How to execute prophet functions using multi-processing? - time-series

I have been generating a few forecast with a small set of 100 products (e.g 100 Aliases), but I want to scale to 20k Aliases. Currently in a anaconda environment using default configuration (1 core). Forecast function has been running for 2 days and I have processes like 4k Aliases. How would I implement multiprocessing with prophet. I think that would help reduce processing time (if I understand correctly?). I currently have a machine with 32 gb of ram, and Intel Xeon(R) CPU E5-2697 v4 (18 cores). Any help would be amazing....
import warnings;
warnings.simplefilter('ignore')
!pip install pystan
!pip install fbprophet
import pandas as pd
from fbprophet import Prophet
import os
from pandas_visual_analysis import VisualAnalysis
import pyodbc
from datetime import datetime, timedelta, date
import psycopg2
from distributed import Client, performance_report
import multiprocessing as mp
from multiprocessing import Process, current_process
#organized Alias in groups
x = df_group_final.groupby('Alias')
#create empty dataframe
y = pd.DataFrame()
def forecast(Aliases, target):
#for loop to pick all Aliases and loop through 1 at a time
for Alias in Aliases.groups:
#assigns group variable and get all data for that group (aka Alias)
group = Aliases.get_group(Alias)
#initialized model with 95% confidence interval.
m = Prophet(interval_width=0.95, seasonality_mode='multiplicative')
#fit model based on past data by each group at a time
m.fit(group)
#based on model predict 18 months in future
future = m.make_future_dataframe(periods=18, freq='m')
forecast = m.predict(future)
#m.plot(forecast)
#creates forecast for ALias
forecast = forecast.rename(columns={'yhat':'yhat_'+Alias})
target = pd.merge(target, forecast.set_index('ds'), how='outer', left_index=True, right_index=True)
forecast(x,y)

Related

Reasons why swifter/dask/ray only use one core for an apply task?

I have this function that I would like to apply to a large dataframe in parallel:
from rdkit import Chem
from rdkit.Chem.MolStandardize import rdMolStandardize
from rdkit import RDLogger
RDLogger.DisableLog('rdApp.*')
def standardize_smiles(smiles):
if smiles is None:
return None
try:
mol = Chem.MolFromSmiles(smiles)
# removeHs, disconnect metal atoms, normalize the molecule, reionize the molecule
clean_mol = rdMolStandardize.Cleanup(mol)
# if many fragments, get the "parent" (the actual mol we are interested in)
parent_clean_mol = rdMolStandardize.FragmentParent(clean_mol)
# try to neutralize molecule
uncharger = rdMolStandardize.Uncharger() # annoying, but necessary as no convenience method exists
uncharged_parent_clean_mol = uncharger.uncharge(parent_clean_mol)
# note that no attempt is made at reionization at this step
# nor at ionization at some pH (rdkit has no pKa caculator)
# the main aim to to represent all molecules from different sources
# in a (single) standard way, for use in ML, catalogue, etc.
te = rdMolStandardize.TautomerEnumerator() # idem
taut_uncharged_parent_clean_mol = te.Canonicalize(uncharged_parent_clean_mol)
return Chem.MolToSmiles(taut_uncharged_parent_clean_mol)
#except:
# return False
standardize_smiles('CCC')
'CCC'
However, neither Dask, nor Swifter, nor Ray can do the job. All frameworks use a single CPU for some reason.
Native Pandas
import pandas as pd
N = 1000
smilest_test = pd.DataFrame({'smiles': ['CCC']*N})
smilest_test
CPU times: user 3.58 s, sys: 0 ns, total: 3.58 s
Wall time: 3.58 s
Swifter 1.3.4
smiles_test['standardized_siles'] = smiles_test.smiles.swifter.allow_dask_on_strings(True).apply(standardize_smiles)
CPU times: user 892 ms, sys: 31.4 ms, total: 923 ms
Wall time: 5.14 s
While this WORKS with the dummy data, it does not with the real data, which looks like this:
The strings are a bit more complicated than the ones in the dummy data.
it seems first swifter needs some time to prepare the parallel execution and only uses one core, but then uses more cores. However, for the real data, it only uses 3 out of 8 cores.
I have the same issue with other frameworks such as dask, ray, modin, swifter.
Is there something that I miss here? Is there a problem when the dataframe contains stings? Why does the parallel execution take so much time even on a single computer (with multiple cores)? Or is there an issue with the RDKit library that I am using that makes it difficult to parallelize the above function?

