train_size = 2000
decay = 0.0
hack_trivial = False
epochs = 3000
lr = 8e-4
# How many solutions I want to generate
n_draws = 10
# How many times I want to fit a single trajectory, getting the best result
n_trials = 3
fit_epochs = 1000
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50)
model_name = 'i_0={}_r_0={}_betas={}_gammas={}.pt'.format(i_0_set, r_0_set,
betas,
gammas)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/{}'.format(model_name))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
writer = SummaryWriter(
'runs/' + '{}'.format(model_name))
sir, train_losses, run_time, optimizer = train_bundle(sir, initial_conditions_set, t_final=t_final,
import matplotlib.pyplot as plt
if __name__ == '__main__':
source_s_0_bundle = [0.995, 1.0]
source_beta_bundle = [0.72, 0.88]
source_gamma_bundle = [0.2, 0.35]
source_sigma = 0.0
initial_conditions_set = []
t_0 = 0
t_final = 20
initial_conditions_set.append(t_0)
initial_conditions_set.append(source_s_0_bundle)
# Init model
sir = SIRNetwork(input=4, layers=4, hidden=50)
lr = 8e-4
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(ROOT_DIR + '/models/SIR_bundle_total/b_s_0={}'
'_betas={}_gammas={}_noise_{}.pt'.format(
source_s_0_bundle, source_beta_bundle,
source_gamma_bundle, source_sigma))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
source_epochs = 5000
source_hack_trivial = False
source_train_size = 2500
source_decay = 1e-4
initial_conditions_set.append(t_0)
initial_conditions_set.append(s_0_bundle)
# Sanity check
assert 0 not in beta_bundle and 0 not in gamma_bundle
# Model parameters
train_size = 2000
decay = 0.0
hack_trivial = False
epochs = 3000
lr = 8e-4
sigma = 0.0
# Init model
sir = SIRNetwork(input=4, layers=4, hidden=50)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/b_s_0={}'
'_betas={}_gammas={}_noise_{}.pt'.format(s_0_bundle, beta_bundle,
gamma_bundle, sigma))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
writer = SummaryWriter(
'runs/' + 'b_s_0={}'
'_betas={}_gammas={}_noise_{}.pt'.format(s_0_bundle, beta_bundle,
gamma_bundle, sigma))
sir, train_losses, run_time, optimizer = train_bundle_total(
initial_conditions_set.append(i_0_set)
initial_conditions_set.append(r_0_set)
# How many times I want to fit the trajectory, getting the best result
n_trials = 1
fit_epochs = 300
# Model parameters
train_size = 2000
decay = 1e-4
hack_trivial = False
epochs = 3000
lr = 8e-4
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50)
model_name = 'i_0={}_r_0={}_betas={}_gammas={}.pt'.format(i_0_set, r_0_set,
betas,
gammas)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/{}'.format(model_name))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
writer = SummaryWriter('runs/{}'.format(model_name))
sir, train_losses, run_time, optimizer = train_bundle(sir, initial_conditions_set, t_final=t_final,
epochs=epochs,
if __name__ == '__main__':
# File to apply finetuning on a pretrained model
source_i_0_set = [0.1, 0.2]
source_r_0_set = [0.1, 0.2]
source_betas = [0.6, 0.8]
source_gammas = [0.1, 0.2]
initial_conditions_set = []
t_0 = 0
t_final = 20
initial_conditions_set.append(t_0)
initial_conditions_set.append(source_i_0_set)
initial_conditions_set.append(source_r_0_set)
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50)
lr = 8e-4
source_model_name = 'i_0={}_r_0={}_betas={}_gammas={}.pt'.format(
source_i_0_set, source_r_0_set, source_betas, source_gammas)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/{}'.format(source_model_name))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
source_epochs = 20000
source_hack_trivial = 0
source_train_size = 2000
gammas = [0.4, 0.7]
# Model parameters
initial_conditions_set = []
initial_conditions_set.append(t_0)
initial_conditions_set.append(i_0_set)
initial_conditions_set.append(r_0_set)
train_size = 1000
decay = 1e-3
hack_trivial = 0
epochs = 1000
lr = 8e-4
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50, output=3)
model_name = 'i_0={}_r_0={}_betas={}_gammas={}.pt'.format(
i_0_set, r_0_set, betas, gammas)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/{}'.format(model_name))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
writer = SummaryWriter('runs/{}'.format(model_name))
sir, train_losses, run_time, optimizer = train_bundle(
sir,
initial_conditions_set,
t_final=t_final,
log_params_losses = [
item for sublist in log_params_losses for item in sublist
]
log_params_losses = [float(x) for x in log_params_losses]
with open('csv\\log_inits_losses.csv', newline='') as f:
reader = csv.reader(f)
log_inits_losses = list(reader)
log_inits_losses = [
item for sublist in log_inits_losses for item in sublist
]
log_inits_losses = [float(x) for x in log_inits_losses]
except:
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50)
models = []
for model_index in range(1, max_model_index + 1, 1):
model_name = '({})_i_0={}_r_0={}_betas={}_gammas={}.pt'.format(
model_index, i_0_set, r_0_set, betas, gammas)
# Init model
sir = SIRNetwork(input=5, layers=4, hidden=50)
checkpoint = torch.load(
ROOT_DIR + '/models/SIR_bundle_total/{}'.format(model_name))
# Load the model
sir.load_state_dict(checkpoint['model_state_dict'])
models.append(sir)
n_draws = 300
steps = 20
assert i_0 + s_0 + r_0 == rescaling_factor
# Model parameters
t_final = 20
train_size = 2500
decay = 0.0
hack_trivial = False
epochs = 1000
lr = 8e-4
# Scipy solver solution
t = np.linspace(0, t_final, t_final)
s_p, i_p, r_p = SIR_solution(t, s_0, i_0, r_0, beta, gamma)
# Init model
sir = SIRNetwork(layers=2, hidden=50)
try:
# It tries to load the model, otherwise it trains it
checkpoint = torch.load(
ROOT_DIR + '/models/SIR/s_0={:.2f}-i_0={:.2f}-r_0={:.2f}'
'-t_0={}-t_f={:.2f}_beta={}_gamma={}.pt'.format(s_0,
i_0, r_0,
initial_conditions[0],
t_final, beta,
gamma))
except FileNotFoundError:
# Train
optimizer = torch.optim.Adam(sir.parameters(), lr=lr)
writer = SummaryWriter(
'runs/' + 's_0={:.2f}-i_0={:.2f}-r_0={:.2f}-t_0={:.2f}-t_f={:.2f}_beta={}_gamma={}.pt'.format(s_0,