 epoch 9 - iter 12/17 - loss 3.08011819 - samples/sec: 262.50 - lr: 0.250000
2020-09-11 09:30:37,437 epoch 9 - iter 13/17 - loss 3.15795074 - samples/sec: 290.04 - lr: 0.250000
2020-09-11 09:30:37,554 epoch 9 - iter 14/17 - loss 3.13511590 - samples/sec: 276.32 - lr: 0.250000
2020-09-11 09:30:37,658 epoch 9 - iter 15/17 - loss 3.09624709 - samples/sec: 312.35 - lr: 0.250000
2020-09-11 09:30:37,770 epoch 9 - iter 16/17 - loss 3.20836692 - samples/sec: 288.37 - lr: 0.250000
2020-09-11 09:30:37,857 epoch 9 - iter 17/17 - loss 3.14222338 - samples/sec: 372.18 - lr: 0.250000
2020-09-11 09:30:37,859 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:37,865 EPOCH 9 done: loss 3.1422 - lr 0.2500000
2020-09-11 09:30:37,942 DEV : loss 2.601285934448242 - score 0.8783
2020-09-11 09:30:37,944 BAD EPOCHS (no improvement): 1
2020-09-11 09:30:37,949 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:38,064 epoch 10 - iter 1/17 - loss 2.18587756 - samples/sec: 286.91 - lr: 0.250000
2020-09-11 09:30:38,173 epoch 10 - iter 2/17 - loss 2.85696280 - samples/sec: 297.21 - lr: 0.250000
2020-09-11 09:30:38,279 epoch 10 - iter 3/17 - loss 3.14432240 - samples/sec: 308.52 - lr: 0.250000
2020-09-11 09:30:38,385 epoch 10 - iter 4/17 - loss 3.15336317 - samples/sec: 303.25 - lr: 0.250000
2020-09-11 09:30:38,514 epoch 10 - iter 5/17 - loss 3.14261246 - samples/sec: 250.19 - lr: 0.250000
2020-09-11 09:30:38,621 epoch 10 - iter 6/17 - loss 3.17861811 - samples/sec: 303.24 - lr: 0.250000
2020-09-11 09:30:38,734 epoch 10 - iter 7/17 - loss 3.23335017 - samples/sec: 286.76 - lr: 0.250000
2020-09-11 09:30:38,844 epoch 10 - iter 8/17 - loss 3.17415664 - samples/sec: 293.62 - lr: 0.250000
2020-09-11 09:30:38,955 epoch 10 - iter 9/17 - loss 3.23231522 - samples/sec: 291.92 - lr: 0.250000
2020-09-11 09:30:39,075 epoch 10 - iter 10/17 - loss 3.25380137 - samples/sec: 269.14 - lr: 0.250000
2020-09-11 09:30:39,190 epoch 10 - iter 11/17 - loss 3.16060584 - samples/sec: 280.60 - lr: 0.250000
2020-09-11 09:30:39,294 epoch 10 - iter 12/17 - loss 3.02024135 - samples/sec: 312.62 - lr: 0.250000
2020-09-11 09:30:39,401 epoch 10 - iter 13/17 - loss 3.02866225 - samples/sec: 301.38 - lr: 0.250000
2020-09-11 09:30:39,530 epoch 10 - iter 14/17 - loss 3.05385321 - samples/sec: 250.16 - lr: 0.250000
2020-09-11 09:30:39,644 epoch 10 - iter 15/17 - loss 3.06730847 - samples/sec: 283.84 - lr: 0.250000
2020-09-11 09:30:39,757 epoch 10 - iter 16/17 - loss 3.21189360 - samples/sec: 286.73 - lr: 0.250000
2020-09-11 09:30:39,832 epoch 10 - iter 17/17 - loss 3.28693191 - samples/sec: 432.01 - lr: 0.250000
2020-09-11 09:30:39,833 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:39,838 EPOCH 10 done: loss 3.2869 - lr 0.2500000
2020-09-11 09:30:39,919 DEV : loss 2.621568202972412 - score 0.8878
2020-09-11 09:30:39,921 BAD EPOCHS (no improvement): 2
2020-09-11 09:30:39,923 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:40,028 epoch 11 - iter 1/17 - loss 1.97674811 - samples/sec: 312.16 - lr: 0.250000
2020-09-11 09:30:40,146 epoch 11 - iter 2/17 - loss 2.63742667 - samples/sec: 273.63 - lr: 0.250000
2020-09-11 09:30:40,247 epoch 11 - iter 3/17 - loss 3.29801786 - samples/sec: 321.25 - lr: 0.250000
2020-09-11 09:30:40,365 epoch 11 - iter 4/17 - loss 3.60576668 - samples/sec: 272.73 - lr: 0.250000
2020-09-11 09:30:40,474 epoch 11 - iter 5/17 - loss 3.71707928 - samples/sec: 303.02 - lr: 0.250000
2020-09-11 09:30:40,589 epoch 11 - iter 6/17 - loss 3.40174667 - samples/sec: 281.19 - lr: 0.250000
2020-09-11 09:30:40,712 epoch 11 - iter 7/17 - loss 3.18708919 - samples/sec: 261.80 - lr: 0.250000
2020-09-11 09:30:40,826 epoch 11 - iter 8/17 - loss 3.04836126 - samples/sec: 287.11 - lr: 0.250000
2020-09-11 09:30:40,943 epoch 11 - iter 9/17 - loss 2.98250650 - samples/sec: 276.91 - lr: 0.250000
2020-09-11 09:30:41,058 epoch 11 - iter 10/17 - loss 2.92936121 - samples/sec: 282.13 - lr: 0.250000
2020-09-11 09:30:41,167 epoch 11 - iter 11/17 - loss 2.92226912 - samples/sec: 294.92 - lr: 0.250000
2020-09-11 09:30:41,281 epoch 11 - iter 12/17 - loss 2.89158462 - samples/sec: 285.59 - lr: 0.250000
2020-09-11 09:30:41,389 epoch 11 - iter 13/17 - loss 3.05706126 - samples/sec: 300.46 - lr: 0.250000
2020-09-11 09:30:41,497 epoch 11 - iter 14/17 - loss 3.01038414 - samples/sec: 304.51 - lr: 0.250000
2020-09-11 09:30:41,619 epoch 11 - iter 15/17 - loss 3.13056888 - samples/sec: 266.18 - lr: 0.250000
2020-09-11 09:30:41,749 epoch 11 - iter 16/17 - loss 3.20361467 - samples/sec: 248.63 - lr: 0.250000
2020-09-11 09:30:41,824 epoch 11 - iter 17/17 - loss 3.17225321 - samples/sec: 450.57 - lr: 0.250000
2020-09-11 09:30:41,825 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:41,831 EPOCH 11 done: loss 3.1723 - lr 0.2500000
2020-09-11 09:30:41,909 DEV : loss 2.6496293544769287 - score 0.9175
2020-09-11 09:30:41,911 BAD EPOCHS (no improvement): 0
saving best model
2020-09-11 09:30:44,727 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:44,843 epoch 12 - iter 1/17 - loss 3.06218433 - samples/sec: 278.55 - lr: 0.250000
2020-09-11 09:30:44,962 epoch 12 - iter 2/17 - loss 2.53818643 - samples/sec: 273.01 - lr: 0.250000
2020-09-11 09:30:45,080 epoch 12 - iter 3/17 - loss 2.66141121 - samples/sec: 276.04 - lr: 0.250000
2020-09-11 09:30:45,196 epoch 12 - iter 4/17 - loss 2.73923683 - samples/sec: 278.75 - lr: 0.250000
2020-09-11 09:30:45,304 epoch 12 - iter 5/17 - loss 2.70134916 - samples/sec: 302.96 - lr: 0.250000
2020-09-11 09:30:45,411 epoch 12 - iter 6/17 - loss 2.47173454 - samples/sec: 301.51 - lr: 0.250000
2020-09-11 09:30:45,524 epoch 12 - iter 7/17 - loss 2.56464466 - samples/sec: 288.11 - lr: 0.250000
2020-09-11 09:30:45,638 epoch 12 - iter 8/17 - loss 2.83810382 - samples/sec: 284.02 - lr: 0.250000
2020-09-11 09:30:45,753 epoch 12 - iter 9/17 - loss 2.96720710 - samples/sec: 283.20 - lr: 0.250000
2020-09-11 09:30:45,863 epoch 12 - iter 10/17 - loss 3.00092806 - samples/sec: 293.61 - lr: 0.250000
2020-09-11 09:30:45,980 epoch 12 - iter 11/17 - loss 3.04209977 - samples/sec: 282.71 - lr: 0.250000
2020-09-11 09:30:46,103 epoch 12 - iter 12/17 - loss 3.06752718 - samples/sec: 261.65 - lr: 0.250000
2020-09-11 09:30:46,219 epoch 12 - iter 13/17 - loss 3.10638168 - samples/sec: 278.15 - lr: 0.250000
2020-09-11 09:30:46,325 epoch 12 - iter 14/17 - loss 3.11180851 - samples/sec: 307.11 - lr: 0.250000
2020-09-11 09:30:46,435 epoch 12 - iter 15/17 - loss 3.11393329 - samples/sec: 295.05 - lr: 0.250000
2020-09-11 09:30:46,563 epoch 12 - iter 16/17 - loss 3.17738984 - samples/sec: 251.88 - lr: 0.250000
2020-09-11 09:30:46,640 epoch 12 - iter 17/17 - loss 3.13963230 - samples/sec: 419.02 - lr: 0.250000
2020-09-11 09:30:46,642 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:46,647 EPOCH 12 done: loss 3.1396 - lr 0.2500000
2020-09-11 09:30:46,753 DEV : loss 2.456434726715088 - score 0.8571
2020-09-11 09:30:46,757 BAD EPOCHS (no improvement): 1
2020-09-11 09:30:46,759 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:46,872 epoch 13 - iter 1/17 - loss 3.74742985 - samples/sec: 288.46 - lr: 0.250000
2020-09-11 09:30:46,978 epoch 13 - iter 2/17 - loss 3.21253848 - samples/sec: 306.03 - lr: 0.250000
2020-09-11 09:30:47,084 epoch 13 - iter 3/17 - loss 3.31751029 - samples/sec: 306.26 - lr: 0.250000
2020-09-11 09:30:47,192 epoch 13 - iter 4/17 - loss 2.90003487 - samples/sec: 298.68 - lr: 0.250000
2020-09-11 09:30:47,310 epoch 13 - iter 5/17 - loss 2.77682316 - samples/sec: 274.49 - lr: 0.250000
2020-09-11 09:30:47,423 epoch 13 - iter 6/17 - loss 2.92825923 - samples/sec: 284.24 - lr: 0.250000
2020-09-11 09:30:47,535 epoch 13 - iter 7/17 - loss 2.93475987 - samples/sec: 290.65 - lr: 0.250000
2020-09-11 09:30:47,660 epoch 13 - iter 8/17 - loss 3.03747167 - samples/sec: 258.99 - lr: 0.250000
2020-09-11 09:30:47,782 epoch 13 - iter 9/17 - loss 3.01528635 - samples/sec: 265.17 - lr: 0.250000
2020-09-11 09:30:47,887 epoch 13 - iter 10/17 - loss 3.02738467 - samples/sec: 308.06 - lr: 0.250000
2020-09-11 09:30:47,995 epoch 13 - iter 11/17 - loss 3.00082893 - samples/sec: 297.64 - lr: 0.250000
2020-09-11 09:30:48,106 epoch 13 - iter 12/17 - loss 3.03533484 - samples/sec: 298.56 - lr: 0.250000
2020-09-11 09:30:48,214 epoch 13 - iter 13/17 - loss 2.99576386 - samples/sec: 299.72 - lr: 0.250000
2020-09-11 09:30:48,320 epoch 13 - iter 14/17 - loss 2.99190849 - samples/sec: 304.07 - lr: 0.250000
2020-09-11 09:30:48,430 epoch 13 - iter 15/17 - loss 3.02248323 - samples/sec: 293.57 - lr: 0.250000
2020-09-11 09:30:48,539 epoch 13 - iter 16/17 - loss 3.00460766 - samples/sec: 297.44 - lr: 0.250000
2020-09-11 09:30:48,619 epoch 13 - iter 17/17 - loss 2.95338992 - samples/sec: 409.80 - lr: 0.250000
2020-09-11 09:30:48,620 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:48,626 EPOCH 13 done: loss 2.9534 - lr 0.2500000
2020-09-11 09:30:48,702 DEV : loss 2.8066673278808594 - score 0.8548
2020-09-11 09:30:48,704 BAD EPOCHS (no improvement): 2
2020-09-11 09:30:48,705 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:48,833 epoch 14 - iter 1/17 - loss 2.78732395 - samples/sec: 257.25 - lr: 0.250000
2020-09-11 09:30:48,945 epoch 14 - iter 2/17 - loss 2.74681008 - samples/sec: 288.53 - lr: 0.250000
2020-09-11 09:30:49,061 epoch 14 - iter 3/17 - loss 2.75188263 - samples/sec: 278.90 - lr: 0.250000
2020-09-11 09:30:49,182 epoch 14 - iter 4/17 - loss 2.79956216 - samples/sec: 266.16 - lr: 0.250000
2020-09-11 09:30:49,289 epoch 14 - iter 5/17 - loss 2.81255398 - samples/sec: 303.61 - lr: 0.250000
2020-09-11 09:30:49,399 epoch 14 - iter 6/17 - loss 2.67327150 - samples/sec: 293.32 - lr: 0.250000
2020-09-11 09:30:49,509 epoch 14 - iter 7/17 - loss 2.64780256 - samples/sec: 296.43 - lr: 0.250000
2020-09-11 09:30:49,626 epoch 14 - iter 8/17 - loss 2.89155030 - samples/sec: 274.99 - lr: 0.250000
2020-09-11 09:30:49,732 epoch 14 - iter 9/17 - loss 2.80368842 - samples/sec: 304.67 - lr: 0.250000
2020-09-11 09:30:49,848 epoch 14 - iter 10/17 - loss 3.01142316 - samples/sec: 280.74 - lr: 0.250000
2020-09-11 09:30:49,969 epoch 14 - iter 11/17 - loss 3.08519667 - samples/sec: 266.39 - lr: 0.250000
2020-09-11 09:30:50,095 epoch 14 - iter 12/17 - loss 3.10426335 - samples/sec: 257.15 - lr: 0.250000
2020-09-11 09:30:50,209 epoch 14 - iter 13/17 - loss 3.06903934 - samples/sec: 282.02 - lr: 0.250000
2020-09-11 09:30:50,324 epoch 14 - iter 14/17 - loss 3.03085591 - samples/sec: 295.41 - lr: 0.250000
2020-09-11 09:30:50,424 epoch 14 - iter 15/17 - loss 3.02578127 - samples/sec: 321.50 - lr: 0.250000
2020-09-11 09:30:50,545 epoch 14 - iter 16/17 - loss 2.98861359 - samples/sec: 267.74 - lr: 0.250000
2020-09-11 09:30:50,625 epoch 14 - iter 17/17 - loss 3.04630093 - samples/sec: 407.75 - lr: 0.250000
2020-09-11 09:30:50,626 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:50,630 EPOCH 14 done: loss 3.0463 - lr 0.2500000
2020-09-11 09:30:50,726 DEV : loss 2.356717109680176 - score 0.9039
2020-09-11 09:30:50,730 BAD EPOCHS (no improvement): 3
2020-09-11 09:30:50,732 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:50,868 epoch 15 - iter 1/17 - loss 4.18320131 - samples/sec: 237.43 - lr: 0.250000
2020-09-11 09:30:50,981 epoch 15 - iter 2/17 - loss 3.43749928 - samples/sec: 287.59 - lr: 0.250000
2020-09-11 09:30:51,089 epoch 15 - iter 3/17 - loss 3.32088296 - samples/sec: 302.31 - lr: 0.250000
2020-09-11 09:30:51,195 epoch 15 - iter 4/17 - loss 3.32438457 - samples/sec: 303.65 - lr: 0.250000
2020-09-11 09:30:51,317 epoch 15 - iter 5/17 - loss 3.04261053 - samples/sec: 265.50 - lr: 0.250000
2020-09-11 09:30:51,431 epoch 15 - iter 6/17 - loss 3.01294063 - samples/sec: 284.89 - lr: 0.250000
2020-09-11 09:30:51,547 epoch 15 - iter 7/17 - loss 3.07561300 - samples/sec: 279.72 - lr: 0.250000
2020-09-11 09:30:51,650 epoch 15 - iter 8/17 - loss 3.14126436 - samples/sec: 312.54 - lr: 0.250000
2020-09-11 09:30:51,755 epoch 15 - iter 9/17 - loss 3.12974224 - samples/sec: 311.04 - lr: 0.250000
2020-09-11 09:30:51,872 epoch 15 - iter 10/17 - loss 3.00686265 - samples/sec: 280.14 - lr: 0.250000
2020-09-11 09:30:51,984 epoch 15 - iter 11/17 - loss 3.00323490 - samples/sec: 288.18 - lr: 0.250000
2020-09-11 09:30:52,091 epoch 15 - iter 12/17 - loss 2.97433997 - samples/sec: 302.88 - lr: 0.250000
2020-09-11 09:30:52,203 epoch 15 - iter 13/17 - loss 2.99207349 - samples/sec: 290.19 - lr: 0.250000
2020-09-11 09:30:52,333 epoch 15 - iter 14/17 - loss 2.94220261 - samples/sec: 249.46 - lr: 0.250000
2020-09-11 09:30:52,447 epoch 15 - iter 15/17 - loss 2.89441790 - samples/sec: 283.54 - lr: 0.250000
2020-09-11 09:30:52,554 epoch 15 - iter 16/17 - loss 2.89061851 - samples/sec: 302.51 - lr: 0.250000
2020-09-11 09:30:52,632 epoch 15 - iter 17/17 - loss 2.86109092 - samples/sec: 415.69 - lr: 0.250000
2020-09-11 09:30:52,634 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:52,637 EPOCH 15 done: loss 2.8611 - lr 0.2500000
2020-09-11 09:30:52,715 DEV : loss 2.2419240474700928 - score 0.8964
Epoch 15: reducing learning rate of group 0 to 1.2500e-01.
