# training
for i in range(n_train):
# print(thetas_archive)
t_sample = time()
batch, n_reused, idx_reused, scores_reused = sampler.ask(
pop_size, optimizer)
scores = np.zeros(pop_size)
t_sample = time() - t_sample
# reused samples
for j in range(n_reused):
scores[j] = scores_reused[j]
sample_archive.add_gen(idx_reused[j], curr_gen)
thetas_archive[curr_gen].samples.append(idx_reused[j])
# newly drawn samples
t_eval = time()
for j in range(n_reused, pop_size):
nn.set_params(batch[j])
score = env.eval(nn, render=False)
sample_archive.add_sample(Sample(batch[j], score, [curr_gen]))
thetas_archive[curr_gen].samples.append(n_ind)
scores[j] = score
n_ind += 1
if score > best_score:
best_params = batch[j]