Pandas time series index attribute error when using TsTables & PyTables in creating a table class

I am trying to create a table structure through tb.IsDescription class, then create a .h5 file and populate it with a Pandas Dataframe with Datetime index, using TsTables package. I have already tested the Dataframe and the date time Indexing and both seem to be fine. I believe the issue is with the TsTable package, as it remains 'Unused import statement'. The error I get is: " AttributeError: module 'pandas.tseries' has no attribute 'index' ". The reason I am using the TsTAble is that I have heard it is faster than other modules. Any suggestions how to resolve this issue, or any substitute method?
import numpy as np
import pandas as pd
import tables as tb
import datetime as dt
path = r'C:\Users\--------\PycharmProjects\pythonProject2'
no = 5000000 # number of time steps
co = 3 # number of time series
interval = 1. / (12 * 30 * 24 * 60) # the time interval as a year fraction
vol = 0.2 # volatility
rn = np.random.standard_normal((no, co))
rn[0] = 0.0 # sets the initial random numbers to zero
paths = 100 * np.exp(np.cumsum(-0.5 * vol ** 2 * interval + vol * np.sqrt(interval) * rn, axis=0))
# simulation based on an Euler discretization
paths[0] = 100 # Sets the initial values of the paths to 100
dr = pd.date_range('2019-1-1', periods=no, freq='1s')
print(dr[-6:]) # the date range appears fine
df = pd.DataFrame(paths, index=dr, columns=['ts1', 'ts2', 'ts3'])
print(df.info(verbose=True)) # df is pandas Dataframe and appears fine
print(df.head()) # tested a fraction of the data, it is fine
import tstables as tstab # I get Unused import statement
class ts_desc(tb.IsDescription):
timestamp = tb.Int64Col(pos=0) # The column for the timestamps
ts1 = tb.Float64Col(pos=1) # The column to store numerical data
ts2 = tb.Float64Col(pos=2)
ts3 = tb.Float64Col(pos=3)
h5 = tb.open_file(path + 'tstab.h5', 'w')
ts = h5.create_ts('/', 'ts', ts_desc)
ts.append(df) # !!!!! the error I get is from this code line !!!!
# value error raised is: if rows.index.__class__ != pandas.tseries.index.DatetimeIndex:
AttributeError: module 'pandas.tseries' has no attribute 'index' `
I suspect you have run into a version compatibility issue between tstables and your pandas versions (assuming you are running any recent pandas version). Based on the tstables PyPI page, the last release of tstables was in 2015. A check of the tstables github project page shows there was an issue with Pandas 0.20.3 and use of datetime. The error message is the same as yours: module 'pandas.tseries' has no attribute 'index' in tstables See this: tstables breaks down with Pandas 20.3
The issue has a link to another build that works with Pandas 0.20.3. Development notes state "Removed 'convert_datetime64' parameter on line 245". Not sure if it will work with more recent versions, but worth a try. See this: schwed2 tstables build
If that doesn't solve the problem, I suggest running the simple examples provided or run the setup tests. (Note: I could not find the bpi_2014_01.csv file to test the bitcoin/bpi example.)
Good luck.

I'm using Dask to apply LabelingFunction using Snorkel on multiple datasets but it seems to take forever. Is this normal?

