PVNet 4.0.6__tar.gz → 4.0.8__tar.gz

{pvnet-4.0.6 → pvnet-4.0.8}/PKG-INFO

@@ -1,11 +1,11 @@
 Metadata-Version: 2.4
 Name: PVNet
-Version: 4.0.6
+Version: 4.0.8
 Summary: PVNet
 Author-email: Peter Dudfield <info@openclimatefix.org>
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: ocf_data_sampler==0.1.11
+Requires-Dist: ocf-data-sampler==0.1.16
 Requires-Dist: ocf_datapipes>=3.3.34
 Requires-Dist: ocf_ml_metrics>=0.0.11
 Requires-Dist: numpy

{pvnet-4.0.6 → pvnet-4.0.8}/PVNet.egg-info/PKG-INFO

@@ -1,11 +1,11 @@
 Metadata-Version: 2.4
 Name: PVNet
-Version: 4.0.6
+Version: 4.0.8
 Summary: PVNet
 Author-email: Peter Dudfield <info@openclimatefix.org>
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: ocf_data_sampler==0.1.11
+Requires-Dist: ocf-data-sampler==0.1.16
 Requires-Dist: ocf_datapipes>=3.3.34
 Requires-Dist: ocf_ml_metrics>=0.0.11
 Requires-Dist: numpy

{pvnet-4.0.6 → pvnet-4.0.8}/PVNet.egg-info/requires.txt

@@ -1,4 +1,4 @@
-ocf_data_sampler==0.1.11
+ocf-data-sampler==0.1.16
 ocf_datapipes>=3.3.34
 ocf_ml_metrics>=0.0.11
 numpy

pvnet-4.0.8/pvnet/__init__.py

@@ -0,0 +1,2 @@
+"""PVNet"""
+__version__ = "4.0.8"

{pvnet-4.0.6 → pvnet-4.0.8}/pvnet/models/base_model.py

@@ -1,4 +1,5 @@
 """Base model for all PVNet submodels"""
+
 import json
 import logging
 import os
@@ -722,6 +723,35 @@ class BaseModel(pl.LightningModule, PVNetModelHubMixin):
         try:
             # join together validation results, and save to wandb
             validation_results_df = pd.concat(self.validation_epoch_results)
+            validation_results_df["error"] = (
+                validation_results_df["y"] - validation_results_df["y_quantile_0.5"]
+            )
+
+            if isinstance(self.logger, pl.loggers.WandbLogger):
+                # log error distribution metrics
+                wandb.log(
+                    {
+                        "2nd_percentile_median_forecast_error": validation_results_df[
+                            "error"
+                        ].quantile(0.02),
+                        "5th_percentile_median_forecast_error": validation_results_df[
+                            "error"
+                        ].quantile(0.05),
+                        "95th_percentile_median_forecast_error": validation_results_df[
+                            "error"
+                        ].quantile(0.95),
+                        "98th_percentile_median_forecast_error": validation_results_df[
+                            "error"
+                        ].quantile(0.98),
+                        "95th_percentile_median_forecast_absolute_error": abs(
+                            validation_results_df["error"]
+                        ).quantile(0.95),
+                        "98th_percentile_median_forecast_absolute_error": abs(
+                            validation_results_df["error"]
+                        ).quantile(0.98),
+                    }
+                )
+
             with tempfile.TemporaryDirectory() as tempdir:
                 filename = os.path.join(tempdir, f"validation_results_{self.current_epoch}.csv")
                 validation_results_df.to_csv(filename, index=False)

{pvnet-4.0.6 → pvnet-4.0.8}/pvnet/models/multimodal/multimodal.py

@@ -345,10 +345,11 @@ class Model(MultimodalBaseModel):
             modes["id"] = id_embedding
 
         if self.include_sun:
+            # Use only new direct keys
             sun = torch.cat(
                 (
-                    x[f"{self._target_key}_solar_azimuth"],
-                    x[f"{self._target_key}_solar_elevation"],
+                    x["solar_azimuth"],
+                    x["solar_elevation"],
                 ),
                 dim=1,
             ).float()

{pvnet-4.0.6 → pvnet-4.0.8}/pvnet/models/multimodal/multimodal_base.py

@@ -46,11 +46,10 @@ class MultimodalBaseModel(BaseModel):
                 )[:, : self.nwp_encoders_dict[nwp_source].sequence_length]
 
         if self.include_sun:
-            # Slice off the end of the solar coords data
+            sun_len = self.forecast_len + self.history_len + 1
+            # Slice off end of solar coords
             for s in ["solar_azimuth", "solar_elevation"]:
-                key = f"{self._target_key}_{s}"
-                if key in new_batch.keys():
-                    sun_len = self.forecast_len + self.history_len + 1
-                    new_batch[key] = new_batch[key][:, :sun_len]
+                if s in new_batch.keys():
+                    new_batch[s] = new_batch[s][:, :sun_len]
 
         return new_batch

{pvnet-4.0.6 → pvnet-4.0.8}/pvnet/models/multimodal/unimodal_teacher.py

@@ -305,10 +305,11 @@ class Model(MultimodalBaseModel):
             modes["id"] = id_embedding
 
         if self.include_sun:
+            # Use only new direct keys
             sun = torch.cat(
                 (
-                    x["gsp_solar_azimuth"],
-                    x["gsp_solar_elevation"],
+                    x["solar_azimuth"],
+                    x["solar_elevation"],
                 ),
                 dim=1,
             ).float()

{pvnet-4.0.6 → pvnet-4.0.8}/pyproject.toml

@@ -6,7 +6,7 @@ dynamic = ["version", "readme"]
 license={file="LICENCE"}
 
 dependencies = [
-    "ocf_data_sampler==0.1.11",
+    "ocf-data-sampler==0.1.16",
     "ocf_datapipes>=3.3.34",
     "ocf_ml_metrics>=0.0.11",
     "numpy",

{pvnet-4.0.6 → pvnet-4.0.8}/tests/test_end2end.py

@@ -1,9 +1,13 @@
 import lightning
 
-
 def test_model_trainer_fit(multimodal_model, sample_train_val_datamodule):
+    """Test end-to-end training."""
+    # Get a sample batch for testing
     batch = next(iter(sample_train_val_datamodule.train_dataloader()))
+
+    # Run a forward pass to verify the model works with the data
     y = multimodal_model(batch)
 
+    # Train the model for one batch
     trainer = lightning.pytorch.trainer.trainer.Trainer(fast_dev_run=True, accelerator="cpu")
     trainer.fit(model=multimodal_model, datamodule=sample_train_val_datamodule)

pvnet-4.0.6/pvnet/__init__.py

@@ -1,2 +0,0 @@
-"""PVNet"""
-__version__ = "4.0.6"