slow progress

Author: robotastic

.gitignore

@@ -12,3 +12,7 @@
 *-checkpoint.ipynb
 .DS_Store
 .python-version
+experiment_logs/
+lightning_logs/
+spec_logs/
+

poetry.lock

@@ -11,15 +11,15 @@ scikit-image = "^0.22.0"
 matplotlib = "^3.8.3"
 numpy = "^1.26.4"
 opencv-python = "^4.8.0.74"
-scipy = "^1.12.0"
+scipy = "1.11.4"
 zstandard = "^0.22.0"
 pyyaml = "^6.0.1"
 pillow = "^10.2.0"
 python-on-whales = "^0.69.0"
 sigmf = "^1.2.0"
 tqdm = "^4.66.4"
 cupy = "^13.2.0"
-torchsig = {path = "../torchsig"}
+torchsig = {path = "torchsig"}
 
 [tool.poetry.group.dev.dependencies]
 jupyter = "^1.0.0"

pyproject.toml

@@ -2,16 +2,37 @@
 
 ## Poetry
 
+Follow the instructions here to install Poetry: https://python-poetry.org/docs/#installation
+
 ## Torchsig
 
-Download and install from [Github](https://github.com/TorchDSP/torchsig)
+Download from [Github](https://github.com/TorchDSP/torchsig) and then use Poetry to install it.
 
 ```
 git clone https://github.com/TorchDSP/torchsig.git
 cd torchsig
-sudo pip install .
+git checkout 8049b43
+```
+
+Make you have the Poetry environment activated, using `poetry shell`, then run:
+
+```
+poetry add ./torchsig
 ```
 
+(update it with the correct path the directoy where Torchsig is)
+
+## Torch Model Archiver
+
+Install the Torch Model Archiver:
+```
+sudo pip install torch-model-archiver
+```
+
+More information about this tool is available here: 
+https://github.com/pytorch/serve/blob/master/model-archiver/README.md
+
+## Inspectrum (optional)
 
 
 # Building a Model

rfml-dev/README.md

@@ -176,7 +176,7 @@ def get_occupied_bandwidth(samples, sample_rate, center_frequency, spectral_ener
         else: 
             upper_idx += 1
 
-        if freq_power_normalized[lower_idx:upper_idx].sum() >= spectral_energy_threshold:
+        if freq_power_normalized[lower_idx:upper_idx+1].sum() >= spectral_energy_threshold:
             break
 
         if lower_idx == 0 and upper_idx == freq_power_normalized.shape[0]-1:

rfml-dev/annotation_utils.py

@@ -8,9 +8,9 @@
 # Ensure that data directories have sigmf-meta files with annotations 
 # Annotations can be generated using Label_WiFi.ipynb and Label_DJI.ipynb 
 experiments = {
-    "experiment_0": {
-        "experiment_name": "experiment_0",
-        "class_list": ["mavic3_video","mavic3_remoteid","env"],
+    "experiment_1": {
+        "experiment_name": "experiment_1",
+        "class_list": ["mavic3_video","mavic3_remoteid","environment"],
         "train_dir": ["/home/iqt/lberndt/gamutrf-depoly/data/samples/mavic-30db", "/home/iqt/lberndt/gamutrf-depoly/data/samples/mavic-0db", "/home/iqt/lberndt/gamutrf-depoly/data/samples/environment"],
         "iq_epochs": 10,
         "spec_epochs": 10,
@@ -138,39 +138,39 @@
 
 if __name__ == "__main__":
 
-    for i in [9]: #range(7):
+    for i in [1]: #range(7):
         print(f"Experiment {i}")    
         try:
             exp = Experiment(**experiments[f"experiment_{i}"])
 
             logs_timestamp = datetime.now().strftime("%m_%d_%Y_%H_%M_%S")
 
-            # train_iq(
-            #     train_dataset_path = exp.train_dir,
-            #     val_dataset_path = exp.val_dir,
-            #     num_iq_samples = exp.iq_num_samples, 
-            #     only_use_start_of_burst = exp.iq_only_start_of_burst,
-            #     epochs = exp.iq_epochs, 
-            #     batch_size = exp.iq_batch_size, 
-            #     class_list = exp.class_list, 
-            #     output_dir = Path("experiment_logs",exp.experiment_name),
-            #     logs_dir = Path("iq_logs", logs_timestamp),
-            # )
-
-            
-            train_spec(
+            train_iq(
                 train_dataset_path = exp.train_dir,
                 val_dataset_path = exp.val_dir,
-                n_fft = exp.spec_n_fft, 
-                time_dim = exp.spec_time_dim,
-                epochs = exp.spec_epochs, 
-                batch_size = exp.spec_batch_size, 
+                num_iq_samples = exp.iq_num_samples, 
+                only_use_start_of_burst = exp.iq_only_start_of_burst,
+                epochs = exp.iq_epochs, 
+                batch_size = exp.iq_batch_size, 
                 class_list = exp.class_list, 
-                yolo_augment = exp.spec_yolo_augment,
-                skip_export = exp.spec_skip_export,
-                force_yolo_label_larger = exp.spec_force_yolo_label_larger,
                 output_dir = Path("experiment_logs",exp.experiment_name),
-                logs_dir = Path("spec_logs", logs_timestamp),
+                logs_dir = Path("iq_logs", logs_timestamp),
             )
+
+            
+            # train_spec(
+            #     train_dataset_path = exp.train_dir,
+            #     val_dataset_path = exp.val_dir,
+            #     n_fft = exp.spec_n_fft, 
+            #     time_dim = exp.spec_time_dim,
+            #     epochs = exp.spec_epochs, 
+            #     batch_size = exp.spec_batch_size, 
+            #     class_list = exp.class_list, 
+            #     yolo_augment = exp.spec_yolo_augment,
+            #     skip_export = exp.spec_skip_export,
+            #     force_yolo_label_larger = exp.spec_force_yolo_label_larger,
+            #     output_dir = Path("experiment_logs",exp.experiment_name),
+            #     logs_dir = Path("spec_logs", logs_timestamp),
+            # )
         except Exception as error: 
             print(f"Error: {error}")

rfml-dev/run_experiments.py

@@ -294,7 +294,8 @@ def train_iq(
 
     print(f"\n\nI/Q TRAINING COMPLETE\n\n")
     print(f"Find results in {str(Path(logs_dir))}\n")
-    
+    print(f"Total Accuracy: {acc*100:.2f}%")
+    print(f"Best Model Checkpoint: {checkpoint_callback.best_model_path}")
 
 
 def visualize_dataset(dataset_path, num_iq_samples, logs_dir, class_list):

rfml-dev/train_iq.py

@@ -0,0 +1 @@
+Subproject commit 8049b430baca5d7892f68c4436e5cf7884f5c416