My problem is as follow:
I have several datasets (900K, 1M7 and 1M7 entries) in csv format which I load into multiple Dask Dataframe.
Then I concatenate them all in one Dask Dataframe that I can feed to my Snorkel Applier, which applies a bunch of Labeling Function to each row of my Dataframe and return a numpy array with as many rows as there are in the Dataframe and as many columns as there are Labeling Functions.
The call to Snorkel Applier seems to take forever when I do that with 3 datasets (more than 2 days...). However if I just run the code with only the first dataset, the call takes around 2 hours. Of course I don't do the concatenation step.
So I was wondering how can this be ? Should I change the number of partitions in the concatenated Dataframe ? Or maybe I'm using Dask badly in the first place ?
Here is the code I'm using:
from snorkel.labeling.apply.dask import DaskLFApplier
import dask.dataframe as dd
import numpy as np
import os
start = time.time()
applier = DaskLFApplier(lfs) # lfs are the function that are going to be applied, one of them featurize one of the column of my Dataframe and apply a sklearn classifier (I put n_jobs to None when loading the model)
# If I have only one CSV to read
if isinstance(PATH_TO_CSV, str):
training_data = dd.read_csv(PATH_TO_CSV, lineterminator=os.linesep, na_filter=False, dtype={'size': 'int32'})
slices = None
# If I have several CSV
elif isinstance(PATH_TO_CSV, list):
training_data_list = [dd.read_csv(path, lineterminator=os.linesep, na_filter=False, dtype={'size': 'int32'}) for path in PATH_TO_CSV]
training_data = dd.concat(training_data_list, axis=0)
# some useful things I do to know where to slice my final result and be sure I can assign each part to each dataset
df_sizes = [len(df) for df in training_data_list]
cut_idx = np.insert(np.cumsum(df_sizes), 0, 0)
slices = list(zip(cut_idx[:-1], cut_idx[1:]))
# The call that lasts forever: I tested all the code above without that line on my 3 datasets and it runs perfectly fine
L_train = applier.apply(training_data)
end = time.time()
print('Time elapsed: {}'.format(timedelta(seconds=end-start)))
If you need more info I will try to get them to you as much as I can.
Thank in you advance for your help :)
It seems that by default applier function is using processes, so does not benefit from additional workers you might have available:
# add this to the beginning of your code
from dask.distributed import Client
client = Client()
# you can see the address of the client by typing `client` and opening the dashboard
# skipping your other code
# you need to pass the client explicitly to the applier
# after launching this open the dashboard and watch the workers work :)
L_train = applier.apply(training_data, scheduler=client)

Troubles using dask distributed with datashader: 'can't pickle weakref objects'

I'm working with datashader and dask but I'm having problems when trying to plot with a cluster running. To make it more concrete, I have the following example (embedded in a bokeh plot):
import holoviews as hv
import pandas as pd
import dask.dataframe as dd
import numpy as np
from holoviews.operation.datashader import datashade
import datashader.transfer_functions as tf
#initialize the client/cluster
cluster = LocalCluster(n_workers=4, threads_per_worker=1)
dask_client = Client(cluster)
def datashade_plot():
hv.extension('bokeh')
#create some random data (in the actual code this is a parquet file with millions of rows, this is just an example)
delta = 1/1000
x = np.arange(0, 1, delta)
y = np.cumsum(np.sqrt(delta)*np.random.normal(size=len(x)))
df = pd.DataFrame({'X':x, 'Y':y})
#create dask dataframe
points_dd = dd.from_pandas(df, npartitions=3)
#create plot
points = hv.Curve(points_dd)
return hd.datashade(points)
dask_client.submit(datashade_plot,).result()
This raises a:
TypeError: can't pickle weakref objects
I have the theory that this happens because you can't distribute the datashade operations in the cluster. Sorry if this is a noob question, I'd be very grateful for any advice you could give me.
I think you want to go the other way. That is, pass datashader a dask dataframe instead of a pandas dataframe:
>>> from dask import dataframe as dd
>>> import multiprocessing as mp
>>> dask_df = dd.from_pandas(df, npartitions=mp.cpu_count())
>>> dask_df.persist()
...
>>> cvs = datashader.Canvas(...)
>>> agg = cvs.points(dask_df, ...)
XREF: https://datashader.org/user_guide/Performance.html

dask.ml.xgboost raises UnboundLocalError: local variable 'result' referenced before assignment

I am using dask_xgboost and I don't understand the error stated in the subject. I have successfully trained a model and saved it with joblib.dump.
Later on, during the prediction step I use it like this:
import dask
import dask.dataframe as dd
import dask.distributed as ddst
from dask_jobqueue import PBSCluster
from dask.distributed import Client
import dask_xgboost as dxgb
import geopandas as gp
from sklearn.externals import joblib
def predict(zs_files: List[str], model_name: str, client) -> None:
delayed_dfs = [dask.delayed(gp.read_file)(zsf) for zsf in zs_files]
model = joblib.load(model_name)
delayed_predictions = [
dxgb.predict(client, model, df).to_parquet(f"{fn}_predicted.parquet")
for df, fn in zip(delayed_dfs, zs_files)
]
delayed_predictions.compute()
I read a set of GeoJSON files with geopandas and then just feed the model with them. I am using a client on a PBS cluster.
Any help would be appreciated.
Thanks.
I found the issue. I wass missing a from_delayed call to transform the geopandas dataframe to a dask one:
dxgb.predict(client, model, dd.from_delayed(df))

Resources