def __init__(self, z, base_dir):
self.z = z
self.base_dir = base_dir
A10 = 2.85e-15 # s^-1
kb = 1.381e-23 # m^2 kg s^-2 K^-1
# 21cmGEM code
h = 0.6704
H0 = 100 * h * 1000 / 3.089e22
omega_c = 0.12038 / h**2
omega_b = 0.022032 / h**2
omega_m = omega_b + omega_c
T_cmb0 = 2.725 # K
planck_h = 6.626e-34 # m^2 kg s^-1
c = 3e8 # m/s
self.T_r = T_cmb0 * (1 + self.z)
z_ref = 40
T_K_ref = 33.7340 # K
self.T_K = T_K_ref * ((1 + self.z) / (1 + z_ref))**2
Y = 0.274 # Helium abundance by mass
rhoc = 1.36e11 * (h / 0.7)**2 # M_sol/cMpc^3
mp = 8.40969762e-58 # m_p in M_sol
nH = (rhoc / mp) * (1 - Y) * omega_b * (1 + self.z)**3 * 3.40368e-68
Tstar = 0.068 # K
try:
t, kappa10_HH_data = np.loadtxt('kappa_HH.txt', unpack=True)
except OSError:
download().kappa()
t, kappa10_HH_data = np.loadtxt('kappa_HH.txt', unpack=True)
kappa10_HH = np.interp(self.T_K, t, kappa10_HH_data)
xc = (nH * kappa10_HH * 1e-6 * Tstar) / (A10 * self.T_r)
invT_s = (1 / self.T_r + xc * (1 / self.T_K)) / (1 + xc)
self.T_s = 1 / invT_s
xHI = 1
nu0 = 1420.4e6
Hz = H0 * np.sqrt(omega_m * (1 + self.z)**3)
tau = (3*planck_h*c**3*A10*xHI*nH) / \
(32*np.pi*kb*self.T_s*nu0**2*(1+self.z)*Hz/(1+self.z))
deltaT = (self.T_s - self.T_r) / (1 + self.z) * (1 - np.exp(-tau))
norm_factor = np.load(self.base_dir + 'AFB_norm_factor.npy')
self.deltaT = deltaT / np.abs(deltaT).max() * np.abs(norm_factor) * 1e3
np.savetxt(self.base_dir + 'AFB.txt', self.deltaT)
def test_existing_dir():
if os.path.exists('kappa_HH.txt'):
os.remove('kappa_HH.txt')
download().kappa()
with pytest.raises(TypeError):
download(xHI=2).kappa()
# for use in later tests...
download_21cmGEM_data()
def test_existing_dir():
if os.path.exists('T_release/'):
shutil.rmtree('T_release/')
if os.path.exists('xHI_release/'):
shutil.rmtree('xHI_release/')
download().model()
download(xHI=True).model()
predictor = evaluate(z=originalz, base_dir='T_release/', gc=True)
signal, z = predictor(params)
assert (len(signal.shape) == 1)
params_array = []
for i in range(10):
params_array.append(np.random.uniform(0, 1, 7))
params_array = np.array(params_array)
data_max = np.loadtxt('T_release/data_maxs.txt')
data_min = np.loadtxt('T_release/data_mins.txt')
for i in range(params_array.shape[1]):
params_array[:, i] = params_array[:, i] * \
(data_max[i] - data_min[i]) + data_min[i]
signal, z = predictor(params_array)
assert (len(signal.shape) == 2)
predictor = evaluate(z=10, base_dir='T_release/')
signal, z = predictor(params)
preprocess = predictor.preprocess_settings
assert (preprocess['AFB'] is True)
assert (preprocess['resampling'] is True)
assert (preprocess['std_division'] is True)
predictor = evaluate(z=originalz, base_dir='xHI_release/')
signal, z = predictor(params)
with pytest.raises(KeyError):
evaluate(z=originalz, basedir='T_release/')
with pytest.raises(TypeError):
predictor = evaluate(z=originalz, base_dir='T_release/')
predictor(10)
with pytest.raises(TypeError):
predictor = evaluate(z=originalz, base_dir=100)
with pytest.raises(TypeError):
predictor = evaluate(z=originalz, base_dir='T_release/', logs=10)
with pytest.raises(TypeError):
predictor = evaluate(z='foo', base_dir='T_release/')
with pytest.raises(TypeError):
predictor = evaluate(z=originalz, base_dir='T_release/', gc='false')
def test_config():
if os.path.exists('T_release/'):
shutil.rmtree('T_release/')
if os.path.exists('xHI_release/'):
shutil.rmtree('xHI_release/')
download().model()
download(xHI=True).model()
paramnames = [
r'$\log(f_*)$', r'$\log(V_c)$', r'$\log(f_X)$', r'$\nu_\mathrm{min}$',
r'$\tau$', r'$\alpha$', r'$R_\mathrm{mfp}$'
]