2020-09-11 09:30:52,719 BAD EPOCHS (no improvement): 4
2020-09-11 09:30:52,720 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:52,835 epoch 16 - iter 1/17 - loss 3.22716141 - samples/sec: 281.04 - lr: 0.125000
2020-09-11 09:30:52,945 epoch 16 - iter 2/17 - loss 2.45503211 - samples/sec: 296.44 - lr: 0.125000
2020-09-11 09:30:53,049 epoch 16 - iter 3/17 - loss 2.48855003 - samples/sec: 323.45 - lr: 0.125000
2020-09-11 09:30:53,154 epoch 16 - iter 4/17 - loss 2.59882224 - samples/sec: 307.89 - lr: 0.125000
2020-09-11 09:30:53,269 epoch 16 - iter 5/17 - loss 2.60306301 - samples/sec: 281.55 - lr: 0.125000
2020-09-11 09:30:53,382 epoch 16 - iter 6/17 - loss 2.49932460 - samples/sec: 286.07 - lr: 0.125000
2020-09-11 09:30:53,488 epoch 16 - iter 7/17 - loss 2.50845940 - samples/sec: 305.77 - lr: 0.125000
2020-09-11 09:30:53,596 epoch 16 - iter 8/17 - loss 2.45025101 - samples/sec: 299.80 - lr: 0.125000
2020-09-11 09:30:53,706 epoch 16 - iter 9/17 - loss 2.47416223 - samples/sec: 296.94 - lr: 0.125000
2020-09-11 09:30:53,842 epoch 16 - iter 10/17 - loss 2.48455787 - samples/sec: 237.17 - lr: 0.125000
2020-09-11 09:30:53,951 epoch 16 - iter 11/17 - loss 2.49519675 - samples/sec: 297.89 - lr: 0.125000
2020-09-11 09:30:54,074 epoch 16 - iter 12/17 - loss 2.41682629 - samples/sec: 262.08 - lr: 0.125000
2020-09-11 09:30:54,179 epoch 16 - iter 13/17 - loss 2.43769492 - samples/sec: 310.28 - lr: 0.125000
2020-09-11 09:30:54,297 epoch 16 - iter 14/17 - loss 2.46912466 - samples/sec: 274.40 - lr: 0.125000
2020-09-11 09:30:54,410 epoch 16 - iter 15/17 - loss 2.47079910 - samples/sec: 285.36 - lr: 0.125000
2020-09-11 09:30:54,521 epoch 16 - iter 16/17 - loss 2.49755961 - samples/sec: 292.36 - lr: 0.125000
2020-09-11 09:30:54,597 epoch 16 - iter 17/17 - loss 2.52624600 - samples/sec: 426.15 - lr: 0.125000
2020-09-11 09:30:54,598 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:54,600 EPOCH 16 done: loss 2.5262 - lr 0.1250000
2020-09-11 09:30:54,681 DEV : loss 2.1843576431274414 - score 0.8964
2020-09-11 09:30:54,684 BAD EPOCHS (no improvement): 1
2020-09-11 09:30:54,686 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:54,806 epoch 17 - iter 1/17 - loss 2.76969957 - samples/sec: 270.53 - lr: 0.125000
2020-09-11 09:30:54,922 epoch 17 - iter 2/17 - loss 2.50936592 - samples/sec: 280.56 - lr: 0.125000
2020-09-11 09:30:55,046 epoch 17 - iter 3/17 - loss 2.33596226 - samples/sec: 258.96 - lr: 0.125000
2020-09-11 09:30:55,162 epoch 17 - iter 4/17 - loss 2.50554565 - samples/sec: 279.75 - lr: 0.125000
2020-09-11 09:30:55,265 epoch 17 - iter 5/17 - loss 2.27989476 - samples/sec: 316.59 - lr: 0.125000
2020-09-11 09:30:55,383 epoch 17 - iter 6/17 - loss 2.29474705 - samples/sec: 274.08 - lr: 0.125000
2020-09-11 09:30:55,494 epoch 17 - iter 7/17 - loss 2.42519081 - samples/sec: 290.64 - lr: 0.125000
2020-09-11 09:30:55,600 epoch 17 - iter 8/17 - loss 2.52284651 - samples/sec: 303.65 - lr: 0.125000
2020-09-11 09:30:55,702 epoch 17 - iter 9/17 - loss 2.41699484 - samples/sec: 318.36 - lr: 0.125000
2020-09-11 09:30:55,808 epoch 17 - iter 10/17 - loss 2.39273806 - samples/sec: 303.67 - lr: 0.125000
2020-09-11 09:30:55,944 epoch 17 - iter 11/17 - loss 2.39208945 - samples/sec: 238.63 - lr: 0.125000
2020-09-11 09:30:56,054 epoch 17 - iter 12/17 - loss 2.32023218 - samples/sec: 293.26 - lr: 0.125000
2020-09-11 09:30:56,160 epoch 17 - iter 13/17 - loss 2.35094486 - samples/sec: 305.68 - lr: 0.125000
2020-09-11 09:30:56,270 epoch 17 - iter 14/17 - loss 2.42827324 - samples/sec: 296.42 - lr: 0.125000
2020-09-11 09:30:56,396 epoch 17 - iter 15/17 - loss 2.45238497 - samples/sec: 257.97 - lr: 0.125000
2020-09-11 09:30:56,507 epoch 17 - iter 16/17 - loss 2.39506280 - samples/sec: 292.19 - lr: 0.125000
2020-09-11 09:30:56,584 epoch 17 - iter 17/17 - loss 2.49167655 - samples/sec: 422.23 - lr: 0.125000
2020-09-11 09:30:56,585 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:56,590 EPOCH 17 done: loss 2.4917 - lr 0.1250000
2020-09-11 09:30:56,669 DEV : loss 2.249250888824463 - score 0.8912
2020-09-11 09:30:56,672 BAD EPOCHS (no improvement): 2
2020-09-11 09:30:56,674 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:56,786 epoch 18 - iter 1/17 - loss 1.70234942 - samples/sec: 297.37 - lr: 0.125000
2020-09-11 09:30:56,888 epoch 18 - iter 2/17 - loss 1.85333061 - samples/sec: 315.17 - lr: 0.125000
2020-09-11 09:30:56,998 epoch 18 - iter 3/17 - loss 1.89275738 - samples/sec: 294.29 - lr: 0.125000
2020-09-11 09:30:57,114 epoch 18 - iter 4/17 - loss 1.86746427 - samples/sec: 278.81 - lr: 0.125000
2020-09-11 09:30:57,222 epoch 18 - iter 5/17 - loss 2.20966289 - samples/sec: 299.47 - lr: 0.125000
2020-09-11 09:30:57,332 epoch 18 - iter 6/17 - loss 2.16520981 - samples/sec: 293.48 - lr: 0.125000
2020-09-11 09:30:57,443 epoch 18 - iter 7/17 - loss 2.11290937 - samples/sec: 292.71 - lr: 0.125000
2020-09-11 09:30:57,543 epoch 18 - iter 8/17 - loss 2.15310632 - samples/sec: 325.15 - lr: 0.125000
2020-09-11 09:30:57,652 epoch 18 - iter 9/17 - loss 2.18450381 - samples/sec: 296.34 - lr: 0.125000
2020-09-11 09:30:57,766 epoch 18 - iter 10/17 - loss 2.19876078 - samples/sec: 282.87 - lr: 0.125000
2020-09-11 09:30:57,908 epoch 18 - iter 11/17 - loss 2.24738829 - samples/sec: 228.46 - lr: 0.125000
2020-09-11 09:30:58,022 epoch 18 - iter 12/17 - loss 2.29583320 - samples/sec: 283.50 - lr: 0.125000
2020-09-11 09:30:58,126 epoch 18 - iter 13/17 - loss 2.42597394 - samples/sec: 310.80 - lr: 0.125000
2020-09-11 09:30:58,225 epoch 18 - iter 14/17 - loss 2.40854189 - samples/sec: 326.30 - lr: 0.125000
2020-09-11 09:30:58,330 epoch 18 - iter 15/17 - loss 2.55437588 - samples/sec: 307.40 - lr: 0.125000
2020-09-11 09:30:58,442 epoch 18 - iter 16/17 - loss 2.60625351 - samples/sec: 289.05 - lr: 0.125000
2020-09-11 09:30:58,514 epoch 18 - iter 17/17 - loss 2.60324173 - samples/sec: 452.42 - lr: 0.125000
2020-09-11 09:30:58,515 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:58,519 EPOCH 18 done: loss 2.6032 - lr 0.1250000
2020-09-11 09:30:58,605 DEV : loss 2.1045773029327393 - score 0.8901
2020-09-11 09:30:58,608 BAD EPOCHS (no improvement): 3
2020-09-11 09:30:58,610 ----------------------------------------------------------------------------------------------------
2020-09-11 09:30:58,727 epoch 19 - iter 1/17 - loss 2.05735612 - samples/sec: 278.66 - lr: 0.125000
2020-09-11 09:30:58,847 epoch 19 - iter 2/17 - loss 2.11045420 - samples/sec: 269.04 - lr: 0.125000
2020-09-11 09:30:58,974 epoch 19 - iter 3/17 - loss 2.07742627 - samples/sec: 254.10 - lr: 0.125000
2020-09-11 09:30:59,084 epoch 19 - iter 4/17 - loss 2.27933270 - samples/sec: 295.28 - lr: 0.125000
2020-09-11 09:30:59,190 epoch 19 - iter 5/17 - loss 2.52328839 - samples/sec: 303.98 - lr: 0.125000
2020-09-11 09:30:59,299 epoch 19 - iter 6/17 - loss 2.53958567 - samples/sec: 303.08 - lr: 0.125000
2020-09-11 09:30:59,415 epoch 19 - iter 7/17 - loss 2.43096699 - samples/sec: 280.08 - lr: 0.125000
2020-09-11 09:30:59,519 epoch 19 - iter 8/17 - loss 2.36722346 - samples/sec: 310.25 - lr: 0.125000
2020-09-11 09:30:59,625 epoch 19 - iter 9/17 - loss 2.32014529 - samples/sec: 305.11 - lr: 0.125000
2020-09-11 09:30:59,733 epoch 19 - iter 10/17 - loss 2.40720191 - samples/sec: 301.33 - lr: 0.125000
2020-09-11 09:30:59,859 epoch 19 - iter 11/17 - loss 2.39145192 - samples/sec: 255.91 - lr: 0.125000
2020-09-11 09:30:59,974 epoch 19 - iter 12/17 - loss 2.35232800 - samples/sec: 281.75 - lr: 0.125000
2020-09-11 09:31:00,092 epoch 19 - iter 13/17 - loss 2.34161122 - samples/sec: 280.33 - lr: 0.125000
2020-09-11 09:31:00,202 epoch 19 - iter 14/17 - loss 2.39985812 - samples/sec: 291.89 - lr: 0.125000
2020-09-11 09:31:00,315 epoch 19 - iter 15/17 - loss 2.34814619 - samples/sec: 289.62 - lr: 0.125000
2020-09-11 09:31:00,447 epoch 19 - iter 16/17 - loss 2.40359858 - samples/sec: 245.59 - lr: 0.125000
2020-09-11 09:31:00,524 epoch 19 - iter 17/17 - loss 2.37741771 - samples/sec: 445.32 - lr: 0.125000
2020-09-11 09:31:00,525 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:00,530 EPOCH 19 done: loss 2.3774 - lr 0.1250000
2020-09-11 09:31:00,607 DEV : loss 1.987985372543335 - score 0.9039
Epoch 19: reducing learning rate of group 0 to 6.2500e-02.
2020-09-11 09:31:00,611 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:00,614 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:00,726 epoch 20 - iter 1/17 - loss 1.71020281 - samples/sec: 296.47 - lr: 0.062500
2020-09-11 09:31:00,832 epoch 20 - iter 2/17 - loss 2.09335452 - samples/sec: 305.37 - lr: 0.062500
2020-09-11 09:31:00,942 epoch 20 - iter 3/17 - loss 2.28337387 - samples/sec: 293.37 - lr: 0.062500
2020-09-11 09:31:01,055 epoch 20 - iter 4/17 - loss 2.47729316 - samples/sec: 284.93 - lr: 0.062500
2020-09-11 09:31:01,159 epoch 20 - iter 5/17 - loss 2.56094611 - samples/sec: 314.81 - lr: 0.062500
2020-09-11 09:31:01,257 epoch 20 - iter 6/17 - loss 2.50572254 - samples/sec: 328.52 - lr: 0.062500
2020-09-11 09:31:01,380 epoch 20 - iter 7/17 - loss 2.47985882 - samples/sec: 262.47 - lr: 0.062500
2020-09-11 09:31:01,495 epoch 20 - iter 8/17 - loss 2.40888305 - samples/sec: 281.91 - lr: 0.062500
2020-09-11 09:31:01,597 epoch 20 - iter 9/17 - loss 2.43964536 - samples/sec: 320.29 - lr: 0.062500
2020-09-11 09:31:01,718 epoch 20 - iter 10/17 - loss 2.36184093 - samples/sec: 266.20 - lr: 0.062500
2020-09-11 09:31:01,834 epoch 20 - iter 11/17 - loss 2.33077885 - samples/sec: 277.68 - lr: 0.062500
2020-09-11 09:31:01,945 epoch 20 - iter 12/17 - loss 2.31888257 - samples/sec: 291.73 - lr: 0.062500
2020-09-11 09:31:02,060 epoch 20 - iter 13/17 - loss 2.30530426 - samples/sec: 281.27 - lr: 0.062500
2020-09-11 09:31:02,163 epoch 20 - iter 14/17 - loss 2.32191645 - samples/sec: 314.84 - lr: 0.062500
2020-09-11 09:31:02,267 epoch 20 - iter 15/17 - loss 2.34071354 - samples/sec: 309.11 - lr: 0.062500
2020-09-11 09:31:02,367 epoch 20 - iter 16/17 - loss 2.40924118 - samples/sec: 325.52 - lr: 0.062500
2020-09-11 09:31:02,444 epoch 20 - iter 17/17 - loss 2.38642045 - samples/sec: 422.96 - lr: 0.062500
2020-09-11 09:31:02,445 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:02,449 EPOCH 20 done: loss 2.3864 - lr 0.0625000
2020-09-11 09:31:02,524 DEV : loss 2.02628755569458 - score 0.9039
2020-09-11 09:31:02,526 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:02,531 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:02,634 epoch 21 - iter 1/17 - loss 2.01805162 - samples/sec: 313.46 - lr: 0.062500
2020-09-11 09:31:02,740 epoch 21 - iter 2/17 - loss 1.95761949 - samples/sec: 307.95 - lr: 0.062500
2020-09-11 09:31:02,865 epoch 21 - iter 3/17 - loss 2.05638238 - samples/sec: 257.24 - lr: 0.062500
2020-09-11 09:31:02,986 epoch 21 - iter 4/17 - loss 2.24450067 - samples/sec: 268.72 - lr: 0.062500
2020-09-11 09:31:03,099 epoch 21 - iter 5/17 - loss 2.27903502 - samples/sec: 285.77 - lr: 0.062500
2020-09-11 09:31:03,203 epoch 21 - iter 6/17 - loss 2.38084318 - samples/sec: 308.65 - lr: 0.062500
2020-09-11 09:31:03,310 epoch 21 - iter 7/17 - loss 2.30562082 - samples/sec: 303.05 - lr: 0.062500
2020-09-11 09:31:03,444 epoch 21 - iter 8/17 - loss 2.33702309 - samples/sec: 240.19 - lr: 0.062500
2020-09-11 09:31:03,550 epoch 21 - iter 9/17 - loss 2.24217778 - samples/sec: 309.72 - lr: 0.062500
2020-09-11 09:31:03,652 epoch 21 - iter 10/17 - loss 2.26832607 - samples/sec: 316.97 - lr: 0.062500
2020-09-11 09:31:03,762 epoch 21 - iter 11/17 - loss 2.26648381 - samples/sec: 294.96 - lr: 0.062500
2020-09-11 09:31:03,867 epoch 21 - iter 12/17 - loss 2.28058559 - samples/sec: 307.86 - lr: 0.062500
2020-09-11 09:31:03,976 epoch 21 - iter 13/17 - loss 2.25714514 - samples/sec: 296.97 - lr: 0.062500
2020-09-11 09:31:04,091 epoch 21 - iter 14/17 - loss 2.23559930 - samples/sec: 281.10 - lr: 0.062500
2020-09-11 09:31:04,206 epoch 21 - iter 15/17 - loss 2.27486047 - samples/sec: 283.20 - lr: 0.062500
2020-09-11 09:31:04,312 epoch 21 - iter 16/17 - loss 2.25765611 - samples/sec: 310.23 - lr: 0.062500
2020-09-11 09:31:04,382 epoch 21 - iter 17/17 - loss 2.24482668 - samples/sec: 463.98 - lr: 0.062500
2020-09-11 09:31:04,383 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:04,389 EPOCH 21 done: loss 2.2448 - lr 0.0625000
2020-09-11 09:31:04,476 DEV : loss 1.9274215698242188 - score 0.9039
2020-09-11 09:31:04,479 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:04,482 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:04,591 epoch 22 - iter 1/17 - loss 2.05779243 - samples/sec: 299.29 - lr: 0.062500
2020-09-11 09:31:04,712 epoch 22 - iter 2/17 - loss 2.54358244 - samples/sec: 267.17 - lr: 0.062500
2020-09-11 09:31:04,843 epoch 22 - iter 3/17 - loss 2.38451076 - samples/sec: 246.03 - lr: 0.062500
2020-09-11 09:31:04,950 epoch 22 - iter 4/17 - loss 2.44637775 - samples/sec: 303.17 - lr: 0.062500
2020-09-11 09:31:05,057 epoch 22 - iter 5/17 - loss 2.30466504 - samples/sec: 302.91 - lr: 0.062500
2020-09-11 09:31:05,178 epoch 22 - iter 6/17 - loss 2.26858521 - samples/sec: 266.10 - lr: 0.062500
2020-09-11 09:31:05,277 epoch 22 - iter 7/17 - loss 2.12332932 - samples/sec: 328.32 - lr: 0.062500
2020-09-11 09:31:05,377 epoch 22 - iter 8/17 - loss 2.16235070 - samples/sec: 321.38 - lr: 0.062500
2020-09-11 09:31:05,507 epoch 22 - iter 9/17 - loss 2.22368626 - samples/sec: 253.23 - lr: 0.062500
2020-09-11 09:31:05,611 epoch 22 - iter 10/17 - loss 2.20117749 - samples/sec: 310.81 - lr: 0.062500
2020-09-11 09:31:05,711 epoch 22 - iter 11/17 - loss 2.23854591 - samples/sec: 323.73 - lr: 0.062500
2020-09-11 09:31:05,832 epoch 22 - iter 12/17 - loss 2.23895742 - samples/sec: 264.75 - lr: 0.062500
2020-09-11 09:31:05,976 epoch 22 - iter 13/17 - loss 2.28941275 - samples/sec: 224.52 - lr: 0.062500
2020-09-11 09:31:06,084 epoch 22 - iter 14/17 - loss 2.28183858 - samples/sec: 302.04 - lr: 0.062500
2020-09-11 09:31:06,196 epoch 22 - iter 15/17 - loss 2.27302086 - samples/sec: 286.23 - lr: 0.062500
2020-09-11 09:31:06,309 epoch 22 - iter 16/17 - loss 2.25989293 - samples/sec: 287.92 - lr: 0.062500
2020-09-11 09:31:06,386 epoch 22 - iter 17/17 - loss 2.19085824 - samples/sec: 419.62 - lr: 0.062500
2020-09-11 09:31:06,388 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:06,393 EPOCH 22 done: loss 2.1909 - lr 0.0625000
2020-09-11 09:31:06,480 DEV : loss 1.9331698417663574 - score 0.9039
2020-09-11 09:31:06,484 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:06,486 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:06,601 epoch 23 - iter 1/17 - loss 1.70134401 - samples/sec: 283.94 - lr: 0.062500
2020-09-11 09:31:06,711 epoch 23 - iter 2/17 - loss 2.13410854 - samples/sec: 299.57 - lr: 0.062500
2020-09-11 09:31:06,818 epoch 23 - iter 3/17 - loss 2.12737036 - samples/sec: 301.17 - lr: 0.062500
2020-09-11 09:31:06,945 epoch 23 - iter 4/17 - loss 2.07283798 - samples/sec: 254.23 - lr: 0.062500
2020-09-11 09:31:07,062 epoch 23 - iter 5/17 - loss 2.13042910 - samples/sec: 277.10 - lr: 0.062500
2020-09-11 09:31:07,180 epoch 23 - iter 6/17 - loss 2.22374584 - samples/sec: 273.38 - lr: 0.062500
2020-09-11 09:31:07,293 epoch 23 - iter 7/17 - loss 2.24066106 - samples/sec: 285.75 - lr: 0.062500
2020-09-11 09:31:07,402 epoch 23 - iter 8/17 - loss 2.25720285 - samples/sec: 298.11 - lr: 0.062500
2020-09-11 09:31:07,520 epoch 23 - iter 9/17 - loss 2.20347146 - samples/sec: 273.03 - lr: 0.062500
2020-09-11 09:31:07,633 epoch 23 - iter 10/17 - loss 2.24449781 - samples/sec: 285.70 - lr: 0.062500
2020-09-11 09:31:07,744 epoch 23 - iter 11/17 - loss 2.31409958 - samples/sec: 292.72 - lr: 0.062500
2020-09-11 09:31:07,860 epoch 23 - iter 12/17 - loss 2.29987424 - samples/sec: 279.58 - lr: 0.062500
2020-09-11 09:31:07,986 epoch 23 - iter 13/17 - loss 2.30229404 - samples/sec: 255.65 - lr: 0.062500
2020-09-11 09:31:08,088 epoch 23 - iter 14/17 - loss 2.32035975 - samples/sec: 317.76 - lr: 0.062500
2020-09-11 09:31:08,205 epoch 23 - iter 15/17 - loss 2.29056011 - samples/sec: 275.83 - lr: 0.062500
2020-09-11 09:31:08,321 epoch 23 - iter 16/17 - loss 2.27713012 - samples/sec: 280.00 - lr: 0.062500
2020-09-11 09:31:08,399 epoch 23 - iter 17/17 - loss 2.26789435 - samples/sec: 413.97 - lr: 0.062500
2020-09-11 09:31:08,401 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:08,406 EPOCH 23 done: loss 2.2679 - lr 0.0625000
2020-09-11 09:31:08,484 DEV : loss 1.9181852340698242 - score 0.9039
Epoch 23: reducing learning rate of group 0 to 3.1250e-02.
