Added glc24 pre_extracted habitat example in inference version

examples/inference/geolifeclef2024_pre_extracted/config/glc24_cnn_multimodal_ensemble.yaml

@@ -4,7 +4,7 @@ hydra:
 
 run:
   predict_type: 'test_dataset'  # choose from ['test_dataset', 'test_point']
-  checkpoint_path: "outputs_training/glc24_cnn_multimodal_ensemble/runOK_2024-08-12_11-50-01/last.ckpt"
+  checkpoint_path: ???
 
 data:
   root: "dataset/geolifeclef-2024/"

examples/inference/geolifeclef2024_pre_extracted/config/glc24_cnn_multimodal_ensemble_habitat.yaml

@@ -3,8 +3,8 @@ hydra:
     dir: outputs/${hydra.job.name}/${now:%Y-%m-%d_%H-%M-%S}
 
 run:
-  predict: false
-  checkpoint_path: # "outputs/glc24_cnn_multimodal_ensemble_habitat/2024-08-30_20-21-26_multiclass/last.ckpt"
+  predict_type: false
+  checkpoint_path: ???
 
 data:
   root: "dataset/geolifeclef-2024_habitats/"

examples/inference/geolifeclef2024_pre_extracted/glc24_cnn_multimodal_ensemble_habitat.py

@@ -65,39 +65,51 @@ def main(cfg: DictConfig):
                                                checkpoint_path=cfg.run.checkpoint_path,
                                                weights_dir=log_dir,
                                                num_classes=cfg.data.num_classes)  # multiclass
+    model_loaded = ClassificationSystemGLC24.load_from_checkpoint(classif_system.checkpoint_path,
+                                                                  model=classif_system.model,
+                                                                  hparams_preprocess=False,
+                                                                  strict=False)
 
     # Lightning Trainer
     callbacks = [
         Summary(),
         ModelCheckpoint(
             dirpath=log_dir,
-            filename="checkpoint-{epoch:02d}-{step}-{loss/val:.4f}",
-            monitor="loss/val",
+            filename="checkpoint-{epoch:02d}-{step}-{" + f"loss/val" + ":.4f}",
+            monitor=f"loss/val",
             mode="min",
-            save_on_train_epoch_end=True,
-            save_last=True,
-            every_n_train_steps=75,
         ),
     ]
     trainer = pl.Trainer(logger=[logger_csv, logger_tb], callbacks=callbacks, **cfg.trainer, deterministic=True)
 
     # Run
-    if cfg.run.predict:
-        model_loaded = ClassificationSystemGLC24.load_from_checkpoint(classif_system.checkpoint_path,
-                                                                      model=classif_system.model,
-                                                                      hparams_preprocess=False,
-                                                                      strict=False)
-
+    if cfg.run.predict_type == 'test_dataset':
+        # Option 1: Predict on the entire test dataset (Pytorch Lightning)
         predictions = model_loaded.predict(datamodule, trainer)
         preds, probas = datamodule.predict_logits_to_class(predictions,
-                                                           np.arange(cfg.data.num_classes))
+                                                           np.arange(cfg.data.num_classes),
+                                                           activation_fn=torch.nn.Sigmoid())
         datamodule.export_predict_csv(preds, probas,
-                                      out_dir=log_dir, out_name='predictions_test_dataset', top_k=None, return_csv=True)
+                                      out_dir=log_dir, out_name='predictions_test_dataset', top_k=25, return_csv=True)
         print('Test dataset prediction (extract) : ', predictions[:1])
 
-    else:
-        trainer.fit(classif_system, datamodule=datamodule, ckpt_path=classif_system.checkpoint_path)
-        trainer.validate(classif_system, datamodule=datamodule)
+    elif cfg.run.predict_type == 'test_point':
+        # Option 2: Predict 1 data point (Pytorch)
+        test_data = datamodule.get_test_dataset()
+        test_data_point = list(test_data[0][:3])
+        for i, d in enumerate(test_data_point):
+            test_data_point[i] = d.unsqueeze(0)
+        query_point = {'observation_id': [test_data[0][-1]],
+                       'lon': None, 'lat': None,
+                       'crs': None,
+                       'species_id': [test_data[0][-1]]}
+        prediction = model_loaded.predict_point(cfg.run.checkpoint_path,
+                                                test_data_point)
+        preds, probas = datamodule.predict_logits_to_class(prediction,
+                                                           np.arange(cfg.data.num_classes))
+        datamodule.export_predict_csv(preds, probas,
+                                      out_dir=log_dir, out_name='prediction_point', single_point_query=query_point, return_csv=True)
+        print('Point prediction : ', prediction.shape, prediction)
 
 
 if __name__ == "__main__":