def test_empty_init(self):
"Test the default Observations creation"
# 1. Create default Device object
myobj = observations.Observations()
# 2. Make sure it has the default values
self.assertEqual(myobj.text, None)
self.assertEqual(len(myobj.data), 0)
def __init__(self, text=None, part2=False):
# 1. Set the initial values
self.part2 = part2
self.text = text
self.obs = None
self.program = None
self.opcodes = None
# 2. Process text (if any)
if text is not None:
self.obs = observations.Observations(text=text)
self.program = self.read_program(text=text)
def test_text_init(self):
"Test the Observation object creation from text"
# 1. Create Device object from text
myobj = observations.Observations(
text=aoc_16.from_text(EXAMPLE_TEXT_MULTIPLE))
# 2. Make sure it has the expected values
self.assertEqual(len(myobj.text), 55)
self.assertEqual(len(myobj.data), 13)
self.assertEqual(myobj.data[0].before, [3, 2, 1, 1])
self.assertEqual(myobj.data[0].instruction, [9, 2, 1, 2])
self.assertEqual(myobj.data[0].after, [3, 2, 2, 1])
def test_opcodes(self):
"Test determining opcodes"
# 1. Create Device object from text
myobj = observations.Observations(
text=aoc_16.from_text(EXAMPLE_TEXT_MULTIPLE))
# 2. Make sure it has the expected values
self.assertEqual(len(myobj.text), 55)
self.assertEqual(len(myobj.data), 13)
# 3. Process the observations
myobj.process_observations(device.INSTS)
self.assertEqual(myobj.data[0].names, set(['seti', 'addi', 'mulr']))
self.assertEqual(
myobj.data[1].names,
set([
'bani', 'eqri', 'gtrr', 'muli', 'borr', 'bori', 'addr', 'setr',
'gtir'
]))
self.assertEqual(myobj.data[2].names,
set(['seti', 'banr', 'borr', 'setr', 'bani', 'bori']))
# 4. Evaluate the opcodes
self.assertEqual(myobj.determine_opcodes(device.INSTS)[15], None)
def obs_single_star_job(tn, obsfile='Tests/cl_0.fits', outdir='Tests/cl_0'):
filter_names = 'hst_wfc3_f275w hst_wfc3_f336w hst_acs_wfc_f475w hst_acs_wfc_f814w hst_wfc3_f110w hst_wfc3_f160w'.upper(
).split()
#Load the initial model grid
g0 = grid.FileSpectralGrid(
'libs/stellib_kurucz2004_padovaiso.spectralgrid.fits')
#load the filters
filts = phot.load_filters(filter_names, interp=True, lamb=g0.lamb)
# define the extinction priors
oAv = extinction.Cardelli()
#Observations
obs = observations.Observations(obsfile)
obs.setFilters(filter_names)
##define Av with a step size
#Av = numpy.arange(0.,3., 0.1)
##define Av with a number of points
Av = numpy.linspace(0, 1, 3)
#Av = numpy.array([0.0], dtype=float)
Rv = numpy.array([3.1], dtype=float)
fb = numpy.array(
[0.5],
dtype=float) # only one value still needs to be array/list/tuple
#get the grid iterator
iter_grid = iter_Av_grid(g0, oAv, Av=Av, Rv=Rv, f_bump=fb)
# Parameters from which we generate fake data
idx = tn
fakesed, fakeerr, mask = obs.getObs(idx)
with timeit('Likelihood Object %d' % tn):
#define output filename
outname = outdir + '/star_%d.fits' % (tn)
for gk, thetak in iter_grid:
with timeit('\t * Computing SEDS'):
seds = gk.getSEDs(filts, absFlux=True)
with timeit('\t * Computing Lnp'):
lnp = computeLogLikelihood(fakesed,
fakeerr,
seds.seds,
normed=False,
mask=mask)
with timeit('\t * Writing outputs'):
t = Table(name='SEDOUT')
#t = gk.grid