2020-09-11 09:31:08,489 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:08,490 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:08,603 epoch 24 - iter 1/17 - loss 2.76224279 - samples/sec: 300.34 - lr: 0.031250
2020-09-11 09:31:08,713 epoch 24 - iter 2/17 - loss 2.52293849 - samples/sec: 293.60 - lr: 0.031250
2020-09-11 09:31:08,826 epoch 24 - iter 3/17 - loss 2.32014473 - samples/sec: 285.65 - lr: 0.031250
2020-09-11 09:31:08,939 epoch 24 - iter 4/17 - loss 2.27968848 - samples/sec: 288.03 - lr: 0.031250
2020-09-11 09:31:09,047 epoch 24 - iter 5/17 - loss 2.19507091 - samples/sec: 300.27 - lr: 0.031250
2020-09-11 09:31:09,143 epoch 24 - iter 6/17 - loss 2.15499967 - samples/sec: 335.25 - lr: 0.031250
2020-09-11 09:31:09,262 epoch 24 - iter 7/17 - loss 2.34728897 - samples/sec: 271.65 - lr: 0.031250
2020-09-11 09:31:09,371 epoch 24 - iter 8/17 - loss 2.23135819 - samples/sec: 297.18 - lr: 0.031250
2020-09-11 09:31:09,487 epoch 24 - iter 9/17 - loss 2.35803510 - samples/sec: 279.54 - lr: 0.031250
2020-09-11 09:31:09,603 epoch 24 - iter 10/17 - loss 2.30058753 - samples/sec: 279.13 - lr: 0.031250
2020-09-11 09:31:09,719 epoch 24 - iter 11/17 - loss 2.25945546 - samples/sec: 280.01 - lr: 0.031250
2020-09-11 09:31:09,835 epoch 24 - iter 12/17 - loss 2.27051020 - samples/sec: 278.96 - lr: 0.031250
2020-09-11 09:31:09,945 epoch 24 - iter 13/17 - loss 2.23163500 - samples/sec: 292.98 - lr: 0.031250
2020-09-11 09:31:10,057 epoch 24 - iter 14/17 - loss 2.25586351 - samples/sec: 290.07 - lr: 0.031250
2020-09-11 09:31:10,167 epoch 24 - iter 15/17 - loss 2.26975231 - samples/sec: 291.75 - lr: 0.031250
2020-09-11 09:31:10,284 epoch 24 - iter 16/17 - loss 2.23412069 - samples/sec: 278.97 - lr: 0.031250
2020-09-11 09:31:10,378 epoch 24 - iter 17/17 - loss 2.21847627 - samples/sec: 346.19 - lr: 0.031250
2020-09-11 09:31:10,379 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:10,385 EPOCH 24 done: loss 2.2185 - lr 0.0312500
2020-09-11 09:31:10,464 DEV : loss 1.9145869016647339 - score 0.9039
2020-09-11 09:31:10,467 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:10,469 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:10,576 epoch 25 - iter 1/17 - loss 2.68180037 - samples/sec: 308.77 - lr: 0.031250
2020-09-11 09:31:10,701 epoch 25 - iter 2/17 - loss 2.58654761 - samples/sec: 259.33 - lr: 0.031250
2020-09-11 09:31:10,806 epoch 25 - iter 3/17 - loss 2.54156987 - samples/sec: 306.55 - lr: 0.031250
2020-09-11 09:31:10,922 epoch 25 - iter 4/17 - loss 2.45042753 - samples/sec: 278.73 - lr: 0.031250
2020-09-11 09:31:11,030 epoch 25 - iter 5/17 - loss 2.37050085 - samples/sec: 300.68 - lr: 0.031250
2020-09-11 09:31:11,131 epoch 25 - iter 6/17 - loss 2.17418148 - samples/sec: 320.10 - lr: 0.031250
2020-09-11 09:31:11,238 epoch 25 - iter 7/17 - loss 2.03742301 - samples/sec: 303.48 - lr: 0.031250
2020-09-11 09:31:11,349 epoch 25 - iter 8/17 - loss 2.02971898 - samples/sec: 293.46 - lr: 0.031250
2020-09-11 09:31:11,475 epoch 25 - iter 9/17 - loss 1.98951248 - samples/sec: 257.54 - lr: 0.031250
2020-09-11 09:31:11,590 epoch 25 - iter 10/17 - loss 2.04102901 - samples/sec: 279.25 - lr: 0.031250
2020-09-11 09:31:11,702 epoch 25 - iter 11/17 - loss 2.12209180 - samples/sec: 289.60 - lr: 0.031250
2020-09-11 09:31:11,808 epoch 25 - iter 12/17 - loss 2.12565103 - samples/sec: 307.27 - lr: 0.031250
2020-09-11 09:31:11,916 epoch 25 - iter 13/17 - loss 2.21383809 - samples/sec: 301.98 - lr: 0.031250
2020-09-11 09:31:12,021 epoch 25 - iter 14/17 - loss 2.22879870 - samples/sec: 308.28 - lr: 0.031250
2020-09-11 09:31:12,143 epoch 25 - iter 15/17 - loss 2.25462717 - samples/sec: 265.38 - lr: 0.031250
2020-09-11 09:31:12,261 epoch 25 - iter 16/17 - loss 2.24462736 - samples/sec: 273.45 - lr: 0.031250
2020-09-11 09:31:12,339 epoch 25 - iter 17/17 - loss 2.26452670 - samples/sec: 422.20 - lr: 0.031250
2020-09-11 09:31:12,340 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:12,342 EPOCH 25 done: loss 2.2645 - lr 0.0312500
2020-09-11 09:31:12,425 DEV : loss 1.8926992416381836 - score 0.9039
2020-09-11 09:31:12,429 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:12,431 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:12,547 epoch 26 - iter 1/17 - loss 1.52716124 - samples/sec: 279.51 - lr: 0.031250
2020-09-11 09:31:12,670 epoch 26 - iter 2/17 - loss 1.83199435 - samples/sec: 264.82 - lr: 0.031250
2020-09-11 09:31:12,783 epoch 26 - iter 3/17 - loss 1.96283241 - samples/sec: 291.04 - lr: 0.031250
2020-09-11 09:31:12,890 epoch 26 - iter 4/17 - loss 1.96469370 - samples/sec: 304.40 - lr: 0.031250
2020-09-11 09:31:13,001 epoch 26 - iter 5/17 - loss 2.03075287 - samples/sec: 290.77 - lr: 0.031250
2020-09-11 09:31:13,114 epoch 26 - iter 6/17 - loss 2.18692909 - samples/sec: 286.41 - lr: 0.031250
2020-09-11 09:31:13,238 epoch 26 - iter 7/17 - loss 2.29399492 - samples/sec: 261.06 - lr: 0.031250
2020-09-11 09:31:13,359 epoch 26 - iter 8/17 - loss 2.21845753 - samples/sec: 266.94 - lr: 0.031250
2020-09-11 09:31:13,465 epoch 26 - iter 9/17 - loss 2.27986459 - samples/sec: 304.67 - lr: 0.031250
2020-09-11 09:31:13,574 epoch 26 - iter 10/17 - loss 2.31232477 - samples/sec: 297.41 - lr: 0.031250
2020-09-11 09:31:13,690 epoch 26 - iter 11/17 - loss 2.29177281 - samples/sec: 282.18 - lr: 0.031250
2020-09-11 09:31:13,811 epoch 26 - iter 12/17 - loss 2.25202786 - samples/sec: 267.51 - lr: 0.031250
2020-09-11 09:31:13,918 epoch 26 - iter 13/17 - loss 2.21773954 - samples/sec: 302.48 - lr: 0.031250
2020-09-11 09:31:14,029 epoch 26 - iter 14/17 - loss 2.19228974 - samples/sec: 290.48 - lr: 0.031250
2020-09-11 09:31:14,140 epoch 26 - iter 15/17 - loss 2.16736569 - samples/sec: 292.30 - lr: 0.031250
2020-09-11 09:31:14,245 epoch 26 - iter 16/17 - loss 2.16703309 - samples/sec: 307.62 - lr: 0.031250
2020-09-11 09:31:14,336 epoch 26 - iter 17/17 - loss 2.20482035 - samples/sec: 355.35 - lr: 0.031250
2020-09-11 09:31:14,337 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:14,339 EPOCH 26 done: loss 2.2048 - lr 0.0312500
2020-09-11 09:31:14,418 DEV : loss 1.829909324645996 - score 0.9039
2020-09-11 09:31:14,420 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:14,423 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:14,535 epoch 27 - iter 1/17 - loss 2.94158649 - samples/sec: 289.52 - lr: 0.031250
2020-09-11 09:31:14,650 epoch 27 - iter 2/17 - loss 2.31759334 - samples/sec: 279.59 - lr: 0.031250
2020-09-11 09:31:14,769 epoch 27 - iter 3/17 - loss 2.34357309 - samples/sec: 274.76 - lr: 0.031250
2020-09-11 09:31:14,901 epoch 27 - iter 4/17 - loss 2.31417656 - samples/sec: 252.17 - lr: 0.031250
2020-09-11 09:31:15,019 epoch 27 - iter 5/17 - loss 2.19485738 - samples/sec: 274.71 - lr: 0.031250
2020-09-11 09:31:15,131 epoch 27 - iter 6/17 - loss 2.24387747 - samples/sec: 289.75 - lr: 0.031250
2020-09-11 09:31:15,248 epoch 27 - iter 7/17 - loss 2.31949151 - samples/sec: 276.19 - lr: 0.031250
2020-09-11 09:31:15,372 epoch 27 - iter 8/17 - loss 2.27232495 - samples/sec: 261.82 - lr: 0.031250
2020-09-11 09:31:15,501 epoch 27 - iter 9/17 - loss 2.23626822 - samples/sec: 251.21 - lr: 0.031250
2020-09-11 09:31:15,634 epoch 27 - iter 10/17 - loss 2.19940158 - samples/sec: 243.44 - lr: 0.031250
2020-09-11 09:31:15,745 epoch 27 - iter 11/17 - loss 2.25831646 - samples/sec: 289.98 - lr: 0.031250
2020-09-11 09:31:15,860 epoch 27 - iter 12/17 - loss 2.20441757 - samples/sec: 286.89 - lr: 0.031250
2020-09-11 09:31:15,971 epoch 27 - iter 13/17 - loss 2.22872842 - samples/sec: 297.65 - lr: 0.031250
2020-09-11 09:31:16,081 epoch 27 - iter 14/17 - loss 2.20987686 - samples/sec: 292.10 - lr: 0.031250
2020-09-11 09:31:16,191 epoch 27 - iter 15/17 - loss 2.19334530 - samples/sec: 295.28 - lr: 0.031250
2020-09-11 09:31:16,299 epoch 27 - iter 16/17 - loss 2.17026734 - samples/sec: 297.75 - lr: 0.031250
2020-09-11 09:31:16,393 epoch 27 - iter 17/17 - loss 2.12320886 - samples/sec: 344.81 - lr: 0.031250
2020-09-11 09:31:16,394 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:16,396 EPOCH 27 done: loss 2.1232 - lr 0.0312500
2020-09-11 09:31:16,477 DEV : loss 1.877849817276001 - score 0.9039
Epoch 27: reducing learning rate of group 0 to 1.5625e-02.
2020-09-11 09:31:16,480 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:16,482 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:16,598 epoch 28 - iter 1/17 - loss 1.15935516 - samples/sec: 281.50 - lr: 0.015625
2020-09-11 09:31:16,730 epoch 28 - iter 2/17 - loss 1.71386480 - samples/sec: 244.95 - lr: 0.015625
2020-09-11 09:31:16,843 epoch 28 - iter 3/17 - loss 1.80357365 - samples/sec: 288.82 - lr: 0.015625
2020-09-11 09:31:16,953 epoch 28 - iter 4/17 - loss 1.88561693 - samples/sec: 293.88 - lr: 0.015625
2020-09-11 09:31:17,069 epoch 28 - iter 5/17 - loss 1.87196920 - samples/sec: 280.24 - lr: 0.015625
2020-09-11 09:31:17,191 epoch 28 - iter 6/17 - loss 1.89888237 - samples/sec: 264.96 - lr: 0.015625
2020-09-11 09:31:17,299 epoch 28 - iter 7/17 - loss 2.00117913 - samples/sec: 298.19 - lr: 0.015625
2020-09-11 09:31:17,419 epoch 28 - iter 8/17 - loss 1.97936609 - samples/sec: 269.09 - lr: 0.015625
2020-09-11 09:31:17,526 epoch 28 - iter 9/17 - loss 1.89083740 - samples/sec: 304.52 - lr: 0.015625
2020-09-11 09:31:17,635 epoch 28 - iter 10/17 - loss 2.04024680 - samples/sec: 298.68 - lr: 0.015625
2020-09-11 09:31:17,750 epoch 28 - iter 11/17 - loss 2.15808472 - samples/sec: 280.07 - lr: 0.015625
2020-09-11 09:31:17,862 epoch 28 - iter 12/17 - loss 2.18874393 - samples/sec: 288.12 - lr: 0.015625
2020-09-11 09:31:17,979 epoch 28 - iter 13/17 - loss 2.18398133 - samples/sec: 277.47 - lr: 0.015625
2020-09-11 09:31:18,107 epoch 28 - iter 14/17 - loss 2.23524254 - samples/sec: 252.52 - lr: 0.015625
2020-09-11 09:31:18,209 epoch 28 - iter 15/17 - loss 2.20884586 - samples/sec: 317.97 - lr: 0.015625
2020-09-11 09:31:18,328 epoch 28 - iter 16/17 - loss 2.19283289 - samples/sec: 271.69 - lr: 0.015625
2020-09-11 09:31:18,411 epoch 28 - iter 17/17 - loss 2.17065479 - samples/sec: 389.14 - lr: 0.015625
2020-09-11 09:31:18,412 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:18,416 EPOCH 28 done: loss 2.1707 - lr 0.0156250
2020-09-11 09:31:18,497 DEV : loss 1.8591759204864502 - score 0.9039
2020-09-11 09:31:18,501 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:18,504 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:18,616 epoch 29 - iter 1/17 - loss 2.38848042 - samples/sec: 295.16 - lr: 0.015625
2020-09-11 09:31:18,730 epoch 29 - iter 2/17 - loss 2.01982307 - samples/sec: 281.73 - lr: 0.015625
2020-09-11 09:31:18,842 epoch 29 - iter 3/17 - loss 2.28015947 - samples/sec: 289.15 - lr: 0.015625
2020-09-11 09:31:18,973 epoch 29 - iter 4/17 - loss 2.35284030 - samples/sec: 248.54 - lr: 0.015625
2020-09-11 09:31:19,100 epoch 29 - iter 5/17 - loss 2.46799192 - samples/sec: 253.52 - lr: 0.015625
2020-09-11 09:31:19,208 epoch 29 - iter 6/17 - loss 2.40597574 - samples/sec: 298.77 - lr: 0.015625
2020-09-11 09:31:19,316 epoch 29 - iter 7/17 - loss 2.27038290 - samples/sec: 301.48 - lr: 0.015625
2020-09-11 09:31:19,432 epoch 29 - iter 8/17 - loss 2.20697598 - samples/sec: 277.52 - lr: 0.015625
2020-09-11 09:31:19,545 epoch 29 - iter 9/17 - loss 2.14561096 - samples/sec: 286.38 - lr: 0.015625
2020-09-11 09:31:19,656 epoch 29 - iter 10/17 - loss 2.20748037 - samples/sec: 293.48 - lr: 0.015625
2020-09-11 09:31:19,763 epoch 29 - iter 11/17 - loss 2.22472450 - samples/sec: 302.61 - lr: 0.015625
2020-09-11 09:31:19,878 epoch 29 - iter 12/17 - loss 2.24347988 - samples/sec: 286.50 - lr: 0.015625
2020-09-11 09:31:19,994 epoch 29 - iter 13/17 - loss 2.21828763 - samples/sec: 277.90 - lr: 0.015625
2020-09-11 09:31:20,105 epoch 29 - iter 14/17 - loss 2.18254331 - samples/sec: 294.47 - lr: 0.015625
2020-09-11 09:31:20,213 epoch 29 - iter 15/17 - loss 2.22664884 - samples/sec: 299.30 - lr: 0.015625
2020-09-11 09:31:20,322 epoch 29 - iter 16/17 - loss 2.23254411 - samples/sec: 297.41 - lr: 0.015625
2020-09-11 09:31:20,399 epoch 29 - iter 17/17 - loss 2.21945810 - samples/sec: 422.06 - lr: 0.015625
2020-09-11 09:31:20,400 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:20,402 EPOCH 29 done: loss 2.2195 - lr 0.0156250
2020-09-11 09:31:20,492 DEV : loss 1.8752442598342896 - score 0.9039
2020-09-11 09:31:20,496 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:20,497 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:20,607 epoch 30 - iter 1/17 - loss 1.82208562 - samples/sec: 297.46 - lr: 0.015625
2020-09-11 09:31:20,721 epoch 30 - iter 2/17 - loss 2.00882280 - samples/sec: 284.76 - lr: 0.015625
2020-09-11 09:31:20,838 epoch 30 - iter 3/17 - loss 2.11731188 - samples/sec: 276.32 - lr: 0.015625
2020-09-11 09:31:20,946 epoch 30 - iter 4/17 - loss 1.95303643 - samples/sec: 300.17 - lr: 0.015625
2020-09-11 09:31:21,047 epoch 30 - iter 5/17 - loss 2.05010142 - samples/sec: 322.80 - lr: 0.015625
2020-09-11 09:31:21,164 epoch 30 - iter 6/17 - loss 2.18919861 - samples/sec: 276.19 - lr: 0.015625
2020-09-11 09:31:21,273 epoch 30 - iter 7/17 - loss 2.20131217 - samples/sec: 298.11 - lr: 0.015625
2020-09-11 09:31:21,389 epoch 30 - iter 8/17 - loss 2.24543753 - samples/sec: 282.40 - lr: 0.015625
2020-09-11 09:31:21,506 epoch 30 - iter 9/17 - loss 2.20305248 - samples/sec: 277.27 - lr: 0.015625
2020-09-11 09:31:21,611 epoch 30 - iter 10/17 - loss 2.13625307 - samples/sec: 309.91 - lr: 0.015625
2020-09-11 09:31:21,733 epoch 30 - iter 11/17 - loss 2.14368816 - samples/sec: 264.68 - lr: 0.015625
2020-09-11 09:31:21,870 epoch 30 - iter 12/17 - loss 2.21769957 - samples/sec: 235.08 - lr: 0.015625
2020-09-11 09:31:21,980 epoch 30 - iter 13/17 - loss 2.18367576 - samples/sec: 297.07 - lr: 0.015625
2020-09-11 09:31:22,087 epoch 30 - iter 14/17 - loss 2.16648819 - samples/sec: 303.74 - lr: 0.015625
2020-09-11 09:31:22,193 epoch 30 - iter 15/17 - loss 2.19202752 - samples/sec: 304.22 - lr: 0.015625
2020-09-11 09:31:22,301 epoch 30 - iter 16/17 - loss 2.19048543 - samples/sec: 304.97 - lr: 0.015625
2020-09-11 09:31:22,394 epoch 30 - iter 17/17 - loss 2.21348506 - samples/sec: 351.18 - lr: 0.015625
2020-09-11 09:31:22,395 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:22,400 EPOCH 30 done: loss 2.2135 - lr 0.0156250
2020-09-11 09:31:22,481 DEV : loss 1.8820831775665283 - score 0.9039
2020-09-11 09:31:22,484 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:22,486 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:22,600 epoch 31 - iter 1/17 - loss 2.38624716 - samples/sec: 288.82 - lr: 0.015625
2020-09-11 09:31:22,707 epoch 31 - iter 2/17 - loss 2.14800692 - samples/sec: 304.05 - lr: 0.015625
2020-09-11 09:31:22,816 epoch 31 - iter 3/17 - loss 1.89553030 - samples/sec: 296.20 - lr: 0.015625
2020-09-11 09:31:22,921 epoch 31 - iter 4/17 - loss 1.80254796 - samples/sec: 308.79 - lr: 0.015625
2020-09-11 09:31:23,025 epoch 31 - iter 5/17 - loss 2.05551784 - samples/sec: 312.82 - lr: 0.015625
2020-09-11 09:31:23,132 epoch 31 - iter 6/17 - loss 2.07799266 - samples/sec: 303.13 - lr: 0.015625
2020-09-11 09:31:23,242 epoch 31 - iter 7/17 - loss 2.12334066 - samples/sec: 292.67 - lr: 0.015625
2020-09-11 09:31:23,354 epoch 31 - iter 8/17 - loss 2.12068520 - samples/sec: 289.28 - lr: 0.015625
2020-09-11 09:31:23,464 epoch 31 - iter 9/17 - loss 2.06356608 - samples/sec: 295.16 - lr: 0.015625
2020-09-11 09:31:23,596 epoch 31 - iter 10/17 - loss 2.07550075 - samples/sec: 244.25 - lr: 0.015625
2020-09-11 09:31:23,718 epoch 31 - iter 11/17 - loss 2.08892419 - samples/sec: 264.35 - lr: 0.015625
2020-09-11 09:31:23,850 epoch 31 - iter 12/17 - loss 2.10753208 - samples/sec: 245.54 - lr: 0.015625
2020-09-11 09:31:23,961 epoch 31 - iter 13/17 - loss 2.11423028 - samples/sec: 292.27 - lr: 0.015625
2020-09-11 09:31:24,075 epoch 31 - iter 14/17 - loss 2.10826038 - samples/sec: 284.92 - lr: 0.015625
2020-09-11 09:31:24,180 epoch 31 - iter 15/17 - loss 2.15856535 - samples/sec: 306.64 - lr: 0.015625
2020-09-11 09:31:24,289 epoch 31 - iter 16/17 - loss 2.20597644 - samples/sec: 296.83 - lr: 0.015625
2020-09-11 09:31:24,363 epoch 31 - iter 17/17 - loss 2.15071367 - samples/sec: 441.66 - lr: 0.015625
2020-09-11 09:31:24,364 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:24,369 EPOCH 31 done: loss 2.1507 - lr 0.0156250
2020-09-11 09:31:24,448 DEV : loss 1.87209951877594 - score 0.9039
Epoch 31: reducing learning rate of group 0 to 7.8125e-03.
