Merge pull request #54 from kaylode/dev

A quick fix

Author: kaylode

.gitignore

@@ -6,3 +6,4 @@ runs/
 catboost_info/
 .cache
 /build/
+credentials.json

theseus/base/callbacks/loss_logging_callback.py

@@ -52,6 +52,13 @@ def setup(
         else:
             self.params["valloader_length"] = None
 
+        testloader = pl_module.datamodule.testloader
+        if testloader is not None:
+            batch_size = testloader.batch_size
+            self.params["testloader_length"] = len(testloader)
+        else:
+            self.params["testloader_length"] = None
+
         if self.print_interval is None:
             self.print_interval = self.auto_get_print_interval(pl_module)
             LOGGER.text(
@@ -70,6 +77,8 @@ def auto_get_print_interval(
             self.params["trainloader_length"]
             if self.params["trainloader_length"] is not None
             else self.params["valloader_length"]
+            if self.params["valloader_length"] is not None
+            else self.params["testloader_length"]
         )
         print_interval = max(int(train_fraction * num_iterations_per_epoch), 1)
         return print_interval
@@ -249,4 +258,56 @@ def on_validation_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
             for k, v in self.running_loss.items()
         ]
 
+        self.running_loss = {}
+        LOGGER.log(log_dict)
+
+    def on_test_epoch_start(self, *args, **kwargs):
+        """
+        Before main test loops
+        """
+        return self.on_validation_epoch_start(*args, **kwargs)
+
+    def on_test_batch_end(self, *args, **kwargs):
+        """
+        After finish a batch
+        """
+        return self.on_validation_batch_end(*args, **kwargs)
+
+    def on_test_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
+        """
+        After finish validation
+        """
+
+        iters = trainer.global_step
+        num_iterations = self.params["num_iterations"]
+        epoch_time = time.time() - self.running_time
+
+        # Log loss
+        for key in self.running_loss.keys():
+            self.running_loss[key] = np.round(np.mean(self.running_loss[key]), 5)
+        loss_string = (
+            "{}".format(self.running_loss)[1:-1].replace("'", "").replace(",", " ||")
+        )
+        LOGGER.text(
+            "[{}|{}] || {} || Time: {:10.4f} (it/s)".format(
+                iters,
+                num_iterations,
+                loss_string,
+                self.params["testloader_length"] / epoch_time,
+            ),
+            level=LoggerObserver.INFO,
+        )
+
+        # Call other loggers
+        log_dict = [
+            {
+                "tag": f"Test/{k} Loss",
+                "value": v,
+                "type": LoggerObserver.SCALAR,
+                "kwargs": {"step": iters},
+            }
+            for k, v in self.running_loss.items()
+        ]
+
         LOGGER.log(log_dict)
+        self.running_loss = {}

theseus/base/callbacks/metric_logging_callback.py

@@ -21,12 +21,8 @@ class MetricLoggerCallback(Callback):
     def __init__(self, save_json: bool = True, **kwargs) -> None:
         super().__init__()
         self.save_json = save_json
-        if self.save_json:
-            self.save_dir = kwargs.get("save_dir", None)
-            if self.save_dir is not None:
-                self.save_dir = osp.join(self.save_dir, "Validation")
-                os.makedirs(self.save_dir, exist_ok=True)
-            self.output_dict = []
+        self.save_dir = kwargs.get("save_dir", None)
+        self.output_dict = []
 
     def on_validation_end(
         self, trainer: pl.Trainer, pl_module: pl.LightningModule
@@ -64,14 +60,57 @@ def on_validation_end(
 
         LOGGER.log(log_dict)
 
