fixed syntax

Reviewed-on: https://git.riahub.ai/qoherent/modrec-workflow/pulls/3

.riahub/workflows/workflow.yaml

@@ -2,11 +2,9 @@ name: Modulation Recognition Demo
 
 on:
   push:
-    branches:
-      [main]
+    branches: [main]
   pull_request:
-    branches:
-      [main]
+    branches: [main]
 
 jobs:
   ria-demo:
@@ -46,17 +44,19 @@ jobs:
           fi
           pip install -r requirements.txt
 
-
       - name: 1. Generate Recordings
         run: |
           mkdir -p data/recordings
           PYTHONPATH=. python scripts/dataset_manager/data_gen.py --output-dir data/recordings
 
+      - name: 📦 Compress Recordings
+        run: tar -czf recordings.tar.gz -C data/recordings .
+
       - name: ⬆️ Upload recordings
         uses: actions/upload-artifact@v3
         with:
           name: recordings
-          path: data/recordings/**
+          path: recordings.tar.gz
 
       - name: 2. Build HDF5 Dataset
         run: |
@@ -113,7 +113,7 @@ jobs:
         uses: actions/upload-artifact@v3
         with:
           name: profile-data
-          path: '**/onnxruntime_profile_*.json'
+          path: "**/onnxruntime_profile_*.json"
 
       - name: 7. Convert ONNX graph to an ORT file
         run: |

conf/app.yaml

@@ -24,7 +24,7 @@ dataset:
   snr_step: 3
 
   # Number of iterations (signal recordings) per modulation and SNR combination
-  num_iterations: 3
+  num_iterations: 10
 
   # Modulation scheme settings; keys must match the `modulation_types` list above
   # Each entry includes:

helpers/app_settings.py

@@ -41,7 +41,11 @@ class AppConfig:
 
 
 class AppSettings:
-    """Application settings, to be initialized from app.yaml configuration file."""
+    """
+    Application settings,
+    to be initialized from
+    app.yaml configuration file.
+    """
 
     def __init__(self, config_file: str):
         # Load the YAML configuration file

scripts/application_packager/convert_to_onnx.py

@@ -2,9 +2,9 @@ import os
 
 import numpy as np
 import torch
-from scripts.model_builder.mobilenetv3 import RFClassifier, mobilenetv3
 
 from helpers.app_settings import get_app_settings
+from scripts.model_builder.mobilenetv3 import RFClassifier, mobilenetv3
 
 
 def convert_to_onnx(ckpt_path: str, fp16: bool = False) -> None:
@@ -21,7 +21,7 @@ def convert_to_onnx(ckpt_path: str, fp16: bool = False) -> None:
 
     in_channels = 2
     batch_size = 1
-    slice_length = int(1024 / dataset_cfg.num_slices)
+    slice_length = int(dataset_cfg.recording_length / dataset_cfg.num_slices)
     num_classes = len(dataset_cfg.modulation_types)
 
     model = RFClassifier(
@@ -42,7 +42,7 @@ def convert_to_onnx(ckpt_path: str, fp16: bool = False) -> None:
     model.eval()
 
     # Generate random sample data
-    base, ext = os.path.splitext(os.path.basename(ckpt_path))
+    base, _ = os.path.splitext(os.path.basename(ckpt_path))
     if fp16:
         output_path = os.path.join("onnx_files", f"{base}.onnx")
         sample_input = torch.from_numpy(

scripts/dataset_manager/data_gen.py

@@ -29,7 +29,7 @@ def generate_modulated_signals(output_dir: str) -> None:
 
     for modulation in settings.modulation_types:
         for snr in np.arange(settings.snr_start, settings.snr_stop, settings.snr_step):
-            for i in range(3):
+            for _ in range(settings.num_iterations):
                 recording_length = settings.recording_length
                 beta = (
                     settings.beta

scripts/dataset_manager/produce_dataset.py

@@ -49,8 +49,6 @@ def write_hdf5_file(records: List, output_path: str, dataset_name: str = "data")
             int(md["sps"]),
         )
 
-    first_rec, _ = records[0]  # records[0] is a tuple of (data, md)
-
     with h5py.File(output_path, "w") as hf:
         data_arr = np.stack([rec[0] for rec in records])
         dset = hf.create_dataset(dataset_name, data=data_arr, compression="gzip")

scripts/dataset_manager/split_dataset.py

@@ -90,7 +90,7 @@ def split_recording(
     snippet_list = []
 
     for data, md in recording_list:
-        C, N = data.shape
+        _, N = data.shape
         L = N // num_snippets
         for i in range(num_snippets):
             start = i * L

scripts/model_builder/mobilenetv3.py

@@ -1,5 +1,4 @@
 import lightning as L
-import numpy as np
 import timm
 import torch
 from torch import nn

scripts/model_builder/plot_data.py

@@ -2,26 +2,26 @@ import os
 
 import numpy as np
 import torch
-from sklearn.metrics import classification_report
-
-os.environ["NNPACK"] = "0"
 from matplotlib import pyplot as plt
 from mobilenetv3 import RFClassifier, mobilenetv3
 from modulation_dataset import ModulationH5Dataset
+from sklearn.metrics import classification_report
 
 from helpers.app_settings import get_app_settings
 
+os.environ["NNPACK"] = "0"
+
 
 def load_validation_data():
     val_dataset = ModulationH5Dataset(
         "data/dataset/val.h5", label_name="modulation", data_key="validation_data"
     )
 
-    X = np.stack([x.numpy() for x, _ in val_dataset])  # shape: (N, C, L)
+    x = np.stack([x.numpy() for x, _ in val_dataset])  # shape: (N, C, L)
     y = np.array([y.item() for _, y in val_dataset])  # shape: (N,)
     class_names = list(val_dataset.label_encoder.classes_)
 