2020-09-11 09:31:24,450 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:24,454 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:24,587 epoch 32 - iter 1/17 - loss 3.31541109 - samples/sec: 244.62 - lr: 0.007812
2020-09-11 09:31:24,699 epoch 32 - iter 2/17 - loss 2.87909269 - samples/sec: 292.61 - lr: 0.007812
2020-09-11 09:31:24,817 epoch 32 - iter 3/17 - loss 2.39622060 - samples/sec: 274.76 - lr: 0.007812
2020-09-11 09:31:24,922 epoch 32 - iter 4/17 - loss 2.27022970 - samples/sec: 307.60 - lr: 0.007812
2020-09-11 09:31:25,030 epoch 32 - iter 5/17 - loss 2.35990863 - samples/sec: 299.72 - lr: 0.007812
2020-09-11 09:31:25,138 epoch 32 - iter 6/17 - loss 2.25238045 - samples/sec: 299.64 - lr: 0.007812
2020-09-11 09:31:25,252 epoch 32 - iter 7/17 - loss 2.31877245 - samples/sec: 281.27 - lr: 0.007812
2020-09-11 09:31:25,359 epoch 32 - iter 8/17 - loss 2.32546023 - samples/sec: 302.22 - lr: 0.007812
2020-09-11 09:31:25,465 epoch 32 - iter 9/17 - loss 2.25857042 - samples/sec: 305.31 - lr: 0.007812
2020-09-11 09:31:25,578 epoch 32 - iter 10/17 - loss 2.16616904 - samples/sec: 284.42 - lr: 0.007812
2020-09-11 09:31:25,678 epoch 32 - iter 11/17 - loss 2.17235703 - samples/sec: 325.38 - lr: 0.007812
2020-09-11 09:31:25,788 epoch 32 - iter 12/17 - loss 2.24928207 - samples/sec: 292.01 - lr: 0.007812
2020-09-11 09:31:25,887 epoch 32 - iter 13/17 - loss 2.20171934 - samples/sec: 327.47 - lr: 0.007812
2020-09-11 09:31:25,994 epoch 32 - iter 14/17 - loss 2.19627576 - samples/sec: 304.41 - lr: 0.007812
2020-09-11 09:31:26,097 epoch 32 - iter 15/17 - loss 2.15969448 - samples/sec: 314.44 - lr: 0.007812
2020-09-11 09:31:26,196 epoch 32 - iter 16/17 - loss 2.15566656 - samples/sec: 324.70 - lr: 0.007812
2020-09-11 09:31:26,277 epoch 32 - iter 17/17 - loss 2.12638341 - samples/sec: 401.92 - lr: 0.007812
2020-09-11 09:31:26,278 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:26,283 EPOCH 32 done: loss 2.1264 - lr 0.0078125
2020-09-11 09:31:26,360 DEV : loss 1.8655517101287842 - score 0.9039
2020-09-11 09:31:26,363 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:26,367 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:26,472 epoch 33 - iter 1/17 - loss 2.11451244 - samples/sec: 307.77 - lr: 0.007812
2020-09-11 09:31:26,571 epoch 33 - iter 2/17 - loss 1.94452286 - samples/sec: 329.06 - lr: 0.007812
2020-09-11 09:31:26,692 epoch 33 - iter 3/17 - loss 2.11479004 - samples/sec: 269.04 - lr: 0.007812
2020-09-11 09:31:26,810 epoch 33 - iter 4/17 - loss 2.03008524 - samples/sec: 273.58 - lr: 0.007812
2020-09-11 09:31:26,921 epoch 33 - iter 5/17 - loss 1.94079068 - samples/sec: 293.75 - lr: 0.007812
2020-09-11 09:31:27,030 epoch 33 - iter 6/17 - loss 2.00068849 - samples/sec: 296.73 - lr: 0.007812
2020-09-11 09:31:27,143 epoch 33 - iter 7/17 - loss 2.01924847 - samples/sec: 288.16 - lr: 0.007812
2020-09-11 09:31:27,251 epoch 33 - iter 8/17 - loss 2.11283641 - samples/sec: 297.76 - lr: 0.007812
2020-09-11 09:31:27,358 epoch 33 - iter 9/17 - loss 2.04315246 - samples/sec: 303.98 - lr: 0.007812
2020-09-11 09:31:27,473 epoch 33 - iter 10/17 - loss 2.06819868 - samples/sec: 280.28 - lr: 0.007812
2020-09-11 09:31:27,583 epoch 33 - iter 11/17 - loss 2.12044837 - samples/sec: 297.79 - lr: 0.007812
2020-09-11 09:31:27,693 epoch 33 - iter 12/17 - loss 2.14381409 - samples/sec: 293.06 - lr: 0.007812
2020-09-11 09:31:27,802 epoch 33 - iter 13/17 - loss 2.14429503 - samples/sec: 297.72 - lr: 0.007812
2020-09-11 09:31:27,913 epoch 33 - iter 14/17 - loss 2.10765451 - samples/sec: 291.99 - lr: 0.007812
2020-09-11 09:31:28,053 epoch 33 - iter 15/17 - loss 2.13644788 - samples/sec: 230.54 - lr: 0.007812
2020-09-11 09:31:28,157 epoch 33 - iter 16/17 - loss 2.16665895 - samples/sec: 310.85 - lr: 0.007812
2020-09-11 09:31:28,240 epoch 33 - iter 17/17 - loss 2.18356979 - samples/sec: 395.16 - lr: 0.007812
2020-09-11 09:31:28,241 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:28,246 EPOCH 33 done: loss 2.1836 - lr 0.0078125
2020-09-11 09:31:28,324 DEV : loss 1.860945701599121 - score 0.9039
2020-09-11 09:31:28,326 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:28,329 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:28,435 epoch 34 - iter 1/17 - loss 2.16651177 - samples/sec: 310.24 - lr: 0.007812
2020-09-11 09:31:28,534 epoch 34 - iter 2/17 - loss 2.25732791 - samples/sec: 329.03 - lr: 0.007812
2020-09-11 09:31:28,648 epoch 34 - iter 3/17 - loss 2.21363282 - samples/sec: 284.57 - lr: 0.007812
2020-09-11 09:31:28,771 epoch 34 - iter 4/17 - loss 2.33480859 - samples/sec: 262.22 - lr: 0.007812
2020-09-11 09:31:28,888 epoch 34 - iter 5/17 - loss 2.34977865 - samples/sec: 275.83 - lr: 0.007812
2020-09-11 09:31:28,998 epoch 34 - iter 6/17 - loss 2.29187767 - samples/sec: 295.34 - lr: 0.007812
2020-09-11 09:31:29,107 epoch 34 - iter 7/17 - loss 2.28540604 - samples/sec: 296.13 - lr: 0.007812
2020-09-11 09:31:29,216 epoch 34 - iter 8/17 - loss 2.21293262 - samples/sec: 297.22 - lr: 0.007812
2020-09-11 09:31:29,324 epoch 34 - iter 9/17 - loss 2.13716773 - samples/sec: 300.87 - lr: 0.007812
2020-09-11 09:31:29,454 epoch 34 - iter 10/17 - loss 2.22819139 - samples/sec: 247.87 - lr: 0.007812
2020-09-11 09:31:29,559 epoch 34 - iter 11/17 - loss 2.28991833 - samples/sec: 309.13 - lr: 0.007812
2020-09-11 09:31:29,670 epoch 34 - iter 12/17 - loss 2.28256600 - samples/sec: 291.14 - lr: 0.007812
2020-09-11 09:31:29,783 epoch 34 - iter 13/17 - loss 2.22029331 - samples/sec: 285.83 - lr: 0.007812
2020-09-11 09:31:29,906 epoch 34 - iter 14/17 - loss 2.21900798 - samples/sec: 263.64 - lr: 0.007812
2020-09-11 09:31:30,017 epoch 34 - iter 15/17 - loss 2.19328976 - samples/sec: 292.50 - lr: 0.007812
2020-09-11 09:31:30,129 epoch 34 - iter 16/17 - loss 2.19865713 - samples/sec: 288.33 - lr: 0.007812
2020-09-11 09:31:30,207 epoch 34 - iter 17/17 - loss 2.23145440 - samples/sec: 416.13 - lr: 0.007812
2020-09-11 09:31:30,208 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:30,214 EPOCH 34 done: loss 2.2315 - lr 0.0078125
2020-09-11 09:31:30,292 DEV : loss 1.8520863056182861 - score 0.9039
2020-09-11 09:31:30,296 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:30,299 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:30,403 epoch 35 - iter 1/17 - loss 2.20372343 - samples/sec: 317.64 - lr: 0.007812
2020-09-11 09:31:30,512 epoch 35 - iter 2/17 - loss 2.14256120 - samples/sec: 297.31 - lr: 0.007812
2020-09-11 09:31:30,617 epoch 35 - iter 3/17 - loss 2.07499452 - samples/sec: 310.13 - lr: 0.007812
2020-09-11 09:31:30,733 epoch 35 - iter 4/17 - loss 2.12319884 - samples/sec: 276.98 - lr: 0.007812
2020-09-11 09:31:30,837 epoch 35 - iter 5/17 - loss 2.20568845 - samples/sec: 310.20 - lr: 0.007812
2020-09-11 09:31:30,949 epoch 35 - iter 6/17 - loss 2.16370652 - samples/sec: 291.01 - lr: 0.007812
2020-09-11 09:31:31,060 epoch 35 - iter 7/17 - loss 2.15263004 - samples/sec: 292.45 - lr: 0.007812
2020-09-11 09:31:31,167 epoch 35 - iter 8/17 - loss 2.19112043 - samples/sec: 300.66 - lr: 0.007812
2020-09-11 09:31:31,281 epoch 35 - iter 9/17 - loss 2.16499941 - samples/sec: 284.53 - lr: 0.007812
2020-09-11 09:31:31,403 epoch 35 - iter 10/17 - loss 2.11889089 - samples/sec: 264.39 - lr: 0.007812
2020-09-11 09:31:31,513 epoch 35 - iter 11/17 - loss 2.06410332 - samples/sec: 294.73 - lr: 0.007812
2020-09-11 09:31:31,622 epoch 35 - iter 12/17 - loss 2.05899282 - samples/sec: 295.55 - lr: 0.007812
2020-09-11 09:31:31,730 epoch 35 - iter 13/17 - loss 2.01894082 - samples/sec: 299.70 - lr: 0.007812
2020-09-11 09:31:31,843 epoch 35 - iter 14/17 - loss 2.08863953 - samples/sec: 285.49 - lr: 0.007812
2020-09-11 09:31:31,958 epoch 35 - iter 15/17 - loss 2.17515968 - samples/sec: 281.82 - lr: 0.007812
2020-09-11 09:31:32,063 epoch 35 - iter 16/17 - loss 2.16473049 - samples/sec: 307.60 - lr: 0.007812
2020-09-11 09:31:32,142 epoch 35 - iter 17/17 - loss 2.15948333 - samples/sec: 410.71 - lr: 0.007812
2020-09-11 09:31:32,143 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:32,148 EPOCH 35 done: loss 2.1595 - lr 0.0078125
2020-09-11 09:31:32,225 DEV : loss 1.8610153198242188 - score 0.9039
Epoch 35: reducing learning rate of group 0 to 3.9062e-03.
2020-09-11 09:31:32,228 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:32,230 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:32,346 epoch 36 - iter 1/17 - loss 2.43557644 - samples/sec: 285.81 - lr: 0.003906
2020-09-11 09:31:32,452 epoch 36 - iter 2/17 - loss 2.08234406 - samples/sec: 305.17 - lr: 0.003906
2020-09-11 09:31:32,556 epoch 36 - iter 3/17 - loss 2.13271157 - samples/sec: 308.61 - lr: 0.003906
2020-09-11 09:31:32,663 epoch 36 - iter 4/17 - loss 2.12066472 - samples/sec: 303.64 - lr: 0.003906
2020-09-11 09:31:32,775 epoch 36 - iter 5/17 - loss 2.19467897 - samples/sec: 288.35 - lr: 0.003906
2020-09-11 09:31:32,884 epoch 36 - iter 6/17 - loss 2.10642157 - samples/sec: 295.36 - lr: 0.003906
2020-09-11 09:31:32,996 epoch 36 - iter 7/17 - loss 2.09218320 - samples/sec: 295.59 - lr: 0.003906
2020-09-11 09:31:33,104 epoch 36 - iter 8/17 - loss 2.07874064 - samples/sec: 297.33 - lr: 0.003906
2020-09-11 09:31:33,210 epoch 36 - iter 9/17 - loss 2.07249725 - samples/sec: 304.95 - lr: 0.003906
2020-09-11 09:31:33,316 epoch 36 - iter 10/17 - loss 2.11360244 - samples/sec: 305.79 - lr: 0.003906
2020-09-11 09:31:33,427 epoch 36 - iter 11/17 - loss 2.14954514 - samples/sec: 292.79 - lr: 0.003906
2020-09-11 09:31:33,532 epoch 36 - iter 12/17 - loss 2.15247686 - samples/sec: 306.40 - lr: 0.003906
2020-09-11 09:31:33,628 epoch 36 - iter 13/17 - loss 2.17567102 - samples/sec: 336.53 - lr: 0.003906
2020-09-11 09:31:33,746 epoch 36 - iter 14/17 - loss 2.15970991 - samples/sec: 274.48 - lr: 0.003906
2020-09-11 09:31:33,877 epoch 36 - iter 15/17 - loss 2.17100244 - samples/sec: 246.64 - lr: 0.003906
2020-09-11 09:31:33,989 epoch 36 - iter 16/17 - loss 2.17317575 - samples/sec: 288.69 - lr: 0.003906
2020-09-11 09:31:34,068 epoch 36 - iter 17/17 - loss 2.16962830 - samples/sec: 414.82 - lr: 0.003906
2020-09-11 09:31:34,069 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:34,075 EPOCH 36 done: loss 2.1696 - lr 0.0039062
2020-09-11 09:31:34,154 DEV : loss 1.8612797260284424 - score 0.9039
2020-09-11 09:31:34,155 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:34,159 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:34,267 epoch 37 - iter 1/17 - loss 2.70274210 - samples/sec: 302.17 - lr: 0.003906
2020-09-11 09:31:34,376 epoch 37 - iter 2/17 - loss 2.30244517 - samples/sec: 295.43 - lr: 0.003906
2020-09-11 09:31:34,487 epoch 37 - iter 3/17 - loss 2.29809332 - samples/sec: 293.06 - lr: 0.003906
2020-09-11 09:31:34,595 epoch 37 - iter 4/17 - loss 2.27606213 - samples/sec: 299.59 - lr: 0.003906
2020-09-11 09:31:34,698 epoch 37 - iter 5/17 - loss 2.30987163 - samples/sec: 313.49 - lr: 0.003906
2020-09-11 09:31:34,802 epoch 37 - iter 6/17 - loss 2.23178925 - samples/sec: 309.89 - lr: 0.003906
2020-09-11 09:31:34,908 epoch 37 - iter 7/17 - loss 2.18345986 - samples/sec: 304.52 - lr: 0.003906
2020-09-11 09:31:35,011 epoch 37 - iter 8/17 - loss 2.11910208 - samples/sec: 320.77 - lr: 0.003906
2020-09-11 09:31:35,133 epoch 37 - iter 9/17 - loss 2.08216679 - samples/sec: 265.51 - lr: 0.003906
2020-09-11 09:31:35,235 epoch 37 - iter 10/17 - loss 2.14111589 - samples/sec: 316.56 - lr: 0.003906
2020-09-11 09:31:35,337 epoch 37 - iter 11/17 - loss 2.13171017 - samples/sec: 316.25 - lr: 0.003906
2020-09-11 09:31:35,447 epoch 37 - iter 12/17 - loss 2.11819906 - samples/sec: 294.52 - lr: 0.003906
2020-09-11 09:31:35,548 epoch 37 - iter 13/17 - loss 2.12408993 - samples/sec: 318.93 - lr: 0.003906
2020-09-11 09:31:35,658 epoch 37 - iter 14/17 - loss 2.14650181 - samples/sec: 292.30 - lr: 0.003906
2020-09-11 09:31:35,759 epoch 37 - iter 15/17 - loss 2.10822772 - samples/sec: 326.46 - lr: 0.003906
2020-09-11 09:31:35,873 epoch 37 - iter 16/17 - loss 2.09593584 - samples/sec: 282.42 - lr: 0.003906
2020-09-11 09:31:35,959 epoch 37 - iter 17/17 - loss 2.13863867 - samples/sec: 375.76 - lr: 0.003906
2020-09-11 09:31:35,961 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:35,963 EPOCH 37 done: loss 2.1386 - lr 0.0039062
2020-09-11 09:31:36,041 DEV : loss 1.862546443939209 - score 0.9039
2020-09-11 09:31:36,045 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:36,047 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:36,153 epoch 38 - iter 1/17 - loss 1.55653930 - samples/sec: 309.42 - lr: 0.003906
2020-09-11 09:31:36,262 epoch 38 - iter 2/17 - loss 1.83484852 - samples/sec: 295.54 - lr: 0.003906
2020-09-11 09:31:36,366 epoch 38 - iter 3/17 - loss 1.97699809 - samples/sec: 310.71 - lr: 0.003906
2020-09-11 09:31:36,478 epoch 38 - iter 4/17 - loss 1.97135368 - samples/sec: 290.51 - lr: 0.003906
2020-09-11 09:31:36,580 epoch 38 - iter 5/17 - loss 2.01099064 - samples/sec: 318.33 - lr: 0.003906
2020-09-11 09:31:36,685 epoch 38 - iter 6/17 - loss 1.91558721 - samples/sec: 306.78 - lr: 0.003906
2020-09-11 09:31:36,811 epoch 38 - iter 7/17 - loss 2.06156763 - samples/sec: 256.64 - lr: 0.003906
2020-09-11 09:31:36,913 epoch 38 - iter 8/17 - loss 2.07127489 - samples/sec: 318.30 - lr: 0.003906
2020-09-11 09:31:37,034 epoch 38 - iter 9/17 - loss 2.12594540 - samples/sec: 266.92 - lr: 0.003906
2020-09-11 09:31:37,138 epoch 38 - iter 10/17 - loss 2.16660880 - samples/sec: 313.26 - lr: 0.003906
2020-09-11 09:31:37,246 epoch 38 - iter 11/17 - loss 2.15714467 - samples/sec: 297.33 - lr: 0.003906
2020-09-11 09:31:37,350 epoch 38 - iter 12/17 - loss 2.16440808 - samples/sec: 314.15 - lr: 0.003906
2020-09-11 09:31:37,460 epoch 38 - iter 13/17 - loss 2.14800671 - samples/sec: 294.64 - lr: 0.003906
2020-09-11 09:31:37,583 epoch 38 - iter 14/17 - loss 2.16008878 - samples/sec: 262.87 - lr: 0.003906
2020-09-11 09:31:37,687 epoch 38 - iter 15/17 - loss 2.08712841 - samples/sec: 309.62 - lr: 0.003906
2020-09-11 09:31:37,788 epoch 38 - iter 16/17 - loss 2.09805407 - samples/sec: 320.24 - lr: 0.003906
2020-09-11 09:31:37,868 epoch 38 - iter 17/17 - loss 2.07644403 - samples/sec: 404.17 - lr: 0.003906
2020-09-11 09:31:37,869 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:37,873 EPOCH 38 done: loss 2.0764 - lr 0.0039062
2020-09-11 09:31:37,949 DEV : loss 1.869424819946289 - score 0.9039
2020-09-11 09:31:37,953 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:37,961 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:38,070 epoch 39 - iter 1/17 - loss 2.91720772 - samples/sec: 299.92 - lr: 0.003906
2020-09-11 09:31:38,174 epoch 39 - iter 2/17 - loss 2.45199174 - samples/sec: 313.26 - lr: 0.003906
2020-09-11 09:31:38,280 epoch 39 - iter 3/17 - loss 2.31430829 - samples/sec: 303.97 - lr: 0.003906
2020-09-11 09:31:38,389 epoch 39 - iter 4/17 - loss 2.37968537 - samples/sec: 296.63 - lr: 0.003906
2020-09-11 09:31:38,484 epoch 39 - iter 5/17 - loss 2.30176828 - samples/sec: 338.59 - lr: 0.003906
2020-09-11 09:31:38,590 epoch 39 - iter 6/17 - loss 2.28818593 - samples/sec: 308.01 - lr: 0.003906
2020-09-11 09:31:38,694 epoch 39 - iter 7/17 - loss 2.20957887 - samples/sec: 315.08 - lr: 0.003906
2020-09-11 09:31:38,796 epoch 39 - iter 8/17 - loss 2.15541568 - samples/sec: 316.55 - lr: 0.003906
2020-09-11 09:31:38,906 epoch 39 - iter 9/17 - loss 2.05849785 - samples/sec: 295.51 - lr: 0.003906
2020-09-11 09:31:39,026 epoch 39 - iter 10/17 - loss 2.14265947 - samples/sec: 268.94 - lr: 0.003906
2020-09-11 09:31:39,131 epoch 39 - iter 11/17 - loss 2.14655584 - samples/sec: 307.74 - lr: 0.003906
2020-09-11 09:31:39,229 epoch 39 - iter 12/17 - loss 2.12833981 - samples/sec: 328.42 - lr: 0.003906
2020-09-11 09:31:39,350 epoch 39 - iter 13/17 - loss 2.23054640 - samples/sec: 268.39 - lr: 0.003906
2020-09-11 09:31:39,455 epoch 39 - iter 14/17 - loss 2.18998856 - samples/sec: 308.36 - lr: 0.003906
2020-09-11 09:31:39,549 epoch 39 - iter 15/17 - loss 2.13317854 - samples/sec: 344.80 - lr: 0.003906
2020-09-11 09:31:39,655 epoch 39 - iter 16/17 - loss 2.14621764 - samples/sec: 305.66 - lr: 0.003906
2020-09-11 09:31:39,735 epoch 39 - iter 17/17 - loss 2.10736282 - samples/sec: 400.25 - lr: 0.003906
2020-09-11 09:31:39,736 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:39,742 EPOCH 39 done: loss 2.1074 - lr 0.0039062
2020-09-11 09:31:39,838 DEV : loss 1.8706583976745605 - score 0.9039
Epoch 39: reducing learning rate of group 0 to 1.9531e-03.
