Merge pull request #53 from kaylode/dev

Small fixes

.github/workflows/docker.yml

@@ -60,6 +60,16 @@ jobs:
           run: |
               pytest tests --capture=no
 
+      - name: Log in to Docker Hub
+        uses: docker/login-action@f4ef78c080cd8ba55a85445d5b36e214a81df20a
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+
+      - name: Push image to DockerHub
+        run: |
+          docker push --all-tags
+
       - name: where am I
         run: |
           pwd

.gitignore

@@ -5,3 +5,4 @@ runs/
 *.zip
 catboost_info/
 .cache
+/build/

Dockerfile

@@ -57,7 +57,7 @@ ENV PYTORCH_VERSION ${PYTORCH_VERSION}
 COPY ./ /workspace/
 WORKDIR /workspace/
 RUN /opt/mamba/bin/python -m pip install --upgrade pip && \
-    /opt/mamba/bin/python -m pip install -e .[cv,cv_classification,cv_semantic,cv_detection,nlp,nlp_retrieval,tabular,tabular_classification,dev] && \
+    /opt/mamba/bin/python -m pip install -e .[cv,cv_classification,cv_semantic,cv_detection,nlp,nlp_retrieval,ml,dev] && \
     /opt/mamba/bin/python -m pip install dvc dvc-gdrive && \
     /opt/mamba/bin/python -m pip install -U timm
 

theseus/base/metrics/confusion_matrix.py

@@ -24,7 +24,7 @@ def plot_cfm(cm, ax, labels: List):
     ax.set_xlabel("\nActual")
     ax.set_ylabel("Predicted ")
 
-    ax.xaxis.set_ticklabels(labels)
+    ax.xaxis.set_ticklabels(labels, rotation=90)
     ax.yaxis.set_ticklabels(labels, rotation=0)
 
 

theseus/cv/classification/callbacks/gradcam_callback.py

@@ -65,17 +65,17 @@ def on_validation_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
         # Vizualize Grad Class Activation Mapping and model predictions
         LOGGER.text("Visualizing model predictions...", level=LoggerObserver.DEBUG)
 
-        images = last_batch["inputs"]
+        images = last_batch["inputs"].cpu()
         targets = last_batch["targets"]
         model.eval()
 
         ## Calculate GradCAM and Grad Class Activation Mapping and
-        model_name = model.model.name
+        model_name = model.name
 
         try:
             grad_cam = CAMWrapper.get_method(
                 name="gradcam",
-                model=model.model.get_model(),
+                model=model.get_model(),
                 model_name=model_name,
                 use_cuda=next(model.parameters()).is_cuda,
             )

theseus/cv/classification/callbacks/visualize_callback.py

@@ -50,7 +50,6 @@ def on_sanity_check_start(
         trainloader = pl_module.datamodule.trainloader
         train_batch = next(iter(trainloader))
         val_batch = next(iter(valloader))
-
         try:
             self.visualize_model(model, train_batch)
         except TypeError as e:
@@ -59,16 +58,19 @@ def on_sanity_check_start(
 
     @torch.no_grad()
     def visualize_model(self, model, batch):
+
+        device = next(model.parameters()).device
+
         # Vizualize Model Graph
         LOGGER.text("Visualizing architecture...", level=LoggerObserver.DEBUG)
         LOGGER.log(
             [
                 {
                     "tag": "Sanitycheck/analysis/architecture",
-                    "value": model.model.get_model(),
+                    "value": model.get_model(),
                     "type": LoggerObserver.TORCH_MODULE,
                     "kwargs": {
-                        "inputs": move_to(batch["inputs"], model.device),
+                        "inputs": move_to(batch["inputs"], device),
                         "log_freq": 100,
                     },
                 }
@@ -82,7 +84,7 @@ def visualize_gt(self, train_batch, val_batch, iters):
         LOGGER.text("Visualizing dataset...", level=LoggerObserver.DEBUG)
 
         # Train batch
-        images = train_batch["inputs"]
+        images = train_batch["inputs"].cpu()
 
         batch = []
         for idx, inputs in enumerate(images):
@@ -107,7 +109,7 @@ def visualize_gt(self, train_batch, val_batch, iters):
         )
 
         # Validation batch
-        images = val_batch["inputs"]
+        images = val_batch["inputs"].cpu()
 
         batch = []
         for idx, inputs in enumerate(images):
@@ -162,12 +164,13 @@ def on_validation_end(self, trainer: pl.Trainer, pl_module: pl.LightningModule):
         # Vizualize model predictions
         LOGGER.text("Visualizing model predictions...", level=LoggerObserver.DEBUG)
 
-        images = last_batch["inputs"]
+        images = last_batch["inputs"].cpu()
         targets = last_batch["targets"]
         model.eval()
+        device = next(model.parameters()).device
 
         ## Get prediction on last batch
-        outputs = model.model.get_prediction(last_batch, device=model.device)
+        outputs = model.get_prediction(last_batch, device=device)
         label_indices = outputs["labels"]
         scores = outputs["confidences"]
 

theseus/cv/classification/metrics/errorcases.py

@@ -53,12 +53,12 @@ def update(self, outputs: Dict[str, Any], batch: Dict[str, Any]):
         targets = targets.squeeze(-1).long()
 
         outputs = outputs.numpy().tolist()
-        targets = targets.numpy().tolist()
+        targets = targets.cpu().numpy().tolist()
         probs = probs.numpy().tolist()
 
         for (output, target, prob, image) in zip(outputs, targets, probs, images):
             if output != target:
-                self.images.append(image)
+                self.images.append(image.cpu())
                 self.preds.append(output)
                 self.targets.append(target)
                 self.probs.append(prob)