+    def on_test_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule) -> None:
+        """
+        After finish validation
+        """
+        iters = trainer.global_step
+        metric_dict = pl_module.metric_dict
+
+        # Save json
+        if self.save_json:
+            item = {}
+            for metric, score in metric_dict.items():
+                if isinstance(score, (int, float)):
+                    item[metric] = float(f"{score:.5f}")
+            if len(item.keys()) > 0:
+                item["iters"] = iters
+                self.output_dict.append(item)
+
+        # Log metric
+        metric_string = ""
+        for metric, score in metric_dict.items():
+            if isinstance(score, (int, float)):
+                metric_string += metric + ": " + f"{score:.5f}" + " | "
+        metric_string += "\n"
+
+        LOGGER.text(metric_string, level=LoggerObserver.INFO)
+
+        # Call other loggers
+        log_dict = [
+            {"tag": f"Test/{k}", "value": v, "kwargs": {"step": iters}}
+            for k, v in metric_dict.items()
+        ]
+
+        LOGGER.log(log_dict)
+
     def teardown(
         self, trainer: pl.Trainer, pl_module: pl.LightningModule, stage: str
     ) -> None:
         """
         After finish everything
         """
+
         if self.save_json:
-            save_json = osp.join(self.save_dir, "metrics.json")
-            if len(self.output_dict) > 0:
-                with open(save_json, "w") as f:
-                    json.dump(self.output_dict, f)
+            if self.save_dir is not None:
+                save_dir = osp.join(self.save_dir, stage.capitalize())
+                os.makedirs(save_dir, exist_ok=True)
+                save_json = osp.join(save_dir, "metrics.json")
+                if len(self.output_dict) > 0:
+                    with open(save_json, "w") as f:
+                        json.dump(
+                            self.output_dict,
+                            f,
+                            indent=4,
+                            default=lambda x: "<not serializable>",
+                        )

theseus/base/callbacks/timer_callback.py

@@ -92,3 +92,29 @@ def on_validation_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
             "================================================================",
             LoggerObserver.INFO,
         )
+
+    def on_test_start(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
+        """
+        Before main test loops
+        """
+        self.test_epoch_start_time = time.time()
+        LOGGER.text(
+            "=============================TEST EVALUATION===================================",
+            LoggerObserver.INFO,
+        )
+
+    def on_test_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
+        """
+        After finish test
+        """
+
+        running_time = time.time() - self.test_epoch_start_time
+        h, m, s = seconds_to_hours(running_time)
+        LOGGER.text(
+            f"Test evaluation epoch running time: {h} hours, {m} minutes and {s} seconds",
+            level=LoggerObserver.INFO,
+        )
+        LOGGER.text(
+            "================================================================",
+            LoggerObserver.INFO,
+        )

theseus/base/callbacks/wandb_callback.py

@@ -77,6 +77,7 @@ def __init__(
             self.run_name = datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
         self.run_name = osp.basename(save_dir)
 
+        self.config_dict = OmegaConf.to_container(self.config_dict, resolve=True)
         if self.resume is None:
             self.id = wandblogger.util.generate_id()
         else:

theseus/base/models/wrapper.py

@@ -64,6 +64,19 @@ def on_validation_epoch_end(self) -> None:
             batch_size=self.datamodule.valloader.batch_size,
         )
 
+    def on_test_epoch_end(self) -> None:
+        self.metric_dict = {}
+        if self.metrics is not None:
+            for metric in self.metrics:
+                self.metric_dict.update(metric.value())
+                metric.reset()
+
+        self.log_dict(
+            self.metric_dict,
+            prog_bar=True,
+            batch_size=self.datamodule.testloader.batch_size,
+        )
+
     def _forward(self, batch: Dict, metrics: List[Any] = None):
         """
         Forward the batch through models, losses and metrics
@@ -94,6 +107,12 @@ def validation_step(self, batch, batch_idx):
         self.log_dict(outputs["loss_dict"], prog_bar=True, on_step=True, on_epoch=False)
         return outputs
 
+    def test_step(self, batch, batch_idx):
+        # this is the test loop
+        outputs = self._forward(batch, metrics=self.metrics)
+        self.log_dict(outputs["loss_dict"], prog_bar=True, on_step=True, on_epoch=False)
+        return outputs
+
     def predict_step(self, batch, batch_idx=None):
         pred = self.model.get_prediction(batch)
         return pred