+ str(len(pred_interval_x1)) \
+ ', ' \
+ str(len(pred_interval_x2)) \
+ ')...'
for k in range(num_chains):
pred_posterior = np.empty([len(pred_interval_x1), len(pred_interval_x2)])
nums_dropped_samples = np.empty(
[len(pred_interval_x1), len(pred_interval_x2)], dtype=np.int64)
for i in range(len(pred_interval_x1)):
for j in range(len(pred_interval_x2)):
print(verbose_msg.format(k + 1, i + 1, j + 1))
integral, num_dropped_samples = model.predictive_posterior(
chain_lists.vals['sample'][k],
torch.tensor([[pred_interval_x1[i], pred_interval_x2[j]]],
dtype=dtype), torch.tensor([[1.]], dtype=dtype))
pred_posterior[i, j] = integral.item()
nums_dropped_samples[i, j] = num_dropped_samples
np.savetxt(
sampler_output_run_paths[k].joinpath('pred_posterior_on_grid.csv'),
pred_posterior,
delimiter=',')
np.savetxt(sampler_output_run_paths[k].joinpath(
'pred_posterior_on_grid_num_dropped_samples.csv'),
nums_dropped_samples,
fmt='%d',
delimiter=',')
chain_lists = ChainLists.from_file(sampler_output_run_paths,
keys=['sample'],
dtype=dtype)
# %% Drop burn-in samples
for i in range(num_chains):
chain_lists.vals['sample'][i] = chain_lists.vals['sample'][i][
pred_iter_thres:]
# %% Compute chain means
means = chain_lists.mean()
# %% Make and save predictions
for k in range(num_chains):
test_pred_probs = np.empty([len(test_dataloader)])
for i, (x, _) in enumerate(test_dataloader):
integral, _ = model.predictive_posterior([means[k, :]], x,
torch.tensor([[1.]],
dtype=dtype))
test_pred_probs[i] = integral.item()
test_preds = test_pred_probs > 0.5
np.savetxt(sampler_output_run_paths[k].joinpath('preds_via_mean.txt'),
test_preds,
fmt='%d')