-    return X, y, class_names
+    return x, y, class_names
 
 
 def build_model_from_ckpt(
@@ -75,7 +75,7 @@ def evaluate_checkpoint(ckpt_path: str):
         classification_report(y_true, y_pred, target_names=class_names, zero_division=0)
     )
 
-    plot_confusion_matrix_with_counts(
+    print_confusion_matrix(
         y_true=np.array(y_true),
         y_pred=np.array(y_pred),
         classes=class_names,
@@ -84,7 +84,7 @@ def evaluate_checkpoint(ckpt_path: str):
     )
 
 
-def plot_confusion_matrix_with_counts(
+def print_confusion_matrix(
     y_true: np.ndarray,
     y_pred: np.ndarray,
     classes: list[str],
@@ -94,6 +94,7 @@ def plot_confusion_matrix_with_counts(
     """
     Plot a confusion matrix showing both raw counts and (optionally) normalized values.
 
+
     Args:
         y_true: true labels (integers 0..C-1)
         y_pred: predicted labels (same shape as y_true)
@@ -102,8 +103,8 @@ def plot_confusion_matrix_with_counts(
         title: title for the plot
     """
     # 1) build raw CM
-    C = len(classes)
-    cm = np.zeros((C, C), dtype=int)
+    c = len(classes)
+    cm = np.zeros((c, c), dtype=int)
     for t, p in zip(y_true, y_pred):
         cm[t, p] += 1
 
@@ -112,33 +113,48 @@ def plot_confusion_matrix_with_counts(
         with np.errstate(divide="ignore", invalid="ignore"):
             cm_norm = cm.astype(float) / cm.sum(axis=1)[:, None]
             cm_norm = np.nan_to_num(cm_norm)
+        print_confusion_matrix_helper(cm_norm, classes)
     else:
-        cm_norm = cm
+        print_confusion_matrix_helper(cm, classes)
 
-    # 3) plot
-    fig, ax = plt.subplots(figsize=(8, 8))
-    im = ax.imshow(cm_norm, interpolation="nearest")
-    ax.set_title(title)
-    ax.set_xlabel("Predicted label")
-    ax.set_ylabel("True label")
-    ax.set_xticks(np.arange(C))
-    ax.set_yticks(np.arange(C))
-    ax.set_xticklabels(classes, rotation=45, ha="right")
-    ax.set_yticklabels(classes)
 
-    # 4) annotate
-    for i in range(C):
-        for j in range(C):
-            count = cm[i, j]
-            val = cm_norm[i, j]
-            txt = f"{count}\n{val:.2f}"
-            ax.text(j, i, txt, ha="center", va="center")
+import numpy as np
 
-    fig.colorbar(im, ax=ax, label="Normalized value" if normalize else "Count")
-    plt.tight_layout()
-    plt.show()
+
+def print_confusion_matrix_helper(matrix, classes=None, normalize=False, digits=2):
+    """
+    Pretty prints a confusion matrix with x/y labels.
+
+    Parameters:
+    - matrix: square 2D numpy array
+    - labels: list of class labels (default: range(num_classes))
+    - normalize: whether to normalize rows to sum to 1
+    - digits: number of decimal places to show for normalized values
+    """
+    matrix = np.array(matrix)
+    num_classes = matrix.shape[0]
+    labels = classes or list(range(num_classes))
+
+    # Header
+    print(" " * 9 + "Ground Truth →")
+    header = "Pred ↓ | " + "  ".join([f"{str(label):>6}" for label in labels])
+    print(header)
+    print("-" * len(header))
+
+    # Rows
+    for i in range(num_classes):
+        row_vals = matrix[i]
+        if normalize:
+            row_sum = row_vals.sum()
+            row_vals = row_vals / row_sum if row_sum != 0 else row_vals
+            row_str = "  ".join([f"{val:>6.{digits}f}" for val in row_vals])
+        else:
+            row_str = "  ".join([f"{int(val):>6}" for val in row_vals])
+        print(f"{str(labels[i]):>7} | {row_str}")
 
 
 if __name__ == "__main__":
     settings = get_app_settings()
-    evaluate_checkpoint(os.path.join("checkpoint_files", "inference_recognition_model.ckpt"))
+    evaluate_checkpoint(
+        os.path.join("checkpoint_files", "inference_recognition_model.ckpt")
+    )

scripts/model_builder/train.py

@@ -1,23 +1,22 @@
 import os
 import sys
 
-os.environ["NNPACK"] = "0"
 import lightning as L
 import mobilenetv3
 import torch
 import torch.nn.functional as F
 import torchmetrics
-from lightning.pytorch.callbacks import ModelCheckpoint
+from lightning.pytorch.callbacks import ModelCheckpoint, TQDMProgressBar
 from modulation_dataset import ModulationH5Dataset
 
 from helpers.app_settings import get_app_settings
 
+os.environ["NNPACK"] = "0"
 script_dir = os.path.dirname(os.path.abspath(__file__))
 data_dir = os.path.abspath(os.path.join(script_dir, ".."))
 project_root = os.path.abspath(os.path.join(os.getcwd(), ".."))
 if project_root not in sys.path:
     sys.path.insert(0, project_root)
-from lightning.pytorch.callbacks import TQDMProgressBar
 
 
 class CustomProgressBar(TQDMProgressBar):
@@ -59,8 +58,6 @@ def train_model():
         print("X shape:", x.shape)
         print("Y values:", y[:10])
         break
-
-    unique_labels = list(set([row[label].decode("utf-8") for row in ds_train.metadata]))
     num_classes = len(ds_train.label_encoder.classes_)
 
     hparams = {