t.addCol(numpy.arange(seds.grid.nrows, dtype=int), name='idx')
t.addCol(lnp, name='lnp')
for ek, nk in enumerate(filter_names):
t.addCol(seds.seds[:, ek], name=nk)
t.header['Av'] = thetak[0]
t.header['Rv'] = thetak[1]
t.header['f_bump'] = thetak[2]
t.write(outname)
with timeit('\t * Writing Original grid'):
g0.grid.header['NAME'] = 'MODELS'
g0.grid.header['EXTNAME'] = 'MODELS'
g0.grid.write(outname)
def getFakeCluster(g,
age,
Z,
filts,
err=0.1,
suffix='0',
Nstars=None,
oAv=None,
**kwargs):
""" Generate a fake sed from a model grid
INPUTS:
idx int index number on the grid
age float age of the desired population
Z float metallicity of the desired population
filters list[filter] list of filter names
OUTPUTS:
fakein int the index of the model on the grid
lamb ndarray[float, ndim=1] wavelength taken from the grid
fakesed ndarray[float, ndim=1] resulting SED
KEYWORDS:
suffix str suffix used to store the cluster table
err float proportional error to consider on the fluxes
Nstars int number of stars in the population (None = all possible stars)
oAv ExtinctionLaw if provided, apply extinction function using **kwargs
**kwargs if provided, extra keywords are used to apply an extinction
"""
# find all the stars matching age and Z criteria
glogA = numpy.unique(g.logA)
gZ = numpy.unique(g.Z)
_logA = glogA[numpy.argmin(abs(glogA - numpy.log10(age)))]
_Z = gZ[numpy.argmin(abs(gZ - Z))]
idx = numpy.where((g.logA == _logA) & (g.Z == _Z))[0]
# Restrict to Nstars is provided
if Nstars is not None:
idx = numpy.random.randint(0, len(idx), min([len(idx), Nstars]))
fakeseds = g.seds[idx]
# Compute extinction is requested
if (oAv is not None) & (len(kwargs) > 0):
tau = oAv(g.lamb * 1e-4, Alambda=True, **kwargs)
fakeseds *= exp(-tau)[None, :]
## extract photometry
gg = grid.SpectralGrid()
gg.lamb = g.lamb
gg.seds = fakeseds
gg.grid = g.grid[idx]
seds = gg.getSEDs(filts, absFlux=True)
seds.grid.addCol(idx, name='idx')
for k, v in kwargs.iteritems():
seds.grid.header[k] = v
filter_names = []
for ek, nk in enumerate(filts):
fname = nk.name
seds.grid.addCol(seds.seds[:, ek], name=fname)
seds.grid.addCol(seds.seds[:, ek] * err, name=fname + 'err')
filter_names.append(nk.name)
seds.grid.header['logA'] = _logA
seds.grid.header['Z'] = _Z
seds.grid.write('Tests/cl_%s.fits' % suffix)
del seds
obs = observations.Observations('Tests/cl_%s.fits' % suffix)
obs.setFilters(filter_names)
return obs
ACTIONS_COUNT = 3
MINI_BATCH_SIZE = 128
SAVE_CHECK_EVERY_X = 5000 # Save checkpoint every 5000 steps
PRINT_EVERY_X = 2000
parser = argparse.ArgumentParser()
parser.add_argument("name", help="The name for the observations and checkpoint files");
parser.add_argument("-n", "--nsteps", type=int,
dest="nsteps", help="The number of steps of training to be performed", default=20000)
args = parser.parse_args()
obsfilename = args.name + ".obs"
chkfilename = "./" + args.name
print("Loading observations from {}".format(obsfilename))
obs = observations.Observations()
obs.loadFromFile(args.name + ".obs")
print("There are {} observations".format(len(obs)))
session = tf.Session()
input_layer, output_layer, action, target, train_operation = nnetwork.create_network()
session.run(tf.global_variables_initializer())