2020-09-11 09:31:39,842 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:39,844 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:39,964 epoch 40 - iter 1/17 - loss 2.40858364 - samples/sec: 286.84 - lr: 0.001953
2020-09-11 09:31:40,096 epoch 40 - iter 2/17 - loss 2.66101336 - samples/sec: 245.63 - lr: 0.001953
2020-09-11 09:31:40,202 epoch 40 - iter 3/17 - loss 2.68021313 - samples/sec: 303.80 - lr: 0.001953
2020-09-11 09:31:40,312 epoch 40 - iter 4/17 - loss 2.57772970 - samples/sec: 293.86 - lr: 0.001953
2020-09-11 09:31:40,422 epoch 40 - iter 5/17 - loss 2.45276420 - samples/sec: 294.24 - lr: 0.001953
2020-09-11 09:31:40,532 epoch 40 - iter 6/17 - loss 2.51650308 - samples/sec: 294.31 - lr: 0.001953
2020-09-11 09:31:40,635 epoch 40 - iter 7/17 - loss 2.36470851 - samples/sec: 315.08 - lr: 0.001953
2020-09-11 09:31:40,742 epoch 40 - iter 8/17 - loss 2.28064451 - samples/sec: 303.42 - lr: 0.001953
2020-09-11 09:31:40,846 epoch 40 - iter 9/17 - loss 2.25968324 - samples/sec: 309.69 - lr: 0.001953
2020-09-11 09:31:40,961 epoch 40 - iter 10/17 - loss 2.25215931 - samples/sec: 281.35 - lr: 0.001953
2020-09-11 09:31:41,076 epoch 40 - iter 11/17 - loss 2.29170622 - samples/sec: 282.39 - lr: 0.001953
2020-09-11 09:31:41,183 epoch 40 - iter 12/17 - loss 2.24143372 - samples/sec: 303.11 - lr: 0.001953
2020-09-11 09:31:41,288 epoch 40 - iter 13/17 - loss 2.27747974 - samples/sec: 308.06 - lr: 0.001953
2020-09-11 09:31:41,393 epoch 40 - iter 14/17 - loss 2.22446014 - samples/sec: 307.05 - lr: 0.001953
2020-09-11 09:31:41,507 epoch 40 - iter 15/17 - loss 2.23860873 - samples/sec: 282.55 - lr: 0.001953
2020-09-11 09:31:41,628 epoch 40 - iter 16/17 - loss 2.21095503 - samples/sec: 266.93 - lr: 0.001953
2020-09-11 09:31:41,708 epoch 40 - iter 17/17 - loss 2.20091358 - samples/sec: 409.14 - lr: 0.001953
2020-09-11 09:31:41,712 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:41,713 EPOCH 40 done: loss 2.2009 - lr 0.0019531
2020-09-11 09:31:41,803 DEV : loss 1.8684940338134766 - score 0.9039
2020-09-11 09:31:41,809 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:41,812 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:41,929 epoch 41 - iter 1/17 - loss 1.65641534 - samples/sec: 281.95 - lr: 0.001953
2020-09-11 09:31:42,047 epoch 41 - iter 2/17 - loss 1.98350984 - samples/sec: 274.74 - lr: 0.001953
2020-09-11 09:31:42,156 epoch 41 - iter 3/17 - loss 1.93206290 - samples/sec: 298.68 - lr: 0.001953
2020-09-11 09:31:42,268 epoch 41 - iter 4/17 - loss 1.83420372 - samples/sec: 287.80 - lr: 0.001953
2020-09-11 09:31:42,374 epoch 41 - iter 5/17 - loss 1.74692581 - samples/sec: 304.46 - lr: 0.001953
2020-09-11 09:31:42,488 epoch 41 - iter 6/17 - loss 1.89210922 - samples/sec: 287.42 - lr: 0.001953
2020-09-11 09:31:42,600 epoch 41 - iter 7/17 - loss 1.92536325 - samples/sec: 289.60 - lr: 0.001953
2020-09-11 09:31:42,717 epoch 41 - iter 8/17 - loss 1.87745994 - samples/sec: 276.82 - lr: 0.001953
2020-09-11 09:31:42,834 epoch 41 - iter 9/17 - loss 1.95671293 - samples/sec: 277.22 - lr: 0.001953
2020-09-11 09:31:42,969 epoch 41 - iter 10/17 - loss 2.00737596 - samples/sec: 238.75 - lr: 0.001953
2020-09-11 09:31:43,081 epoch 41 - iter 11/17 - loss 2.06435234 - samples/sec: 287.18 - lr: 0.001953
2020-09-11 09:31:43,191 epoch 41 - iter 12/17 - loss 2.05423961 - samples/sec: 294.45 - lr: 0.001953
2020-09-11 09:31:43,295 epoch 41 - iter 13/17 - loss 2.07148066 - samples/sec: 311.37 - lr: 0.001953
2020-09-11 09:31:43,403 epoch 41 - iter 14/17 - loss 2.10708749 - samples/sec: 300.12 - lr: 0.001953
2020-09-11 09:31:43,509 epoch 41 - iter 15/17 - loss 2.10328631 - samples/sec: 303.79 - lr: 0.001953
2020-09-11 09:31:43,631 epoch 41 - iter 16/17 - loss 2.10379887 - samples/sec: 264.22 - lr: 0.001953
2020-09-11 09:31:43,705 epoch 41 - iter 17/17 - loss 2.13205387 - samples/sec: 444.99 - lr: 0.001953
2020-09-11 09:31:43,706 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:43,711 EPOCH 41 done: loss 2.1321 - lr 0.0019531
2020-09-11 09:31:43,795 DEV : loss 1.8697052001953125 - score 0.9039
2020-09-11 09:31:43,798 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:43,800 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:43,909 epoch 42 - iter 1/17 - loss 1.95205081 - samples/sec: 303.71 - lr: 0.001953
2020-09-11 09:31:44,025 epoch 42 - iter 2/17 - loss 1.77231854 - samples/sec: 278.23 - lr: 0.001953
2020-09-11 09:31:44,141 epoch 42 - iter 3/17 - loss 1.76212402 - samples/sec: 278.12 - lr: 0.001953
2020-09-11 09:31:44,251 epoch 42 - iter 4/17 - loss 1.68454868 - samples/sec: 296.14 - lr: 0.001953
2020-09-11 09:31:44,358 epoch 42 - iter 5/17 - loss 1.72046323 - samples/sec: 300.28 - lr: 0.001953
2020-09-11 09:31:44,466 epoch 42 - iter 6/17 - loss 1.82062662 - samples/sec: 300.04 - lr: 0.001953
2020-09-11 09:31:44,577 epoch 42 - iter 7/17 - loss 1.82287388 - samples/sec: 290.32 - lr: 0.001953
2020-09-11 09:31:44,680 epoch 42 - iter 8/17 - loss 1.80281527 - samples/sec: 312.72 - lr: 0.001953
2020-09-11 09:31:44,786 epoch 42 - iter 9/17 - loss 1.73538368 - samples/sec: 307.32 - lr: 0.001953
2020-09-11 09:31:44,889 epoch 42 - iter 10/17 - loss 1.86518863 - samples/sec: 312.29 - lr: 0.001953
2020-09-11 09:31:45,024 epoch 42 - iter 11/17 - loss 1.89539980 - samples/sec: 238.57 - lr: 0.001953
2020-09-11 09:31:45,142 epoch 42 - iter 12/17 - loss 1.88407112 - samples/sec: 273.58 - lr: 0.001953
2020-09-11 09:31:45,252 epoch 42 - iter 13/17 - loss 1.93909807 - samples/sec: 296.27 - lr: 0.001953
2020-09-11 09:31:45,366 epoch 42 - iter 14/17 - loss 1.96448104 - samples/sec: 283.21 - lr: 0.001953
2020-09-11 09:31:45,478 epoch 42 - iter 15/17 - loss 2.04754047 - samples/sec: 288.62 - lr: 0.001953
2020-09-11 09:31:45,588 epoch 42 - iter 16/17 - loss 2.09948217 - samples/sec: 295.08 - lr: 0.001953
2020-09-11 09:31:45,660 epoch 42 - iter 17/17 - loss 2.06929702 - samples/sec: 447.93 - lr: 0.001953
2020-09-11 09:31:45,661 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:45,666 EPOCH 42 done: loss 2.0693 - lr 0.0019531
2020-09-11 09:31:45,751 DEV : loss 1.8711341619491577 - score 0.9039
2020-09-11 09:31:45,753 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:45,754 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:45,875 epoch 43 - iter 1/17 - loss 1.59613466 - samples/sec: 272.75 - lr: 0.001953
2020-09-11 09:31:45,986 epoch 43 - iter 2/17 - loss 1.89992344 - samples/sec: 292.91 - lr: 0.001953
2020-09-11 09:31:46,111 epoch 43 - iter 3/17 - loss 2.15681545 - samples/sec: 257.22 - lr: 0.001953
2020-09-11 09:31:46,221 epoch 43 - iter 4/17 - loss 1.94478324 - samples/sec: 296.32 - lr: 0.001953
2020-09-11 09:31:46,327 epoch 43 - iter 5/17 - loss 2.02235644 - samples/sec: 306.35 - lr: 0.001953
2020-09-11 09:31:46,436 epoch 43 - iter 6/17 - loss 2.06755092 - samples/sec: 295.39 - lr: 0.001953
2020-09-11 09:31:46,533 epoch 43 - iter 7/17 - loss 2.09775879 - samples/sec: 332.81 - lr: 0.001953
2020-09-11 09:31:46,635 epoch 43 - iter 8/17 - loss 2.04314528 - samples/sec: 316.79 - lr: 0.001953
2020-09-11 09:31:46,767 epoch 43 - iter 9/17 - loss 2.04059197 - samples/sec: 244.50 - lr: 0.001953
2020-09-11 09:31:46,887 epoch 43 - iter 10/17 - loss 2.13431033 - samples/sec: 269.87 - lr: 0.001953
2020-09-11 09:31:46,997 epoch 43 - iter 11/17 - loss 2.11249408 - samples/sec: 295.32 - lr: 0.001953
2020-09-11 09:31:47,101 epoch 43 - iter 12/17 - loss 2.18600748 - samples/sec: 308.72 - lr: 0.001953
2020-09-11 09:31:47,213 epoch 43 - iter 13/17 - loss 2.19971956 - samples/sec: 290.51 - lr: 0.001953
2020-09-11 09:31:47,323 epoch 43 - iter 14/17 - loss 2.11635588 - samples/sec: 295.03 - lr: 0.001953
2020-09-11 09:31:47,449 epoch 43 - iter 15/17 - loss 2.15256975 - samples/sec: 255.52 - lr: 0.001953
2020-09-11 09:31:47,560 epoch 43 - iter 16/17 - loss 2.12364431 - samples/sec: 290.04 - lr: 0.001953
2020-09-11 09:31:47,632 epoch 43 - iter 17/17 - loss 2.13725646 - samples/sec: 467.01 - lr: 0.001953
2020-09-11 09:31:47,633 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:47,638 EPOCH 43 done: loss 2.1373 - lr 0.0019531
2020-09-11 09:31:47,716 DEV : loss 1.871195912361145 - score 0.9039
Epoch 43: reducing learning rate of group 0 to 9.7656e-04.
2020-09-11 09:31:47,720 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:47,721 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:47,831 epoch 44 - iter 1/17 - loss 1.57494926 - samples/sec: 297.58 - lr: 0.000977
2020-09-11 09:31:47,955 epoch 44 - iter 2/17 - loss 1.70237011 - samples/sec: 265.10 - lr: 0.000977
2020-09-11 09:31:48,086 epoch 44 - iter 3/17 - loss 1.62696763 - samples/sec: 247.22 - lr: 0.000977
2020-09-11 09:31:48,203 epoch 44 - iter 4/17 - loss 1.73749629 - samples/sec: 275.45 - lr: 0.000977
2020-09-11 09:31:48,311 epoch 44 - iter 5/17 - loss 1.85222251 - samples/sec: 300.34 - lr: 0.000977
2020-09-11 09:31:48,425 epoch 44 - iter 6/17 - loss 1.92494589 - samples/sec: 282.86 - lr: 0.000977
2020-09-11 09:31:48,539 epoch 44 - iter 7/17 - loss 2.00947160 - samples/sec: 283.22 - lr: 0.000977
2020-09-11 09:31:48,656 epoch 44 - iter 8/17 - loss 2.03440548 - samples/sec: 276.26 - lr: 0.000977
2020-09-11 09:31:48,761 epoch 44 - iter 9/17 - loss 2.03593189 - samples/sec: 309.42 - lr: 0.000977
2020-09-11 09:31:48,867 epoch 44 - iter 10/17 - loss 1.98007635 - samples/sec: 304.99 - lr: 0.000977
2020-09-11 09:31:48,977 epoch 44 - iter 11/17 - loss 1.96373889 - samples/sec: 295.94 - lr: 0.000977
2020-09-11 09:31:49,095 epoch 44 - iter 12/17 - loss 2.03976662 - samples/sec: 276.61 - lr: 0.000977
2020-09-11 09:31:49,212 epoch 44 - iter 13/17 - loss 2.07122222 - samples/sec: 275.88 - lr: 0.000977
2020-09-11 09:31:49,318 epoch 44 - iter 14/17 - loss 2.06604474 - samples/sec: 306.71 - lr: 0.000977
2020-09-11 09:31:49,442 epoch 44 - iter 15/17 - loss 2.10574220 - samples/sec: 259.62 - lr: 0.000977
2020-09-11 09:31:49,560 epoch 44 - iter 16/17 - loss 2.10223112 - samples/sec: 273.95 - lr: 0.000977
2020-09-11 09:31:49,636 epoch 44 - iter 17/17 - loss 2.08419435 - samples/sec: 427.19 - lr: 0.000977
2020-09-11 09:31:49,637 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:49,641 EPOCH 44 done: loss 2.0842 - lr 0.0009766
2020-09-11 09:31:49,724 DEV : loss 1.8707306385040283 - score 0.9039
2020-09-11 09:31:49,726 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:49,730 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:49,838 epoch 45 - iter 1/17 - loss 1.89361787 - samples/sec: 305.45 - lr: 0.000977
2020-09-11 09:31:49,947 epoch 45 - iter 2/17 - loss 2.21607864 - samples/sec: 299.69 - lr: 0.000977
2020-09-11 09:31:50,049 epoch 45 - iter 3/17 - loss 2.01096229 - samples/sec: 315.51 - lr: 0.000977
2020-09-11 09:31:50,170 epoch 45 - iter 4/17 - loss 1.97172287 - samples/sec: 266.99 - lr: 0.000977
2020-09-11 09:31:50,277 epoch 45 - iter 5/17 - loss 1.93318305 - samples/sec: 301.61 - lr: 0.000977
2020-09-11 09:31:50,372 epoch 45 - iter 6/17 - loss 1.82265991 - samples/sec: 340.79 - lr: 0.000977
2020-09-11 09:31:50,475 epoch 45 - iter 7/17 - loss 1.90316897 - samples/sec: 313.52 - lr: 0.000977
2020-09-11 09:31:50,582 epoch 45 - iter 8/17 - loss 1.94942440 - samples/sec: 302.36 - lr: 0.000977
2020-09-11 09:31:50,684 epoch 45 - iter 9/17 - loss 1.91201617 - samples/sec: 318.28 - lr: 0.000977
2020-09-11 09:31:50,788 epoch 45 - iter 10/17 - loss 1.95709399 - samples/sec: 308.72 - lr: 0.000977
2020-09-11 09:31:50,917 epoch 45 - iter 11/17 - loss 2.03674131 - samples/sec: 251.01 - lr: 0.000977
2020-09-11 09:31:51,031 epoch 45 - iter 12/17 - loss 2.09745031 - samples/sec: 283.16 - lr: 0.000977
2020-09-11 09:31:51,144 epoch 45 - iter 13/17 - loss 2.10905980 - samples/sec: 288.12 - lr: 0.000977
2020-09-11 09:31:51,254 epoch 45 - iter 14/17 - loss 2.12343171 - samples/sec: 293.42 - lr: 0.000977
2020-09-11 09:31:51,371 epoch 45 - iter 15/17 - loss 2.10053473 - samples/sec: 276.05 - lr: 0.000977
2020-09-11 09:31:51,487 epoch 45 - iter 16/17 - loss 2.09765709 - samples/sec: 278.21 - lr: 0.000977
2020-09-11 09:31:51,571 epoch 45 - iter 17/17 - loss 2.11037770 - samples/sec: 387.99 - lr: 0.000977
2020-09-11 09:31:51,572 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:51,577 EPOCH 45 done: loss 2.1104 - lr 0.0009766
2020-09-11 09:31:51,654 DEV : loss 1.8710083961486816 - score 0.9039
2020-09-11 09:31:51,656 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:51,659 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:51,769 epoch 46 - iter 1/17 - loss 2.29029131 - samples/sec: 297.72 - lr: 0.000977
2020-09-11 09:31:51,878 epoch 46 - iter 2/17 - loss 2.57501078 - samples/sec: 295.45 - lr: 0.000977
2020-09-11 09:31:51,995 epoch 46 - iter 3/17 - loss 2.33628249 - samples/sec: 277.47 - lr: 0.000977
2020-09-11 09:31:52,100 epoch 46 - iter 4/17 - loss 2.20042735 - samples/sec: 308.27 - lr: 0.000977
2020-09-11 09:31:52,228 epoch 46 - iter 5/17 - loss 2.15306783 - samples/sec: 252.65 - lr: 0.000977
2020-09-11 09:31:52,337 epoch 46 - iter 6/17 - loss 2.23327601 - samples/sec: 298.16 - lr: 0.000977
2020-09-11 09:31:52,455 epoch 46 - iter 7/17 - loss 2.17818601 - samples/sec: 273.14 - lr: 0.000977
2020-09-11 09:31:52,580 epoch 46 - iter 8/17 - loss 2.29868346 - samples/sec: 262.24 - lr: 0.000977
2020-09-11 09:31:52,692 epoch 46 - iter 9/17 - loss 2.24365924 - samples/sec: 286.62 - lr: 0.000977
2020-09-11 09:31:52,793 epoch 46 - iter 10/17 - loss 2.13943095 - samples/sec: 321.16 - lr: 0.000977
2020-09-11 09:31:52,909 epoch 46 - iter 11/17 - loss 2.22218587 - samples/sec: 278.50 - lr: 0.000977
2020-09-11 09:31:53,011 epoch 46 - iter 12/17 - loss 2.17799040 - samples/sec: 320.62 - lr: 0.000977
2020-09-11 09:31:53,122 epoch 46 - iter 13/17 - loss 2.13923915 - samples/sec: 291.37 - lr: 0.000977
2020-09-11 09:31:53,244 epoch 46 - iter 14/17 - loss 2.13772869 - samples/sec: 263.87 - lr: 0.000977
2020-09-11 09:31:53,350 epoch 46 - iter 15/17 - loss 2.10194511 - samples/sec: 308.98 - lr: 0.000977
2020-09-11 09:31:53,474 epoch 46 - iter 16/17 - loss 2.12772428 - samples/sec: 261.12 - lr: 0.000977
2020-09-11 09:31:53,558 epoch 46 - iter 17/17 - loss 2.14000834 - samples/sec: 387.06 - lr: 0.000977
2020-09-11 09:31:53,559 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:53,564 EPOCH 46 done: loss 2.1400 - lr 0.0009766
2020-09-11 09:31:53,642 DEV : loss 1.8702256679534912 - score 0.9039
2020-09-11 09:31:53,647 BAD EPOCHS (no improvement): 3
2020-09-11 09:31:53,648 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:53,759 epoch 47 - iter 1/17 - loss 2.40958548 - samples/sec: 304.72 - lr: 0.000977
2020-09-11 09:31:53,893 epoch 47 - iter 2/17 - loss 2.19522685 - samples/sec: 242.33 - lr: 0.000977
2020-09-11 09:31:54,004 epoch 47 - iter 3/17 - loss 2.18233558 - samples/sec: 292.21 - lr: 0.000977
2020-09-11 09:31:54,117 epoch 47 - iter 4/17 - loss 2.08626416 - samples/sec: 287.13 - lr: 0.000977
2020-09-11 09:31:54,243 epoch 47 - iter 5/17 - loss 2.13577464 - samples/sec: 257.57 - lr: 0.000977
2020-09-11 09:31:54,352 epoch 47 - iter 6/17 - loss 2.09897268 - samples/sec: 294.61 - lr: 0.000977
2020-09-11 09:31:54,457 epoch 47 - iter 7/17 - loss 2.25570938 - samples/sec: 309.49 - lr: 0.000977
2020-09-11 09:31:54,566 epoch 47 - iter 8/17 - loss 2.36531955 - samples/sec: 299.93 - lr: 0.000977
2020-09-11 09:31:54,692 epoch 47 - iter 9/17 - loss 2.39397939 - samples/sec: 256.87 - lr: 0.000977
2020-09-11 09:31:54,801 epoch 47 - iter 10/17 - loss 2.31865383 - samples/sec: 295.82 - lr: 0.000977
2020-09-11 09:31:54,904 epoch 47 - iter 11/17 - loss 2.23503270 - samples/sec: 314.43 - lr: 0.000977
2020-09-11 09:31:55,006 epoch 47 - iter 12/17 - loss 2.19364679 - samples/sec: 318.50 - lr: 0.000977
2020-09-11 09:31:55,107 epoch 47 - iter 13/17 - loss 2.16499858 - samples/sec: 319.93 - lr: 0.000977
2020-09-11 09:31:55,247 epoch 47 - iter 14/17 - loss 2.17180823 - samples/sec: 229.81 - lr: 0.000977
2020-09-11 09:31:55,356 epoch 47 - iter 15/17 - loss 2.19207067 - samples/sec: 298.37 - lr: 0.000977
2020-09-11 09:31:55,467 epoch 47 - iter 16/17 - loss 2.19204720 - samples/sec: 292.88 - lr: 0.000977
2020-09-11 09:31:55,540 epoch 47 - iter 17/17 - loss 2.16333261 - samples/sec: 439.67 - lr: 0.000977
2020-09-11 09:31:55,542 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:55,546 EPOCH 47 done: loss 2.1633 - lr 0.0009766
2020-09-11 09:31:55,621 DEV : loss 1.8696794509887695 - score 0.9039
Epoch 47: reducing learning rate of group 0 to 4.8828e-04.
