notebook.community

Edit and run

Introduction to iPython

Robert D. French
User Support Specialist
Oak Ridge Leadership Computing Facility
Email: frenchrd@ornl.gov

  • MATLAB / Mathematica style interactive IDE
  • Good for prototyping / verifying new codes
  • Good for analyzing output from HPC codes
    • Easy plots & charts
    • Basic stats and data management capabilities




In [100]:



    


x = linspace(-10, 10., 1000)
plot(x, sin(x))



    


















    
Out[100]:








[<matplotlib.lines.Line2D at 0xaf36ef0>]










    
























In [101]:



    


bins = [n/10.0 for n in range(0,11)]
hist(rand(100),bins);

Setup

  • All-in-one Distributions are easiest:
    • Canopy by Enthought (Good for OS X)
    • Linux distros, check repo for iPython

Notebooks

  • Cell-based programming makes learning easy
  • Distribute content to students / colleagues easily
  • Display graphics and results inline
  • Supports Markdown for text styling
  • Notebooks are just JSON documents!




In [59]:



    


import json
notebook = json.load(open("/Users/rf9/Projects/iPython-Molecular-Dynamics/presentation.ipynb"))
notebook["worksheets"][0]["cells"][4]



    


















    
Out[59]:








{u'cell_type': u'markdown',
 u'metadata': {},
 u'source': [u'# Notebooks\n',
  u'\n',
  u'* Cell-based programming makes learning easy\n',
  u'* Distribute content to students / colleagues easily\n',
  u'* Display graphics and results inline\n',
  u'* Supports [Markdown](http://daringfireball.net/projects/markdown/) for text styling\n',
  u'* Notebooks are just JSON documents!']}

Vector Arithmetic

Vectors can be created by passing a python list to the array() function





In [43]:



    


x = array([1,2,3])
x



    


















    
Out[43]:








array([1, 2, 3])

Vector addition is now supported via the + operator





In [64]:



    


y = array([1,0,1])
x + y



    


















    
Out[64]:








array([2, 2, 4])

Standard tools from linear algebra such as Euclidean inner products and norms





In [65]:



    


x.dot(y)



    


















    
Out[65]:








4

















In [66]:



    


numpy.linalg.norm(x - y)



    


















    
Out[66]:








2.8284271247461903

Scipy functions are automatically applied to each element in the list:





In [36]:



    


sin(x)**2



    


















    
Out[36]:








array([ 0.70807342,  0.82682181,  0.01991486])

This includes any operation of scalar arithmetic, and extends to everyone's favorite trig identity:





In [37]:



    


sin(x)**2 + cos(x)**2



    


















    
Out[37]:








array([ 1.,  1.,  1.])

Example Use Case: Verifying a new MD Code

  • Prototyped a simple MD system in Python (available here)
  • Want to verify that output is correct
  • We can use the Pandas library to analyze the data

Example 1: Particle B drags Particle A

  • 2 particle system, simple spring potential
  • Particle A has zero initial velocity
  • Particle B has small initial velocity along y-axis
  • Particle B should "drag" Particle A

Step 1: Verify the input file





In [69]:



    


input_config = json.load(open("/Users/rf9/Projects/iPython-Molecular-Dynamics/ExampleInput.dragging.json"))
input_config["particles"]



    


















    
Out[69]:








[{u'mass': 1, u'position': [0, 0, 0], u'velocity': [0.0, 0.0, 0.0]},
 {u'mass': 1, u'position': [0, 1, 0], u'velocity': [0.0, 0.5, 0.0]}]

Step 2: Verify the structure of the output file





In [74]:



    


import pandas
output = pandas.read_csv("/Users/rf9/Projects/iPython-Molecular-Dynamics/dragging_output.csv")
output.head()



    


















    
Out[74]:










  
    

      
      time
      x1
      y1
      z1
      x2
      y2
      z2
    

  
  
    

      0
       0.00
       0
       0.000000
       0
       0
       1.000000
       0
    

    

      1
       0.01
       0
       0.000000
       0
       0
       1.005000
       0
    

    

      2
       0.02
       0
       0.000000
       0
       0
       1.009999
       0
    

    

      3
       0.03
       0
       0.000002
       0
       0
       1.014998
       0
    

    

      4
       0.04
       0
       0.000005
       0
       0
       1.019995
       0

Step 3: Visualize the output data





In [90]:



    


plot(output["time"],output["y1"], label='Particle A')
plot(output["time"],output["y2"], label='Particle B')
legend(loc='upper left')



    


















    
Out[90]:








<matplotlib.legend.Legend at 0xdb0ee70>

Example 2: Particles Oscillate "in place"

  • 2 particle system, simple spring potential
  • Particles have same initial velocity, but in opposite directions
  • Particles should oscillate with no net change in position

Step 1: Verify Input data





In [91]:



    


input_config = json.load(open("/Users/rf9/Projects/iPython-Molecular-Dynamics/ExampleInput.oscillation.json"))
input_config["particles"]



    


















    
Out[91]:








[{u'mass': 1, u'position': [0, 0, 0], u'velocity': [0.0, -0.5, 0.0]},
 {u'mass': 1, u'position': [0, 1, 0], u'velocity': [0.0, 0.5, 0.0]}]

Step 2: Verify structure of output data





In [92]:



    


output = pandas.read_csv("/Users/rf9/Projects/iPython-Molecular-Dynamics/oscillation_output.csv")
output.head()



    


















    
Out[92]:










  
    

      
      time
      x1
      y1
      z1
      x2
      y2
      z2
    

  
  
    

      0
       0.00
       0
       0.000000
       0
       0
       1.000000
       0
    

    

      1
       0.01
       0
      -0.005000
       0
       0
       1.005000
       0
    

    

      2
       0.02
       0
      -0.009999
       0
       0
       1.009999
       0
    

    

      3
       0.03
       0
      -0.014996
       0
       0
       1.014996
       0
    

    

      4
       0.04
       0
      -0.019990
       0
       0
       1.019990
       0

Step 3: Visualize data





In [93]:



    


plot(output["time"],output["y1"], label='Particle A')
plot(output["time"],output["y2"], label='Particle B')
legend(loc='upper left')



    


















    
Out[93]:








<matplotlib.legend.Legend at 0xdbcfbf0>

Step 4: Calculate difference between average position and initial position





In [98]:



    


mean(output["y1"]) - input_config["particles"][0]["position"][1]



    


















    
Out[98]:








-0.025606218562874289

















In [99]:



    


mean(output["y2"]) - input_config["particles"][1]["position"][1]



    


















    
Out[99]:








0.025606218562873595

















In [ ]:

Content source: olcf/iPython-Molecular-Dynamics

Similar notebooks:

notebook.community | gallery | about