Update README.md
Browse files
README.md
CHANGED
|
@@ -61,9 +61,20 @@ state_dict_path = hf_hub_download(repo_id="lschmidt/edsr-dsc", filename="pytorch
|
|
| 61 |
state_dict = torch.load(state_dict_path, map_location="cpu")
|
| 62 |
model.load_state_dict(state_dict, strict=False)
|
| 63 |
|
| 64 |
-
#
|
| 65 |
inputs = torch.randn(1, 2, 64, 64) # replace with coarse wind velocity fields
|
| 66 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
# prediction
|
| 68 |
outputs = model(inputs)
|
| 69 |
|
|
|
|
| 61 |
state_dict = torch.load(state_dict_path, map_location="cpu")
|
| 62 |
model.load_state_dict(state_dict, strict=False)
|
| 63 |
|
| 64 |
+
# create random input: must be a 4D tensor (B, C=2, H, W)
|
| 65 |
inputs = torch.randn(1, 2, 64, 64) # replace with coarse wind velocity fields
|
| 66 |
|
| 67 |
+
# or use sample data
|
| 68 |
+
import xarray as xr
|
| 69 |
+
import numpy as np
|
| 70 |
+
|
| 71 |
+
data_path = hf_hub_download(repo_id="lschmidt/edsr-dsc/test_data", filename="test_wind_velocities.nc")
|
| 72 |
+
ds = xr.open_dataset(data_path)
|
| 73 |
+
u = ds["u100"].values[0]
|
| 74 |
+
v = ds["v100"].values[0]
|
| 75 |
+
inputs = torch.from_numpy(np.stack([u, v], axis=0)).unsqueeze(0).float() # shape (1, 2, H, W)
|
| 76 |
+
|
| 77 |
+
|
| 78 |
# prediction
|
| 79 |
outputs = model(inputs)
|
| 80 |
|