2020-09-11 09:31:55,625 BAD EPOCHS (no improvement): 4
2020-09-11 09:31:55,627 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:55,732 epoch 48 - iter 1/17 - loss 2.13647890 - samples/sec: 312.68 - lr: 0.000488
2020-09-11 09:31:55,841 epoch 48 - iter 2/17 - loss 2.60982764 - samples/sec: 295.73 - lr: 0.000488
2020-09-11 09:31:55,955 epoch 48 - iter 3/17 - loss 2.60576868 - samples/sec: 284.55 - lr: 0.000488
2020-09-11 09:31:56,063 epoch 48 - iter 4/17 - loss 2.81221414 - samples/sec: 299.76 - lr: 0.000488
2020-09-11 09:31:56,165 epoch 48 - iter 5/17 - loss 2.56672552 - samples/sec: 320.36 - lr: 0.000488
2020-09-11 09:31:56,284 epoch 48 - iter 6/17 - loss 2.45763489 - samples/sec: 271.14 - lr: 0.000488
2020-09-11 09:31:56,398 epoch 48 - iter 7/17 - loss 2.42420312 - samples/sec: 284.66 - lr: 0.000488
2020-09-11 09:31:56,512 epoch 48 - iter 8/17 - loss 2.40458560 - samples/sec: 282.20 - lr: 0.000488
2020-09-11 09:31:56,637 epoch 48 - iter 9/17 - loss 2.34310838 - samples/sec: 258.63 - lr: 0.000488
2020-09-11 09:31:56,745 epoch 48 - iter 10/17 - loss 2.32288555 - samples/sec: 302.83 - lr: 0.000488
2020-09-11 09:31:56,859 epoch 48 - iter 11/17 - loss 2.23890460 - samples/sec: 282.30 - lr: 0.000488
2020-09-11 09:31:56,972 epoch 48 - iter 12/17 - loss 2.23518967 - samples/sec: 287.37 - lr: 0.000488
2020-09-11 09:31:57,080 epoch 48 - iter 13/17 - loss 2.26823274 - samples/sec: 299.58 - lr: 0.000488
2020-09-11 09:31:57,191 epoch 48 - iter 14/17 - loss 2.26889933 - samples/sec: 296.74 - lr: 0.000488
2020-09-11 09:31:57,307 epoch 48 - iter 15/17 - loss 2.23389218 - samples/sec: 277.80 - lr: 0.000488
2020-09-11 09:31:57,419 epoch 48 - iter 16/17 - loss 2.21442220 - samples/sec: 288.86 - lr: 0.000488
2020-09-11 09:31:57,489 epoch 48 - iter 17/17 - loss 2.15259550 - samples/sec: 463.29 - lr: 0.000488
2020-09-11 09:31:57,490 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:57,495 EPOCH 48 done: loss 2.1526 - lr 0.0004883
2020-09-11 09:31:57,572 DEV : loss 1.869992733001709 - score 0.9039
2020-09-11 09:31:57,575 BAD EPOCHS (no improvement): 1
2020-09-11 09:31:57,577 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:57,689 epoch 49 - iter 1/17 - loss 3.32574272 - samples/sec: 290.88 - lr: 0.000488
2020-09-11 09:31:57,798 epoch 49 - iter 2/17 - loss 3.23043871 - samples/sec: 298.50 - lr: 0.000488
2020-09-11 09:31:57,909 epoch 49 - iter 3/17 - loss 2.52508895 - samples/sec: 293.91 - lr: 0.000488
2020-09-11 09:31:58,014 epoch 49 - iter 4/17 - loss 2.27688953 - samples/sec: 307.34 - lr: 0.000488
2020-09-11 09:31:58,124 epoch 49 - iter 5/17 - loss 2.20293577 - samples/sec: 296.12 - lr: 0.000488
2020-09-11 09:31:58,225 epoch 49 - iter 6/17 - loss 2.22645797 - samples/sec: 318.18 - lr: 0.000488
2020-09-11 09:31:58,343 epoch 49 - iter 7/17 - loss 2.18002132 - samples/sec: 274.32 - lr: 0.000488
2020-09-11 09:31:58,454 epoch 49 - iter 8/17 - loss 2.08970803 - samples/sec: 290.59 - lr: 0.000488
2020-09-11 09:31:58,579 epoch 49 - iter 9/17 - loss 2.12008651 - samples/sec: 258.67 - lr: 0.000488
2020-09-11 09:31:58,698 epoch 49 - iter 10/17 - loss 2.12727408 - samples/sec: 272.61 - lr: 0.000488
2020-09-11 09:31:58,800 epoch 49 - iter 11/17 - loss 2.08404921 - samples/sec: 314.90 - lr: 0.000488
2020-09-11 09:31:58,913 epoch 49 - iter 12/17 - loss 2.07866135 - samples/sec: 288.02 - lr: 0.000488
2020-09-11 09:31:59,034 epoch 49 - iter 13/17 - loss 2.09018061 - samples/sec: 266.84 - lr: 0.000488
2020-09-11 09:31:59,152 epoch 49 - iter 14/17 - loss 2.12898230 - samples/sec: 275.69 - lr: 0.000488
2020-09-11 09:31:59,277 epoch 49 - iter 15/17 - loss 2.11229343 - samples/sec: 257.11 - lr: 0.000488
2020-09-11 09:31:59,413 epoch 49 - iter 16/17 - loss 2.14525196 - samples/sec: 237.68 - lr: 0.000488
2020-09-11 09:31:59,502 epoch 49 - iter 17/17 - loss 2.16026076 - samples/sec: 366.15 - lr: 0.000488
2020-09-11 09:31:59,503 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:59,509 EPOCH 49 done: loss 2.1603 - lr 0.0004883
2020-09-11 09:31:59,602 DEV : loss 1.8705830574035645 - score 0.9039
2020-09-11 09:31:59,605 BAD EPOCHS (no improvement): 2
2020-09-11 09:31:59,606 ----------------------------------------------------------------------------------------------------
2020-09-11 09:31:59,729 epoch 50 - iter 1/17 - loss 1.70025277 - samples/sec: 265.16 - lr: 0.000488
2020-09-11 09:31:59,843 epoch 50 - iter 2/17 - loss 1.75726217 - samples/sec: 285.34 - lr: 0.000488
2020-09-11 09:31:59,947 epoch 50 - iter 3/17 - loss 1.81887865 - samples/sec: 311.87 - lr: 0.000488
2020-09-11 09:32:00,055 epoch 50 - iter 4/17 - loss 1.73993617 - samples/sec: 299.32 - lr: 0.000488
2020-09-11 09:32:00,164 epoch 50 - iter 5/17 - loss 1.70279319 - samples/sec: 295.46 - lr: 0.000488
2020-09-11 09:32:00,290 epoch 50 - iter 6/17 - loss 1.82746798 - samples/sec: 257.35 - lr: 0.000488
2020-09-11 09:32:00,417 epoch 50 - iter 7/17 - loss 1.90154033 - samples/sec: 252.97 - lr: 0.000488
2020-09-11 09:32:00,537 epoch 50 - iter 8/17 - loss 1.92441855 - samples/sec: 270.18 - lr: 0.000488
2020-09-11 09:32:00,650 epoch 50 - iter 9/17 - loss 1.97294697 - samples/sec: 291.55 - lr: 0.000488
2020-09-11 09:32:00,759 epoch 50 - iter 10/17 - loss 1.98562843 - samples/sec: 297.68 - lr: 0.000488
2020-09-11 09:32:00,871 epoch 50 - iter 11/17 - loss 2.00348701 - samples/sec: 287.94 - lr: 0.000488
2020-09-11 09:32:00,991 epoch 50 - iter 12/17 - loss 2.05574360 - samples/sec: 270.36 - lr: 0.000488
2020-09-11 09:32:01,112 epoch 50 - iter 13/17 - loss 2.10646278 - samples/sec: 273.29 - lr: 0.000488
2020-09-11 09:32:01,222 epoch 50 - iter 14/17 - loss 2.14258188 - samples/sec: 297.55 - lr: 0.000488
2020-09-11 09:32:01,338 epoch 50 - iter 15/17 - loss 2.10952766 - samples/sec: 280.93 - lr: 0.000488
2020-09-11 09:32:01,454 epoch 50 - iter 16/17 - loss 2.18073478 - samples/sec: 282.13 - lr: 0.000488
2020-09-11 09:32:01,544 epoch 50 - iter 17/17 - loss 2.18814912 - samples/sec: 361.88 - lr: 0.000488
2020-09-11 09:32:01,546 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:01,555 EPOCH 50 done: loss 2.1881 - lr 0.0004883
2020-09-11 09:32:01,639 DEV : loss 1.8708758354187012 - score 0.9039
2020-09-11 09:32:01,643 BAD EPOCHS (no improvement): 3
2020-09-11 09:32:01,645 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:01,754 epoch 51 - iter 1/17 - loss 2.37925291 - samples/sec: 299.85 - lr: 0.000488
2020-09-11 09:32:01,862 epoch 51 - iter 2/17 - loss 2.31285119 - samples/sec: 299.21 - lr: 0.000488
2020-09-11 09:32:01,983 epoch 51 - iter 3/17 - loss 2.07815230 - samples/sec: 267.30 - lr: 0.000488
2020-09-11 09:32:02,098 epoch 51 - iter 4/17 - loss 2.12176999 - samples/sec: 280.35 - lr: 0.000488
2020-09-11 09:32:02,209 epoch 51 - iter 5/17 - loss 2.13544719 - samples/sec: 296.11 - lr: 0.000488
2020-09-11 09:32:02,315 epoch 51 - iter 6/17 - loss 2.15400670 - samples/sec: 302.91 - lr: 0.000488
2020-09-11 09:32:02,435 epoch 51 - iter 7/17 - loss 2.15878560 - samples/sec: 271.01 - lr: 0.000488
2020-09-11 09:32:02,546 epoch 51 - iter 8/17 - loss 2.10743538 - samples/sec: 292.75 - lr: 0.000488
2020-09-11 09:32:02,652 epoch 51 - iter 9/17 - loss 2.14642469 - samples/sec: 303.95 - lr: 0.000488
2020-09-11 09:32:02,768 epoch 51 - iter 10/17 - loss 2.11396281 - samples/sec: 280.38 - lr: 0.000488
2020-09-11 09:32:02,896 epoch 51 - iter 11/17 - loss 2.13433785 - samples/sec: 252.34 - lr: 0.000488
2020-09-11 09:32:03,009 epoch 51 - iter 12/17 - loss 2.14074990 - samples/sec: 287.48 - lr: 0.000488
2020-09-11 09:32:03,133 epoch 51 - iter 13/17 - loss 2.16428389 - samples/sec: 266.61 - lr: 0.000488
2020-09-11 09:32:03,242 epoch 51 - iter 14/17 - loss 2.17041359 - samples/sec: 297.92 - lr: 0.000488
2020-09-11 09:32:03,348 epoch 51 - iter 15/17 - loss 2.11546025 - samples/sec: 305.25 - lr: 0.000488
2020-09-11 09:32:03,465 epoch 51 - iter 16/17 - loss 2.11760668 - samples/sec: 279.29 - lr: 0.000488
2020-09-11 09:32:03,532 epoch 51 - iter 17/17 - loss 2.15209274 - samples/sec: 491.81 - lr: 0.000488
2020-09-11 09:32:03,533 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:03,538 EPOCH 51 done: loss 2.1521 - lr 0.0004883
2020-09-11 09:32:03,616 DEV : loss 1.8709588050842285 - score 0.9039
Epoch 51: reducing learning rate of group 0 to 2.4414e-04.
2020-09-11 09:32:03,621 BAD EPOCHS (no improvement): 4
2020-09-11 09:32:03,623 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:03,734 epoch 52 - iter 1/17 - loss 1.42433143 - samples/sec: 293.13 - lr: 0.000244
2020-09-11 09:32:03,864 epoch 52 - iter 2/17 - loss 1.94309056 - samples/sec: 248.18 - lr: 0.000244
2020-09-11 09:32:03,970 epoch 52 - iter 3/17 - loss 2.05831647 - samples/sec: 306.61 - lr: 0.000244
2020-09-11 09:32:04,083 epoch 52 - iter 4/17 - loss 2.07525927 - samples/sec: 286.32 - lr: 0.000244
2020-09-11 09:32:04,197 epoch 52 - iter 5/17 - loss 2.31849523 - samples/sec: 284.32 - lr: 0.000244
2020-09-11 09:32:04,302 epoch 52 - iter 6/17 - loss 2.25557834 - samples/sec: 307.19 - lr: 0.000244
2020-09-11 09:32:04,416 epoch 52 - iter 7/17 - loss 2.21537377 - samples/sec: 282.38 - lr: 0.000244
2020-09-11 09:32:04,532 epoch 52 - iter 8/17 - loss 2.16501409 - samples/sec: 278.57 - lr: 0.000244
2020-09-11 09:32:04,633 epoch 52 - iter 9/17 - loss 2.17519458 - samples/sec: 322.09 - lr: 0.000244
2020-09-11 09:32:04,758 epoch 52 - iter 10/17 - loss 2.25405197 - samples/sec: 265.34 - lr: 0.000244
2020-09-11 09:32:04,866 epoch 52 - iter 11/17 - loss 2.22692343 - samples/sec: 298.33 - lr: 0.000244
2020-09-11 09:32:04,971 epoch 52 - iter 12/17 - loss 2.20147764 - samples/sec: 309.76 - lr: 0.000244
2020-09-11 09:32:05,077 epoch 52 - iter 13/17 - loss 2.19556830 - samples/sec: 305.14 - lr: 0.000244
2020-09-11 09:32:05,203 epoch 52 - iter 14/17 - loss 2.16996110 - samples/sec: 261.25 - lr: 0.000244
2020-09-11 09:32:05,307 epoch 52 - iter 15/17 - loss 2.15638318 - samples/sec: 308.44 - lr: 0.000244
2020-09-11 09:32:05,425 epoch 52 - iter 16/17 - loss 2.13956507 - samples/sec: 275.18 - lr: 0.000244
2020-09-11 09:32:05,493 epoch 52 - iter 17/17 - loss 2.07687952 - samples/sec: 475.48 - lr: 0.000244
2020-09-11 09:32:05,494 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:05,499 EPOCH 52 done: loss 2.0769 - lr 0.0002441
2020-09-11 09:32:05,575 DEV : loss 1.8710246086120605 - score 0.9039
2020-09-11 09:32:05,578 BAD EPOCHS (no improvement): 1
2020-09-11 09:32:05,580 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:05,696 epoch 53 - iter 1/17 - loss 2.67301607 - samples/sec: 279.99 - lr: 0.000244
2020-09-11 09:32:05,810 epoch 53 - iter 2/17 - loss 2.87615097 - samples/sec: 284.34 - lr: 0.000244
2020-09-11 09:32:05,925 epoch 53 - iter 3/17 - loss 2.46839376 - samples/sec: 279.79 - lr: 0.000244
2020-09-11 09:32:06,049 epoch 53 - iter 4/17 - loss 2.64226267 - samples/sec: 261.68 - lr: 0.000244
2020-09-11 09:32:06,156 epoch 53 - iter 5/17 - loss 2.48819025 - samples/sec: 300.57 - lr: 0.000244
2020-09-11 09:32:06,263 epoch 53 - iter 6/17 - loss 2.42314504 - samples/sec: 302.87 - lr: 0.000244
2020-09-11 09:32:06,366 epoch 53 - iter 7/17 - loss 2.33482059 - samples/sec: 313.20 - lr: 0.000244
2020-09-11 09:32:06,481 epoch 53 - iter 8/17 - loss 2.23332472 - samples/sec: 281.06 - lr: 0.000244
2020-09-11 09:32:06,593 epoch 53 - iter 9/17 - loss 2.26868233 - samples/sec: 294.98 - lr: 0.000244
2020-09-11 09:32:06,700 epoch 53 - iter 10/17 - loss 2.20803412 - samples/sec: 302.27 - lr: 0.000244
2020-09-11 09:32:06,808 epoch 53 - iter 11/17 - loss 2.17226865 - samples/sec: 299.75 - lr: 0.000244
2020-09-11 09:32:06,903 epoch 53 - iter 12/17 - loss 2.13279492 - samples/sec: 342.01 - lr: 0.000244
2020-09-11 09:32:07,007 epoch 53 - iter 13/17 - loss 2.14206107 - samples/sec: 309.29 - lr: 0.000244
2020-09-11 09:32:07,118 epoch 53 - iter 14/17 - loss 2.15092226 - samples/sec: 293.50 - lr: 0.000244
2020-09-11 09:32:07,229 epoch 53 - iter 15/17 - loss 2.18495049 - samples/sec: 289.12 - lr: 0.000244
2020-09-11 09:32:07,337 epoch 53 - iter 16/17 - loss 2.16140499 - samples/sec: 298.82 - lr: 0.000244
2020-09-11 09:32:07,415 epoch 53 - iter 17/17 - loss 2.15657809 - samples/sec: 418.69 - lr: 0.000244
2020-09-11 09:32:07,416 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:07,421 EPOCH 53 done: loss 2.1566 - lr 0.0002441
2020-09-11 09:32:07,507 DEV : loss 1.8710384368896484 - score 0.9039
2020-09-11 09:32:07,510 BAD EPOCHS (no improvement): 2
2020-09-11 09:32:07,514 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:07,626 epoch 54 - iter 1/17 - loss 2.50339174 - samples/sec: 292.09 - lr: 0.000244
2020-09-11 09:32:07,735 epoch 54 - iter 2/17 - loss 2.18140179 - samples/sec: 315.09 - lr: 0.000244
2020-09-11 09:32:07,850 epoch 54 - iter 3/17 - loss 2.23897684 - samples/sec: 282.39 - lr: 0.000244
2020-09-11 09:32:07,966 epoch 54 - iter 4/17 - loss 2.10332695 - samples/sec: 279.99 - lr: 0.000244
2020-09-11 09:32:08,088 epoch 54 - iter 5/17 - loss 2.13238862 - samples/sec: 263.80 - lr: 0.000244
2020-09-11 09:32:08,206 epoch 54 - iter 6/17 - loss 2.24009635 - samples/sec: 273.54 - lr: 0.000244
2020-09-11 09:32:08,329 epoch 54 - iter 7/17 - loss 2.30253548 - samples/sec: 269.00 - lr: 0.000244
2020-09-11 09:32:08,433 epoch 54 - iter 8/17 - loss 2.18023707 - samples/sec: 313.01 - lr: 0.000244
2020-09-11 09:32:08,562 epoch 54 - iter 9/17 - loss 2.13453245 - samples/sec: 250.65 - lr: 0.000244
2020-09-11 09:32:08,672 epoch 54 - iter 10/17 - loss 2.19992437 - samples/sec: 292.82 - lr: 0.000244
2020-09-11 09:32:08,787 epoch 54 - iter 11/17 - loss 2.15717849 - samples/sec: 281.81 - lr: 0.000244
2020-09-11 09:32:08,893 epoch 54 - iter 12/17 - loss 2.14129857 - samples/sec: 307.05 - lr: 0.000244
2020-09-11 09:32:08,999 epoch 54 - iter 13/17 - loss 2.16953674 - samples/sec: 304.10 - lr: 0.000244
2020-09-11 09:32:09,107 epoch 54 - iter 14/17 - loss 2.17339781 - samples/sec: 297.71 - lr: 0.000244
2020-09-11 09:32:09,219 epoch 54 - iter 15/17 - loss 2.16997722 - samples/sec: 289.97 - lr: 0.000244
2020-09-11 09:32:09,325 epoch 54 - iter 16/17 - loss 2.14441101 - samples/sec: 305.37 - lr: 0.000244
2020-09-11 09:32:09,402 epoch 54 - iter 17/17 - loss 2.09969692 - samples/sec: 418.75 - lr: 0.000244
2020-09-11 09:32:09,404 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:09,409 EPOCH 54 done: loss 2.0997 - lr 0.0002441
2020-09-11 09:32:09,497 DEV : loss 1.8709750175476074 - score 0.9039
2020-09-11 09:32:09,499 BAD EPOCHS (no improvement): 3
2020-09-11 09:32:09,500 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:09,620 epoch 55 - iter 1/17 - loss 1.90178788 - samples/sec: 272.99 - lr: 0.000244
2020-09-11 09:32:09,728 epoch 55 - iter 2/17 - loss 1.98399633 - samples/sec: 301.43 - lr: 0.000244
2020-09-11 09:32:09,841 epoch 55 - iter 3/17 - loss 1.99539125 - samples/sec: 285.31 - lr: 0.000244
2020-09-11 09:32:09,953 epoch 55 - iter 4/17 - loss 1.90213764 - samples/sec: 289.10 - lr: 0.000244
2020-09-11 09:32:10,065 epoch 55 - iter 5/17 - loss 1.96329775 - samples/sec: 288.70 - lr: 0.000244
2020-09-11 09:32:10,176 epoch 55 - iter 6/17 - loss 1.96443625 - samples/sec: 292.03 - lr: 0.000244
2020-09-11 09:32:10,288 epoch 55 - iter 7/17 - loss 2.03388395 - samples/sec: 286.98 - lr: 0.000244
2020-09-11 09:32:10,409 epoch 55 - iter 8/17 - loss 2.09622708 - samples/sec: 269.29 - lr: 0.000244
2020-09-11 09:32:10,530 epoch 55 - iter 9/17 - loss 2.17744046 - samples/sec: 266.37 - lr: 0.000244
2020-09-11 09:32:10,645 epoch 55 - iter 10/17 - loss 2.20716212 - samples/sec: 281.23 - lr: 0.000244
2020-09-11 09:32:10,762 epoch 55 - iter 11/17 - loss 2.20755792 - samples/sec: 275.91 - lr: 0.000244
2020-09-11 09:32:10,885 epoch 55 - iter 12/17 - loss 2.21960968 - samples/sec: 261.72 - lr: 0.000244
2020-09-11 09:32:10,988 epoch 55 - iter 13/17 - loss 2.20926589 - samples/sec: 320.85 - lr: 0.000244
2020-09-11 09:32:11,106 epoch 55 - iter 14/17 - loss 2.12377312 - samples/sec: 272.30 - lr: 0.000244
2020-09-11 09:32:11,211 epoch 55 - iter 15/17 - loss 2.13603848 - samples/sec: 308.71 - lr: 0.000244
2020-09-11 09:32:11,316 epoch 55 - iter 16/17 - loss 2.14775415 - samples/sec: 309.65 - lr: 0.000244
2020-09-11 09:32:11,392 epoch 55 - iter 17/17 - loss 2.12510460 - samples/sec: 423.90 - lr: 0.000244
2020-09-11 09:32:11,394 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:11,399 EPOCH 55 done: loss 2.1251 - lr 0.0002441
2020-09-11 09:32:11,476 DEV : loss 1.8711307048797607 - score 0.9039
Epoch 55: reducing learning rate of group 0 to 1.2207e-04.
