wassname/seq2seq-time
1
0
Fork 0
mirror of https://github.com/wassname/seq2seq-time.git synced 2026-08-18 12:30:31 +08:00
Code Issues Packages Projects Releases Wiki Activity

lots of models, inc NP

Browse Source
This commit is contained in:
wassname
2020-10-19 20:51:23 +08:00
parent e986751e41
commit 7b6c729db5
12 changed files with 3554 additions and 1623 deletions
Download Patch File Download Diff File Expand all files Collapse all files
seq2seq_time/predict.py
+11 -2
View File
@@ -19,11 +19,11 @@ def predict(model, ds_test, batch_size, device='cpu', scaler=None):
load_test = torch.utils.data.dataloader.DataLoader(ds_test, batch_size=batch_size)
freq = ds_test.df.index.freq
xrs = []
for i, batch in enumerate(tqdm(load_test, desc='predict')):
for i, batch in enumerate(tqdm(load_test, desc='predict', leave=False)):
model.eval()
with torch.no_grad():
x_past, y_past, x_future, y_future = [d.to(device) for d in batch]
y_dist = model(x_past, y_past, x_future, y_future)
y_dist, extra = model(x_past, y_past, x_future)
nll = -y_dist.log_prob(y_future)
# Convert to numpy
@@ -70,3 +70,12 @@ def predict(model, ds_test, batch_size, device='cpu', scaler=None):
# Some plots don't like timedeltas, so lets make a coordinate for time ahead in hours
ds_preds = ds_preds.assign_coords(t_ahead_hours=(ds_preds.t_ahead*1.0e-9/60/60).astype(float))
return ds_preds
def predict_multi(model, datasets, batch_size, device='cpu', scaler=None):
"""Predict over multiple datasets."""
ds_preds = [predict(model.to(device),
d,
batch_size,
device=device,
scaler=output_scaler) for d in tqdm(datasets)]
return xr.concat(ds_preds, dim='block')