# Providing this with global signal data as neutral fraction data is
# not publicly available. Will not effect efficacy of the test.
config('xHI_release/', paramnames, '21cmGEM_data/', logs=[0, 1, 2])
assert (os.path.exists('xHI_release/gui_configuration.csv') is True)
res = pd.read_csv('xHI_release/gui_configuration.csv')
logs = res['logs'].tolist()
logs = [int(x) for x in logs if x != '--']
assert (logs == [0, 1, 2])
paramnames = [
r'$\log(f_*)$', r'$\log(V_c)$', r'$\log(f_X)$', r'$\tau$', r'$\alpha$',
r'$\nu_\mathrm{min}$', r'$R_\mathrm{mfp}$'
]
config('T_release/', paramnames, '21cmGEM_data/')
assert (os.path.exists('T_release/gui_configuration.csv') is True)
with pytest.raises(KeyError):
config('T_release', paramnames, '21cmGEM_data/')
with pytest.raises(TypeError):
config(10, paramnames, '21cmGEM_data/')
with pytest.raises(KeyError):
config('T_release/', paramnames, '21cmGEM_data')
with pytest.raises(TypeError):
config('T_release/', paramnames, '21cmGEM_data/', logs='banana')
with pytest.raises(TypeError):
config('T_release/', 5, '21cmGEM_data/')
with pytest.raises(KeyError):
config('T_release/', paramnames, '21cmGEM_data/', color='C0')
def test_existing_dir():
if os.path.exists('T_release/'):
shutil.rmtree('T_release/')
if os.path.exists('xHI_release/'):
shutil.rmtree('xHI_release/')
download().model()
download(xHI=True).model()
predictor = evaluate(base_dir='T_release/')
parameters = np.loadtxt('21cmGEM_data/test_data.txt')
labels = np.loadtxt('21cmGEM_data/test_labels.txt')
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
figsizex=3.15,
figsizey=6,
loss_label='RMSE = {:.4f} [mK]')
assert (os.path.exists('T_release/eval_plot.pdf') is True)
with pytest.raises(KeyError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
figsize_x=3.15)
with pytest.raises(TypeError):
signal_plot(parameters, labels, 'rmse', predictor, 'T_release')
with pytest.raises(TypeError):
signal_plot(1, labels, 'rmse', predictor, 'T_release/')
with pytest.raises(TypeError):
signal_plot(parameters, labels, 'banana', predictor, 'T_release/')
with pytest.raises(TypeError):
signal_plot(parameters, 6, 'rmse', predictor, 'T_release/')
with pytest.raises(TypeError):
signal_plot(parameters, labels, 'rmse', 'foobar', 'T_release/')
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
rtol='banana')
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
atol='banana')
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
figsizex='banana')
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
figsizey='banana')
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
loss_label=60)
with pytest.raises(TypeError):
signal_plot(parameters,
labels,
'rmse',
predictor,
'T_release/',
xHI=60)
def loss_func(labels, signals):
return np.mean(np.square(labels - signals)) * 1000
signal_plot(parameters, labels, loss_func, predictor, 'T_release/')
assert (os.path.exists('T_release/eval_plot.pdf') is True)
signal_plot(parameters, labels, 'mse', predictor, 'T_release/')
signal_plot(parameters, labels, 'GEMLoss', predictor, 'T_release/')