2020-09-11 09:32:11,478 BAD EPOCHS (no improvement): 4
2020-09-11 09:32:11,483 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:11,597 epoch 56 - iter 1/17 - loss 2.13230896 - samples/sec: 287.93 - lr: 0.000122
2020-09-11 09:32:11,706 epoch 56 - iter 2/17 - loss 2.16884816 - samples/sec: 300.55 - lr: 0.000122
2020-09-11 09:32:11,813 epoch 56 - iter 3/17 - loss 2.14700532 - samples/sec: 299.71 - lr: 0.000122
2020-09-11 09:32:11,925 epoch 56 - iter 4/17 - loss 2.10528451 - samples/sec: 289.35 - lr: 0.000122
2020-09-11 09:32:12,042 epoch 56 - iter 5/17 - loss 2.02419295 - samples/sec: 275.70 - lr: 0.000122
2020-09-11 09:32:12,150 epoch 56 - iter 6/17 - loss 2.09250812 - samples/sec: 302.22 - lr: 0.000122
2020-09-11 09:32:12,257 epoch 56 - iter 7/17 - loss 1.99774071 - samples/sec: 302.23 - lr: 0.000122
2020-09-11 09:32:12,380 epoch 56 - iter 8/17 - loss 1.98823553 - samples/sec: 262.22 - lr: 0.000122
2020-09-11 09:32:12,478 epoch 56 - iter 9/17 - loss 2.01300555 - samples/sec: 329.48 - lr: 0.000122
2020-09-11 09:32:12,603 epoch 56 - iter 10/17 - loss 1.96308876 - samples/sec: 260.03 - lr: 0.000122
2020-09-11 09:32:12,722 epoch 56 - iter 11/17 - loss 2.12541857 - samples/sec: 270.64 - lr: 0.000122
2020-09-11 09:32:12,831 epoch 56 - iter 12/17 - loss 2.09370079 - samples/sec: 299.12 - lr: 0.000122
2020-09-11 09:32:12,934 epoch 56 - iter 13/17 - loss 2.14898827 - samples/sec: 312.36 - lr: 0.000122
2020-09-11 09:32:13,042 epoch 56 - iter 14/17 - loss 2.11128387 - samples/sec: 299.37 - lr: 0.000122
2020-09-11 09:32:13,149 epoch 56 - iter 15/17 - loss 2.12494528 - samples/sec: 303.02 - lr: 0.000122
2020-09-11 09:32:13,262 epoch 56 - iter 16/17 - loss 2.16102227 - samples/sec: 286.20 - lr: 0.000122
2020-09-11 09:32:13,333 epoch 56 - iter 17/17 - loss 2.14927550 - samples/sec: 456.69 - lr: 0.000122
2020-09-11 09:32:13,335 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:13,337 EPOCH 56 done: loss 2.1493 - lr 0.0001221
2020-09-11 09:32:13,416 DEV : loss 1.8712515830993652 - score 0.9039
2020-09-11 09:32:13,420 BAD EPOCHS (no improvement): 1
2020-09-11 09:32:13,423 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:13,542 epoch 57 - iter 1/17 - loss 2.06140113 - samples/sec: 272.59 - lr: 0.000122
2020-09-11 09:32:13,658 epoch 57 - iter 2/17 - loss 2.09485781 - samples/sec: 281.52 - lr: 0.000122
2020-09-11 09:32:13,790 epoch 57 - iter 3/17 - loss 2.17190838 - samples/sec: 246.30 - lr: 0.000122
2020-09-11 09:32:13,914 epoch 57 - iter 4/17 - loss 2.08750546 - samples/sec: 261.82 - lr: 0.000122
2020-09-11 09:32:14,037 epoch 57 - iter 5/17 - loss 2.17361174 - samples/sec: 261.26 - lr: 0.000122
2020-09-11 09:32:14,144 epoch 57 - iter 6/17 - loss 2.15775029 - samples/sec: 303.08 - lr: 0.000122
2020-09-11 09:32:14,252 epoch 57 - iter 7/17 - loss 2.11520788 - samples/sec: 300.57 - lr: 0.000122
2020-09-11 09:32:14,354 epoch 57 - iter 8/17 - loss 2.15911621 - samples/sec: 318.04 - lr: 0.000122
2020-09-11 09:32:14,460 epoch 57 - iter 9/17 - loss 2.15239777 - samples/sec: 303.85 - lr: 0.000122
2020-09-11 09:32:14,576 epoch 57 - iter 10/17 - loss 2.22143736 - samples/sec: 277.67 - lr: 0.000122
2020-09-11 09:32:14,675 epoch 57 - iter 11/17 - loss 2.18487239 - samples/sec: 327.67 - lr: 0.000122
2020-09-11 09:32:14,784 epoch 57 - iter 12/17 - loss 2.16919176 - samples/sec: 316.15 - lr: 0.000122
2020-09-11 09:32:14,890 epoch 57 - iter 13/17 - loss 2.11041773 - samples/sec: 305.92 - lr: 0.000122
2020-09-11 09:32:14,997 epoch 57 - iter 14/17 - loss 2.12899177 - samples/sec: 301.15 - lr: 0.000122
2020-09-11 09:32:15,132 epoch 57 - iter 15/17 - loss 2.10911355 - samples/sec: 239.11 - lr: 0.000122
2020-09-11 09:32:15,244 epoch 57 - iter 16/17 - loss 2.15161987 - samples/sec: 288.10 - lr: 0.000122
2020-09-11 09:32:15,322 epoch 57 - iter 17/17 - loss 2.11003646 - samples/sec: 417.30 - lr: 0.000122
2020-09-11 09:32:15,323 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:15,328 EPOCH 57 done: loss 2.1100 - lr 0.0001221
2020-09-11 09:32:15,407 DEV : loss 1.871190071105957 - score 0.9039
2020-09-11 09:32:15,411 BAD EPOCHS (no improvement): 2
2020-09-11 09:32:15,413 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:15,520 epoch 58 - iter 1/17 - loss 2.92842674 - samples/sec: 303.98 - lr: 0.000122
2020-09-11 09:32:15,634 epoch 58 - iter 2/17 - loss 2.42707694 - samples/sec: 281.85 - lr: 0.000122
2020-09-11 09:32:15,771 epoch 58 - iter 3/17 - loss 2.23341076 - samples/sec: 237.25 - lr: 0.000122
2020-09-11 09:32:15,892 epoch 58 - iter 4/17 - loss 2.20908457 - samples/sec: 265.75 - lr: 0.000122
2020-09-11 09:32:16,007 epoch 58 - iter 5/17 - loss 2.17910295 - samples/sec: 283.48 - lr: 0.000122
2020-09-11 09:32:16,104 epoch 58 - iter 6/17 - loss 2.05788509 - samples/sec: 333.94 - lr: 0.000122
2020-09-11 09:32:16,215 epoch 58 - iter 7/17 - loss 2.08856239 - samples/sec: 289.41 - lr: 0.000122
2020-09-11 09:32:16,330 epoch 58 - iter 8/17 - loss 2.14802042 - samples/sec: 281.15 - lr: 0.000122
2020-09-11 09:32:16,439 epoch 58 - iter 9/17 - loss 2.17300524 - samples/sec: 297.83 - lr: 0.000122
2020-09-11 09:32:16,537 epoch 58 - iter 10/17 - loss 2.16579535 - samples/sec: 332.77 - lr: 0.000122
2020-09-11 09:32:16,665 epoch 58 - iter 11/17 - loss 2.21264616 - samples/sec: 251.48 - lr: 0.000122
2020-09-11 09:32:16,770 epoch 58 - iter 12/17 - loss 2.19528157 - samples/sec: 306.98 - lr: 0.000122
2020-09-11 09:32:16,885 epoch 58 - iter 13/17 - loss 2.11050682 - samples/sec: 282.85 - lr: 0.000122
2020-09-11 09:32:16,993 epoch 58 - iter 14/17 - loss 2.14877816 - samples/sec: 298.19 - lr: 0.000122
2020-09-11 09:32:17,100 epoch 58 - iter 15/17 - loss 2.20244874 - samples/sec: 302.51 - lr: 0.000122
2020-09-11 09:32:17,200 epoch 58 - iter 16/17 - loss 2.16341272 - samples/sec: 323.63 - lr: 0.000122
2020-09-11 09:32:17,270 epoch 58 - iter 17/17 - loss 2.08311685 - samples/sec: 465.62 - lr: 0.000122
2020-09-11 09:32:17,271 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:17,276 EPOCH 58 done: loss 2.0831 - lr 0.0001221
2020-09-11 09:32:17,351 DEV : loss 1.8712575435638428 - score 0.9039
2020-09-11 09:32:17,356 BAD EPOCHS (no improvement): 3
2020-09-11 09:32:17,357 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:17,470 epoch 59 - iter 1/17 - loss 3.57667923 - samples/sec: 288.24 - lr: 0.000122
2020-09-11 09:32:17,570 epoch 59 - iter 2/17 - loss 2.45157045 - samples/sec: 324.19 - lr: 0.000122
2020-09-11 09:32:17,685 epoch 59 - iter 3/17 - loss 2.40864130 - samples/sec: 279.59 - lr: 0.000122
2020-09-11 09:32:17,793 epoch 59 - iter 4/17 - loss 2.30390897 - samples/sec: 301.24 - lr: 0.000122
2020-09-11 09:32:17,896 epoch 59 - iter 5/17 - loss 2.12980967 - samples/sec: 314.40 - lr: 0.000122
2020-09-11 09:32:18,015 epoch 59 - iter 6/17 - loss 2.18769928 - samples/sec: 275.16 - lr: 0.000122
2020-09-11 09:32:18,123 epoch 59 - iter 7/17 - loss 2.13403550 - samples/sec: 301.61 - lr: 0.000122
2020-09-11 09:32:18,230 epoch 59 - iter 8/17 - loss 2.11809765 - samples/sec: 299.58 - lr: 0.000122
2020-09-11 09:32:18,340 epoch 59 - iter 9/17 - loss 2.08499732 - samples/sec: 293.75 - lr: 0.000122
2020-09-11 09:32:18,452 epoch 59 - iter 10/17 - loss 2.08781813 - samples/sec: 292.03 - lr: 0.000122
2020-09-11 09:32:18,565 epoch 59 - iter 11/17 - loss 2.06968793 - samples/sec: 287.39 - lr: 0.000122
2020-09-11 09:32:18,690 epoch 59 - iter 12/17 - loss 2.08348020 - samples/sec: 257.11 - lr: 0.000122
2020-09-11 09:32:18,812 epoch 59 - iter 13/17 - loss 2.08096592 - samples/sec: 265.63 - lr: 0.000122
2020-09-11 09:32:18,914 epoch 59 - iter 14/17 - loss 2.06882810 - samples/sec: 317.14 - lr: 0.000122
2020-09-11 09:32:19,041 epoch 59 - iter 15/17 - loss 2.13064247 - samples/sec: 256.00 - lr: 0.000122
2020-09-11 09:32:19,150 epoch 59 - iter 16/17 - loss 2.16788667 - samples/sec: 297.97 - lr: 0.000122
2020-09-11 09:32:19,222 epoch 59 - iter 17/17 - loss 2.12413386 - samples/sec: 451.70 - lr: 0.000122
2020-09-11 09:32:19,223 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:19,225 EPOCH 59 done: loss 2.1241 - lr 0.0001221
2020-09-11 09:32:19,307 DEV : loss 1.8713274002075195 - score 0.9039
Epoch 59: reducing learning rate of group 0 to 6.1035e-05.
2020-09-11 09:32:19,309 BAD EPOCHS (no improvement): 4
2020-09-11 09:32:19,312 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:19,313 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:19,314 learning rate too small - quitting training!
2020-09-11 09:32:19,316 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:22,073 ----------------------------------------------------------------------------------------------------
2020-09-11 09:32:22,074 Testing using best model ...
2020-09-11 09:32:22,075 loading file /content/best-model.pt
2020-09-11 09:32:23,480 0.9176 0.9455 0.9313
2020-09-11 09:32:23,481
Results:
- F1-score (micro) 0.9313
- F1-score (macro) 0.7715
By class:
- tp: 0 - fp: 0 - fn: 1 - precision: 0.0000 - recall: 0.0000 - f1-score: 0.0000
City tp: 29 - fp: 2 - fn: 2 - precision: 0.9355 - recall: 0.9355 - f1-score: 0.9355
Corporation tp: 13 - fp: 1 - fn: 1 - precision: 0.9286 - recall: 0.9286 - f1-score: 0.9286
Country tp: 11 - fp: 0 - fn: 0 - precision: 1.0000 - recall: 1.0000 - f1-score: 1.0000
GivenName tp: 7 - fp: 1 - fn: 0 - precision: 0.8750 - recall: 1.0000 - f1-score: 0.9333
Household tp: 9 - fp: 0 - fn: 1 - precision: 1.0000 - recall: 0.9000 - f1-score: 0.9474
MiddleName tp: 0 - fp: 2 - fn: 0 - precision: 0.0000 - recall: 0.0000 - f1-score: 0.0000
State tp: 29 - fp: 1 - fn: 1 - precision: 0.9667 - recall: 0.9667 - f1-score: 0.9667
StreetAddress tp: 28 - fp: 3 - fn: 3 - precision: 0.9032 - recall: 0.9032 - f1-score: 0.9032
SurName tp: 7 - fp: 1 - fn: 0 - precision: 0.8750 - recall: 1.0000 - f1-score: 0.9333
Zipcode tp: 23 - fp: 3 - fn: 0 - precision: 0.8846 - recall: 1.0000 - f1-score: 0.9388
2020-09-11 09:32:23,484 ----------------------------------------------------------------------------------------------------
{'dev_loss_history': [5.717854976654053,
7.014471054077148,
6.267121315002441,
4.367828369140625,
7.306588172912598,
3.1234371662139893,
2.352734327316284,
2.3950960636138916,
2.601285934448242,
2.621568202972412,
2.6496293544769287,
2.456434726715088,
2.8066673278808594,
2.356717109680176,
2.2419240474700928,
2.1843576431274414,
2.249250888824463,
2.1045773029327393,
1.987985372543335,
2.02628755569458,
1.9274215698242188,
1.9331698417663574,
1.9181852340698242,
1.9145869016647339,
1.8926992416381836,
1.829909324645996,
1.877849817276001,
1.8591759204864502,
1.8752442598342896,
1.8820831775665283,
1.87209951877594,
1.8655517101287842,
1.860945701599121,
1.8520863056182861,
1.8610153198242188,
1.8612797260284424,
1.862546443939209,
1.869424819946289,
1.8706583976745605,
1.8684940338134766,
1.8697052001953125,
1.8711341619491577,
1.871195912361145,
1.8707306385040283,
1.8710083961486816,
1.8702256679534912,
1.8696794509887695,
1.869992733001709,
1.8705830574035645,
1.8708758354187012,
1.8709588050842285,
1.8710246086120605,
1.8710384368896484,
1.8709750175476074,
1.8711307048797607,
1.8712515830993652,
1.871190071105957,
1.8712575435638428,
1.8713274002075195],
'dev_score_history': [0.8697916666666667,
0.7978436657681941,
0.7704918032786885,
0.8491048593350384,
0.8608923884514434,
0.8808290155440414,
0.8164383561643836,
0.8911392405063291,
0.8783068783068783,
0.8877551020408163,
0.9175257731958764,
0.8571428571428572,
0.8548387096774193,
0.903896103896104,
0.8963730569948186,
0.8963730569948186,
0.8911917098445595,
0.8900523560209423,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104,
0.903896103896104],
'test_score': 0.9313432835820895,
'train_loss_history': [9.025349981644574,
6.609698015100816,
6.332132143132827,
5.108250505783978,
5.506425955716302,
3.8211376667022705,
3.362795156591079,
3.259746509439805,
3.1422233791912304,
3.286931914441726,
3.1722532090018776,
3.1396323021720436,
2.9533899181029377,
3.0463009301353905,
2.8610909195507275,
2.526246000738705,
2.491676554960363,
2.6032417311387905,
2.377417711650624,
2.386420453295988,
2.244826681473676,
2.190858244895935,
2.267894345171311,
2.218476274434258,
2.264526696766124,
2.204820352442124,
2.123208859387566,
2.1706547877367806,
2.2194581031799316,
2.2134850586161896,
2.1507136681500603,
2.1263834097806145,
2.1835697889328003,
2.2314544004552506,
2.1594833345974194,
2.1696282975813923,
2.1386386717067047,
2.0764440298080444,
2.107362817315494,
2.200913576518788,
2.132053866105921,
2.069297019173117,
2.1372564610312965,
2.084194351645077,
2.1103776973836563,
2.140008344369776,
2.1633326095693253,
2.1525954989825977,
2.16026076148538,
2.1881491226308487,
2.152092737310073,
2.076879515367396,
2.1565780850017773,
2.0996969166923973,
2.1251045956331143,
2.149275499231675,
2.1100364572861614,
2.0831168504322277,
2.12413386036368]}
from flair.data import Sentence
from flair.models import SequenceTagger
model = SequenceTagger.load('/content/final-model.pt')
text = '<redacted>'
sentence = Sentence(text)
model.predict(sentence)
out = {}
for entity in sentence.get_spans('ner'):
out[entity.tag] = entity.text
2020-09-11 09:33:19,511 loading file /content/final-model.pt
tags = ['EntityType','Recipient','Address',
'GivenName','MiddleName','SurName','Household','Corporation',
'StreetAddress','City','State','Zipcode','Country', 'Confidence']
PAD_TAIL = '<redacted>'
def household_patch(row):
if row.EntityType=='Household':
text = row.Household.split('&')[0] if '&' in row.Household else ' '.join(row.Household.split()[:2])
text = parseit(text+PAD_TAIL)
row['GivenName'] = text.loc[0,'GivenName']
row['SurName'] = text.loc[0,'SurName']
row['MiddleName'] = text.loc[0,'MiddleName']
return row
def add_on(xx):
# Address is the combination of street, city, state, zip and country
xx['Address'] = xx['StreetAddress'] +' '+ xx['City'] +' '+ xx['State'] +' '+ xx['Zipcode'] +' '+ xx['Country']
# default is person, if household column is not empty, then household, same for corporation
xx['EntityType'] = 'Person'
xx.loc[xx.Household!='','EntityType'] = 'Household'
xx.loc[xx.Corporation!='','EntityType'] = 'Corporation'
# default is full name of person, if entity is corporation ,then corporation name, same for household
xx['Recipient'] = xx['SurName'] +' '+ xx['GivenName'] +' '+ xx['MiddleName']
xx['Recipient'] = xx.apply(lambda row: row.Corporation if row.EntityType=='Corporation' else row.Recipient, axis=1)
xx['Recipient'] = xx.apply(lambda row: row.Household if row.EntityType=='Household' else row.Recipient, axis=1)
# adding household to name field patch
xx = xx.apply(household_patch, axis=1)
# converting to dictionary format
xx = xx.replace(r'^\s*$', np.nan, regex=True).dropna(axis=1, how='any')
xx = xx.T.to_dict()[0]
# return the processed data
return xx
def parseit(text):
text = str(text).upper()
text = re.sub('"','',text)
text = re.sub(',','',text)
text = ' '.join(text.split())
output = {}
sentence = Sentence(text)
model.predict(sentence)
for entity in sentence.get_spans('ner'):
output[entity.tag] = entity.text
df = pd.DataFrame(output, index=[0])
dfx = pd.DataFrame(columns=tags)
dfx = dfx.append(df).fillna('')
dfx['Confidence'] = ','.join([f'[{x.tag} {str(round(x.score,2))}]'for x in sentence.get_spans('ner')])
return dfx
def func(text):
X = parseit(text)
X = add_on(X)
X = str(X)
return X
func('<redacted>')
!pip install -q gradio
import gradio as gr
inputs = gr.inputs.Textbox(lines=3, label='Input')
outputs = gr.outputs.Textbox(label='Output')
gr.Interface(fn=func, inputs=inputs, outputs=outputs).launch()
API
!pip install -q probablepeople
!pip install -q usaddress
import os
import re
import numpy as np
import pandas as pd
from tqdm import tqdm
import seaborn as sns
import matplotlib.pyplot as plt
import probablepeople as pp
import usaddress as ua
import warnings
warnings.filterwarnings("ignore")
tqdm.pandas()
%reload_ext autoreload
%autoreload 2
%reload_ext google.colab.data_table
%config InlineBackend.figure_format = 'retina'
plt.style.use('fivethirtyeight')
plt.style.use('seaborn-notebook')
df = pd.read_pickle(os.path.join(path,'sample.p'))
data = df.oor.tolist()
data[:10]
def patch_country(text):
combos = [' US ',' USA ',' U.S. ',' U.S.A. ']
cflagz = ''
otext = text
text = text[-10:]
text = ' ' + text + ' '
for i in combos:
if i in text:
cflagz = i
text = otext[:-10] + re.sub(cflagz, '', text)
cflagz = cflagz.strip()
return text, cflagz
def patch_household(xz):
xx = pd.DataFrame(pp.tag(xz)[0], index=[0])
xx['Type'] = 'Person'
if 'And' in xx.columns.tolist():
xx = pd.DataFrame({'Household':xz}, index=[0])
xx['Type'] = 'Household'
return xx
tags = pp.LABELS
tags.extend(ua.LABELS)
tags = list(set(tags))
additionals = ['CountryName', 'Household', 'Type',
'NameConfidence', 'AddrConfidence', 'Text']
tags.extend(additionals)
dfx = pd.DataFrame(columns=tags)
dfx = dfx.loc[:,~dfx.columns.duplicated()]
def process_text(text):
text = re.sub(',','',text)
text = ' '.join(text.split())
text = re.sub('UNITED STATES OF AMERICA','US',text)
text = re.sub('TEXAS','TX',text)
return text
def temp(text):
text = process_text(text)
text, cflag = patch_country(text)
# address parsing
try:
df1 = pd.DataFrame(ua.tag(text)[0], index=[0])
df1['AddrConfidence'] = ua.tag(text)[1]
except:
df1 = pd.DataFrame(ua.parse(text)).groupby(1).agg({0: lambda x: ' '.join(x)}).T
df1['AddrConfidence'] = 'Error'
# address to name linking
try:
xz = df1.Recipient.values[0]
except:
xz = text
# name parsing
try:
df2 = pd.DataFrame(pp.tag(xz)[0], index=[0])
df2['NameConfidence'] = pp.tag(xz)[1]
except:
df2 = pd.DataFrame(pp.parse(xz)).groupby(1).agg({0: lambda x: ' '.join(x)}).T
df2['NameConfidence'] = 'Error'
df2['Type'] = 'Person'
# household name patch
if 'And' in df2.columns.tolist():
df2 = pd.DataFrame({'Household':xz}, index=[0])
df2['Type'] = 'Household'
# concatenation
df = pd.concat([df1,df2], axis=1)
df['Text'] = text
df['CountryName'] = cflag if cflag!='' else np.NaN
dfx = pd.DataFrame(columns=tags)
dfx = dfx.loc[:,~dfx.columns.duplicated()]
df = dfx.append(df)
df = df.fillna('')
df.loc[df['CorporationName']!='', 'Type'] = 'Corporation'
if (df['MiddleInitial'][0]=='') & (df['MiddleName'][0]!=''):
gname, mname, sname = df.MiddleName, df.Surname, df.GivenName
df['MiddleName'], df['Surname'], dfx['GivenName'] = mname[0], sname[0], gname[0]
# elif (df['MiddleInitial'][0]=='') & (df['MiddleName'][0]==''):
# gname, sname = df.Surname, df.GivenName
# df['Surname'], df['GivenName'] = sname[0], gname[0]
return df
tagmap = pd.read_excel(os.path.join(path,'tag_map.xlsx')).reset_index()
tagmap.columns = ['_'.join(col.split()) for col in tagmap.columns]
tagmap
tagmap = pd.melt(tagmap, id_vars=['index'], value_vars=['FIRSTNAME','MIDDLENAME','LASTNAME',
'OTHERNAME','HOUSEHOLD','CORPORATION',
'STREET_ADDRESS', 'CITY', 'STATE', 'PINCODE',
'COUNTRY', 'META', 'OTHER']).dropna()[['variable','value']].set_index('value').to_dict()['variable']
def custom_patch(text):
text = re.sub(' DALLA S ', ' DALLAS ', text)
text = re.sub(' PRAIRI E ', ' PRAIRIE ', text)
text = re.sub(' EL PAS O ', ' EL PASO ', text)
text = re.sub(' SAN ANT ONIO ', ' SAN ANTONIO ', text)
text = re.sub(' THE WOODL ANDS ', ' THE WOODLANDS ', text)
text = re.sub(' FORT WORT H ', ' FORT WORTH ', text)
text = re.sub(' ARLINGTO N ', ' ARLINGTON ', text)
text = re.sub(' SOCORR O ', ' SOCORRO ', text)
text = re.sub(' HURS T ', ' HURST ', text)
text = re.sub(' HOLLYWOO D ', ' HOLLYWOOD ', text)
return text
def oor_api(text):
texto = text
xx = temp(text)
xx.columns = xx.columns.to_series().map(tagmap)
xx = xx.drop('META', axis=1)
xx = xx.replace(r'^\s*$', np.nan, regex=True).dropna(axis=1, how='any')
try:
xx['PINCODE'] = xx['PINCODE'].apply(lambda x: re.sub(' ','',x))
except:
pass
xx = xx.T
xx = xx.reset_index()
xx.index = xx.index.map(str)
xx = ' '+ xx + ' '
xx[0] = xx[0].apply(custom_patch)
xx = xx.to_dict()
label2tag = {str(v): ' '+str(k)+' ' for k, v in xx[0].items()}
text = process_text(text)
text = ' ' + text + ' '
for key, value in label2tag.items():
text = text.replace(key, value)
x = pd.DataFrame({'entity':text.split(), 'tag':text.split()})
x = x.replace({'entity': xx[0], 'tag': xx['index']})
x['entity'] = x.groupby(['tag'])['entity'].transform(lambda x: ' '.join(x))
x = x.drop_duplicates()
x['tag'] = x['tag'].apply(process_text)
x['entity'] = x['entity'].apply(process_text)
x = x.append(pd.Series({'tag':'TEXT','entity':texto}), ignore_index=True)
x = x.set_index('tag')
x = x.to_dict()['entity']
return x
# zz = data[5]
# oor_api(zz)
oor_cols = ['TEXT','FIRSTNAME','MIDDLENAME','LASTNAME', 'OTHERNAME', 'HOUSEHOLD',
'CORPORATION', 'STREET_ADDRESS', 'CITY', 'STATE', 'PINCODE', 'COUNTRY']
dfx = pd.DataFrame(columns=oor_cols)
for idx, text in tqdm(enumerate(data)):
dfx = dfx.append(pd.DataFrame(oor_api(text), index=[idx]))
dfx = dfx.fillna('')
dfx = dfx[oor_cols]
1000it [01:03, 15.75it/s]
dfx = dfx.replace(r'^\s*$', np.nan, regex=True)
dfx.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 1000 entries, 0 to 999
Data columns (total 12 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 TEXT 997 non-null object
1 FIRSTNAME 403 non-null object
2 MIDDLENAME 155 non-null object
3 LASTNAME 324 non-null object
4 OTHERNAME 9 non-null object
5 HOUSEHOLD 199 non-null object
6 CORPORATION 364 non-null object
7 STREET_ADDRESS 976 non-null object
8 CITY 870 non-null object
9 STATE 954 non-null object
10 PINCODE 838 non-null object
11 COUNTRY 288 non-null object
dtypes: object(12)
memory usage: 101.6+ KB
dfx.to_csv('x.csv')
x = '<redacted>'
temp(x).T.replace('',np.nan).dropna()
oor_api(x)
Deliverables
# !pip install cookiecutter
# !cookiecutter https://github.com/drivendata/cookiecutter-data-science
imports _requirements.txt_
!pip install flair
# utils
import pandas as pd
tags = ['EntityType','Recipient','Address',
'GivenName','MiddleName','SurName','Household','Corporation',
'StreetAddress','City','State','Zipcode','Country', 'Confidence']
name_tags = ['EntityType','Recipient', 'GivenName','MiddleName','SurName','Household','Corporation']
addr_tags = ['Address','StreetAddress','City','State','Zipcode','Country']
PAD_TAIL = '<redacted>'
PAD_HEAD = '<redacted>'
def household_patch(row):
if row.EntityType=='Household':
text = row.Household.split('&')[0] if '&' in row.Household else ' '.join(row.Household.split()[:2])
text = parseit(text+PAD_TAIL)
row['GivenName'] = text.loc[0,'GivenName']
row['SurName'] = text.loc[0,'SurName']
row['MiddleName'] = text.loc[0,'MiddleName']
return row
def add_on(xx):
# Address is the combination of street, city, state, zip and country
xx['Address'] = xx['StreetAddress'] +' '+ xx['City'] +' '+ xx['State'] +' '+ xx['Zipcode'] +' '+ xx['Country']
# default is person, if household column is not empty, then household, same for corporation
xx['EntityType'] = 'Person'
xx.loc[xx.Household!='','EntityType'] = 'Household'
xx.loc[xx.Corporation!='','EntityType'] = 'Corporation'
# default is full name of person, if entity is corporation ,then corporation name, same for household
xx['Recipient'] = xx['SurName'] +' '+ xx['GivenName'] +' '+ xx['MiddleName']
xx['Recipient'] = xx.apply(lambda row: row.Corporation if row.EntityType=='Corporation' else row.Recipient, axis=1)
xx['Recipient'] = xx.apply(lambda row: row.Household if row.EntityType=='Household' else row.Recipient, axis=1)
# adding household to name field patch
xx = xx.apply(household_patch, axis=1)
# converting to dictionary format
xx = xx.replace(r'^\s*$', np.nan, regex=True).dropna(axis=1, how='any')
# return the processed data
return xx
# pred
import re
import numpy as np
import pandas as pd
from flair.data import Sentence
from flair.models import SequenceTagger
model = SequenceTagger.load('models/final-model.pt')
def _parse(text):
text = str(text).upper()
text = re.sub('"','',text)
text = re.sub(',','',text)
text = ' '.join(text.split())
output = {}
sentence = Sentence(text)
model.predict(sentence)
for entity in sentence.get_spans('ner'):
output[entity.tag] = entity.text
df = pd.DataFrame(output, index=[0])
dfx = pd.DataFrame(columns=tags)
dfx = dfx.append(df).fillna('')
dfx['Confidence'] = ','.join([f'[{x.tag} {str(round(x.score,2))}]'for x in sentence.get_spans('ner')])
return dfx
2020-09-11 13:02:40,896 loading file models/final-model.pt
def _predict(text):
X = _parse(text)
X = add_on(X)
X = X.T.to_dict()[0]
return X
def _predict_name(text):
text = text + PAD_TAIL
X = _parse(text)
X = add_on(X)
rqd_cols = list(set(X.columns) & set(name_tags))
X = X[rqd_cols]
X = X.T.to_dict()[0]
return X
def _predict_address(text):
text = PAD_HEAD + text
X = _parse(text)
X = add_on(X)
rqd_cols = list(set(X.columns) & set(addr_tags))
X = X[rqd_cols]
X = X.T.to_dict()[0]
return X
# app
_predict('<redacted>')
%%writefile src/app.py
import re
import time
import numpy as np
import pandas as pd
from pathlib import Path
from flair.data import Sentence
from flair.models import SequenceTagger
from utils import *
from model import loadmodel
path = Path(__file__)
_PPATH = str(path.parents[1])+'/'
start_time = time.time()
model = loadmodel.finalmodel
print("---Encoder--- %s seconds ---" % (time.time() - start_time))
def _parse(text):
text = str(text).upper()
text = re.sub('"','',text)
text = re.sub(',','',text)
text = ' '.join(text.split())
output = {}
sentence = Sentence(text)
model.predict(sentence)
for entity in sentence.get_spans('ner'):
output[entity.tag] = entity.text
df = pd.DataFrame(output, index=[0])
dfx = pd.DataFrame(columns=tags)
dfx = dfx.append(df).fillna('')
dfx['Confidence'] = ','.join([f'[{x.tag} {str(round(x.score,2))}]'for x in sentence.get_spans('ner')])
return dfx
def _predict(text):
X = _parse(text)
X = add_on(X)
X = X.T.to_dict()[0]
return X
def _predict_name(text):
text = text + PAD_TAIL
X = _parse(text)
X = add_on(X)
rqd_cols = list(set(X.columns) & set(name_tags))
X = X[rqd_cols]
X = X.T.to_dict()[0]
return X
def _predict_address(text):
text = PAD_HEAD + text
X = _parse(text)
X = add_on(X)
rqd_cols = list(set(X.columns) & set(addr_tags))
X = X[rqd_cols]
X = X.T.to_dict()[0]
return X
########## FLASK API ##########
from flask import Flask, request, jsonify, send_file
import json
app = Flask(__name__)
@app.route("/oor", methods=["POST"])
def parse_oor():
req_data = request.get_json()
text = req_data['query']
preds = _predict(text)
return jsonify(preds)
@app.route("/name", methods=["POST"])
def parse_name():
req_data = request.get_json()
text = req_data['query']
preds = _predict_name(text)
return jsonify(preds)
@app.route("/address", methods=["POST"])
def parse_address():
req_data = request.get_json()
text = req_data['query']
preds = _predict_address(text)
return jsonify(preds)
if __name__ == "__main__":
app.run(port=5000, debug=True)
Overwriting src/app.py
!pip install -q pyngrok
from pyngrok import ngrok
!ngrok authtoken <redacted>
Building wheel for pyngrok (setup.py) ... [?25l[?25hdone
Authtoken saved to configuration file: /root/.ngrok2/ngrok.yml
ngrok.kill()
public_url = ngrok.connect(port='5000'); public_url
'http://c1b4cabd8e5b.ngrok.io'
!python src/app.py
2020-09-16 16:05:27.961051: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
2020-09-16 16:05:30,365 loading file /content/p28118/models/final-model.pt
---Encoder--- 1.430511474609375e-06 seconds ---
* Serving Flask app "app" (lazy loading)
* Environment: production
[31m WARNING: This is a development server. Do not use it in a production deployment.[0m
[2m Use a production WSGI server instead.[0m
* Debug mode: on
* Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)
* Restarting with stat
2020-09-16 16:05:35.186608: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
2020-09-16 16:05:37,169 loading file /content/p28118/models/final-model.pt
---Encoder--- 1.1920928955078125e-06 seconds ---
* Debugger is active!
* Debugger PIN: 210-465-697
Pipeline
!pip install -q probablepeople
!pip install -q usaddress
import os
import re
import numpy as np
import pandas as pd
from tqdm import tqdm
tqdm.pandas()
import probablepeople as pp
import usaddress as ua
import shutil
# shutil.make_archive(os.path.join(path,'customNER'),'zip','/content/customNER')
shutil.unpack_archive(os.path.join(path,'customNER.zip'), '/content/customNER', 'zip')
df = pd.read_pickle(os.path.join(path,'sample.p'))
data = df.oor.tolist()
data[100:110]
df.info()
<class 'pandas.core.frame.DataFrame'>
Int64Index: 1000 entries, 3605089 to 2516200
Data columns (total 2 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 _id 1000 non-null object
1 oor 1000 non-null object
dtypes: object(2)
memory usage: 23.4+ KB
import spacy
model = spacy.load('/content/customNER')
text1 = '<redacted>'
def name_parsing(text):
try:
text = pp.tag(text)
textdf = pd.DataFrame(text[0], index=[0])
textdf['NameType'] = text[1]
textdf['NameConfidence'] = 'Tag'
textdf = textdf.T.to_dict()[0]
except:
textdf = pp.parse(text)
textdf = pd.DataFrame(textdf).groupby(1).agg({0: lambda x: ' '.join(x)}).T
textdf['NameType'] = 'Ambiguous'
textdf['NameConfidence'] = 'Parse'
textdf = textdf.T.to_dict()[0]
return textdf
name_parsing(text3)
def preprocess(text):
text = re.sub(',','',text)
text = ' '.join(text.split())
text = re.sub('UNITED STATES OF AMERICA','US',text)
return text
text1 = '<redacted>'
def address_parsing(text):
try:
text = preprocess(text)
text = ua.tag(text)
textdf = pd.DataFrame(text[0], index=[0])
textdf['AddressType'] = text[1]
textdf['AddrConfidence'] = 'Tag'
textdf = textdf.T.to_dict()[0]
except:
textdf = ua.parse(text)
textdf = pd.DataFrame(textdf).groupby(1).agg({0: lambda x: ' '.join(x)}).T
textdf['AddressType'] = 'Ambiguous'
textdf['AddrConfidence'] = 'Parse'
textdf = textdf.T.to_dict()[0]
return textdf
xx = address_parsing(text2); xx
list(xx.keys())
['USPSBoxType',
'USPSBoxID',
'PlaceName',
'StateName',
'ZipCode',
'ZipPlus4',
'CountryName',
'AddressType',
'AddrConfidence']
text1 = '<redacted>'
def full_parsing(text):
doc = model(text)
ners = {ent.label_:ent.text for ent in doc.ents}
name = address = {}
try:
address = address_parsing(ners['Address'])
except:
pass
try:
name = name_parsing(ners['Name'])
except:
pass
address.update(name)
try:
del address['Recipient']
except:
pass
if 'And' in list(address.keys()):
xx = pd.DataFrame({'Household':xz}, index=[0])
xx['Type'] = 'Household'
return address
xx = full_parsing(text2); xx
tags = pp.LABELS
tags.extend(ua.LABELS)
tags = list(set(tags))
additionals = ['Text', 'AddressType','AddrConfidence','NameType','NameConfidence']
tags.extend(additionals)
dfx = pd.DataFrame(columns=tags)
dfx = dfx.loc[:,~dfx.columns.duplicated()]
for idx, text in tqdm(enumerate(data)):
dfx = dfx.append(full_parsing(text), ignore_index=True)
dfx.loc[idx, 'Text'] = text
sorted_collist = dfx.isna().sum().sort_values().index
dfx = dfx[sorted_collist]
dfx.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 61 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Text 1000 non-null object
1 NameConfidence 994 non-null object
2 NameType 994 non-null object
3 AddrConfidence 989 non-null object
4 AddressType 989 non-null object
5 PlaceName 967 non-null object
6 StateName 967 non-null object
7 AddressNumber 933 non-null object
8 StreetName 926 non-null object
9 StreetNamePostType 864 non-null object
10 ZipCode 857 non-null object
11 GivenName 627 non-null object
12 Surname 589 non-null object
13 CorporationName 388 non-null object
14 MiddleInitial 286 non-null object
15 And 208 non-null object
16 SecondGivenName 181 non-null object
17 CountryName 157 non-null object
18 CorporationLegalType 109 non-null object
19 StreetNamePreDirectional 87 non-null object
20 OccupancyIdentifier 83 non-null object
21 ZipPlus4 80 non-null object
22 OccupancyType 71 non-null object
23 SecondMiddleInitial 63 non-null object
24 MiddleName 59 non-null object
25 USPSBoxType 57 non-null object
26 USPSBoxID 57 non-null object
27 CorporationNameOrganization 56 non-null object
28 SecondSurname 25 non-null object
29 StreetNamePreType 24 non-null object
30 SuffixGenerational 22 non-null object
31 StreetNamePostDirectional 15 non-null object
32 PrefixOther 13 non-null object
33 Recipient 9 non-null object
34 CorporationNameBranchIdentifier 8 non-null object
35 CorporationNameAndCompany 5 non-null object
36 SuffixOther 4 non-null object
37 BuildingName 4 non-null object
38 SubaddressIdentifier 4 non-null object
39 SecondMiddleName 3 non-null object
40 SubaddressType 3 non-null object
41 Nickname 2 non-null object
42 CorporationNameBranchType 2 non-null object
43 FirstInitial 1 non-null object
44 ShortForm 1 non-null object
45 SecondCorporationNameOrganization 1 non-null object
46 AddressNumberSuffix 1 non-null object
47 SecondCorporationName 1 non-null object
48 CorporationCommitteeType 1 non-null object
49 USPSBoxGroupType 0 non-null object
50 IntersectionSeparator 0 non-null object
51 LastInitial 0 non-null object
52 AKA 0 non-null object
53 LandmarkName 0 non-null object
54 StreetNamePreModifier 0 non-null object
55 AddressNumberPrefix 0 non-null object
56 CornerOf 0 non-null object
57 USPSBoxGroupID 0 non-null object
58 ProxyFor 0 non-null object
59 PrefixMarital 0 non-null object
60 NotAddress 0 non-null object
dtypes: object(61)
memory usage: 476.7+ KB
dfx0 = dfx.dropna(how='all').reset_index(drop=True); print(dfx0.info())
dfx1 = dfx.dropna(how='all', axis=1).dropna(how='all').reset_index(drop=True); dfx1.info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 61 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Text 1000 non-null object
1 NameConfidence 994 non-null object
2 NameType 994 non-null object
3 AddrConfidence 989 non-null object
4 AddressType 989 non-null object
5 PlaceName 967 non-null object
6 StateName 967 non-null object
7 AddressNumber 933 non-null object
8 StreetName 926 non-null object
9 StreetNamePostType 864 non-null object
10 ZipCode 857 non-null object
11 GivenName 627 non-null object
12 Surname 589 non-null object
13 CorporationName 388 non-null object
14 MiddleInitial 286 non-null object
15 And 208 non-null object
16 SecondGivenName 181 non-null object
17 CountryName 157 non-null object
18 CorporationLegalType 109 non-null object
19 StreetNamePreDirectional 87 non-null object
20 OccupancyIdentifier 83 non-null object
21 ZipPlus4 80 non-null object
22 OccupancyType 71 non-null object
23 SecondMiddleInitial 63 non-null object
24 MiddleName 59 non-null object
25 USPSBoxType 57 non-null object
26 USPSBoxID 57 non-null object
27 CorporationNameOrganization 56 non-null object
28 SecondSurname 25 non-null object
29 StreetNamePreType 24 non-null object
30 SuffixGenerational 22 non-null object
31 StreetNamePostDirectional 15 non-null object
32 PrefixOther 13 non-null object
33 Recipient 9 non-null object
34 CorporationNameBranchIdentifier 8 non-null object
35 CorporationNameAndCompany 5 non-null object
36 SuffixOther 4 non-null object
37 BuildingName 4 non-null object
38 SubaddressIdentifier 4 non-null object
39 SecondMiddleName 3 non-null object
40 SubaddressType 3 non-null object
41 Nickname 2 non-null object
42 CorporationNameBranchType 2 non-null object
43 FirstInitial 1 non-null object
44 ShortForm 1 non-null object
45 SecondCorporationNameOrganization 1 non-null object
46 AddressNumberSuffix 1 non-null object
47 SecondCorporationName 1 non-null object
48 CorporationCommitteeType 1 non-null object
49 USPSBoxGroupType 0 non-null object
50 IntersectionSeparator 0 non-null object
51 LastInitial 0 non-null object
52 AKA 0 non-null object
53 LandmarkName 0 non-null object
54 StreetNamePreModifier 0 non-null object
55 AddressNumberPrefix 0 non-null object
56 CornerOf 0 non-null object
57 USPSBoxGroupID 0 non-null object
58 ProxyFor 0 non-null object
59 PrefixMarital 0 non-null object
60 NotAddress 0 non-null object
dtypes: object(61)
memory usage: 476.7+ KB
None
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1000 entries, 0 to 999
Data columns (total 49 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 Text 1000 non-null object
1 NameConfidence 994 non-null object
2 NameType 994 non-null object
3 AddrConfidence 989 non-null object
4 AddressType 989 non-null object
5 PlaceName 967 non-null object
6 StateName 967 non-null object
7 AddressNumber 933 non-null object
8 StreetName 926 non-null object
9 StreetNamePostType 864 non-null object
10 ZipCode 857 non-null object
11 GivenName 627 non-null object
12 Surname 589 non-null object
13 CorporationName 388 non-null object
14 MiddleInitial 286 non-null object
15 And 208 non-null object
16 SecondGivenName 181 non-null object
17 CountryName 157 non-null object
18 CorporationLegalType 109 non-null object
19 StreetNamePreDirectional 87 non-null object
20 OccupancyIdentifier 83 non-null object
21 ZipPlus4 80 non-null object
22 OccupancyType 71 non-null object
23 SecondMiddleInitial 63 non-null object
24 MiddleName 59 non-null object
25 USPSBoxType 57 non-null object
26 USPSBoxID 57 non-null object
27 CorporationNameOrganization 56 non-null object
28 SecondSurname 25 non-null object
29 StreetNamePreType 24 non-null object
30 SuffixGenerational 22 non-null object
31 StreetNamePostDirectional 15 non-null object
32 PrefixOther 13 non-null object
33 Recipient 9 non-null object
34 CorporationNameBranchIdentifier 8 non-null object
35 CorporationNameAndCompany 5 non-null object
36 SuffixOther 4 non-null object
37 BuildingName 4 non-null object
38 SubaddressIdentifier 4 non-null object
39 SecondMiddleName 3 non-null object
40 SubaddressType 3 non-null object
41 Nickname 2 non-null object
42 CorporationNameBranchType 2 non-null object
43 FirstInitial 1 non-null object
44 ShortForm 1 non-null object
45 SecondCorporationNameOrganization 1 non-null object
46 AddressNumberSuffix 1 non-null object
47 SecondCorporationName 1 non-null object
48 CorporationCommitteeType 1 non-null object
dtypes: object(49)
memory usage: 382.9+ KB
# dfx0.to_csv('dfx0.csv')
dfx1.to_csv('dfx1.csv')
def func(text):
text = str(text)
output = full_parsing(text)
return str(output)
txt = '<redacted>'
func(txt)
!pip install -q gradio
import gradio as gr
gr.Interface(fn=func, inputs="text", outputs="text").launch()