reference config in build and train scripts

README.md

@@ -2,12 +2,16 @@
 
 APIs that tap into the quality of the night sky.
 
-## HTTP APIs
+## Building and training the sky brightness model
 
-- artificial sky brightness (light pollution)
-- predictive sky brightness
+> Note that `globe_at_night.tar.gz` will need to unpacked into `./data/globe_at_night`
+
+- `python -m ctts.model.build` to write the csv that the model trains on
+- `python -m ctts.model.train` to train on the data in the csv
 
-### running locally
+## HTTP APIs
+
+### how to run locally
 
 > Note: tested on python 3.11
 
@@ -16,6 +20,10 @@ pip install -r requirements.txt
 python -m uvicorn ctts.api:app --reload
 ```
 
+
+- artificial sky brightness (light pollution)
+- predictive sky brightness
+
 ### endpoints
 
 #### `/api/v1/pollution`
@@ -50,39 +58,3 @@ curl "http://localhost:8000/api/v1/prediction?lat=-30.2466&lon=-70.7494"
 
 Open the [ui](http://localhost:8000/docs) in a browser.
 
-### running with docker
-
-> Note: image size is on the order of 5.13GB
-
-- build the image
-
-```sh
-cd ctts
-docker build -t ctts:latest .
-```
-
-- run the container
-
-```sh
-docker run -d --name ctts -p 8000:80 ctts:latest
-```
-
-### validation
-
-> Note: validation of the model's predictions is an ongoing process.
-
-### running tests
-
-```sh
-cd ctts
-python -m pytest
-```
-
-### env
-
-Location of logfile can be controlled by setting `SKY_BRIGHTNESS_LOGFILE` to
-some filename (in the home directory).
-
-```sh
-export SKY_BRIGHTNESS_LOGFILE=ctts.log
-```

ctts/model/build.py

@@ -9,7 +9,7 @@
 from astropy.time import Time
 
 config = ConfigParser()
-config.read("config.ini")
+config.read(Path(__file__).parent / "config.ini")
 
 max_sqm = config.getint("sqm", "max")
 min_sqm = config.getint("sqm", "min")

ctts/model/config.ini

@@ -3,4 +3,7 @@ max = 22
 min = 16
 
 [csv]
-filename = globe_at_night.csv
+filename = globe_at_night.csv
+
+[train]
+epochs = 100

ctts/model/train.py

@@ -13,13 +13,16 @@
 features_size = len(features)
 
 config = ConfigParser()
-config.read("config.ini")
-# csv_filename = config.get("csv", "filename")
+config.read(Path(__file__).parent / "config.ini")
+
+csv_filename = config.get("csv", "filename")
+epochs = config.getint("train", "epochs")
 
 cwd = Path.cwd()
-csv_filename = "globe_at_night.csv"
+saved_model_path = cwd / "ctts" / "prediction" / "model.pth"
 
-path_to_gan_dataframe = cwd / "data" / csv_filename
+path_to_data_dir = Path(__file__).parent.parent.parent / "data"
+path_to_gan_dataframe = path_to_data_dir / csv_filename
 if not path_to_gan_dataframe.exists():
     raise FileNotFoundError()
 
@@ -87,12 +90,10 @@ def test_model(data_loader: DataLoader, model: NeuralNetwork, loss_fn: nn.HuberL
 
 
 if __name__ == "__main__":
-    epochs = 100
-    saved_model_path = cwd / "ctts" / "prediction" / "model.pth"
     print(f"starting training with {epochs} epochs, and saving state dict to {saved_model_path}")
 
-    for t in range(epochs):
-        print(f"epoch {t+1}")
+    for epoch in range(epochs):
+        print(f"epoch {epoch + 1}")
         train_loop(train_dataloader, model, loss_fn, optimizer)
 
     test_model(test_dataloader, model, loss_fn)

ctts/prediction/constants.py

@@ -11,9 +11,8 @@
 
 HIDDEN_SIZE = 64 * 3
 OUTPUT_SIZE = 1
-MODEL_STATE_DICT_FILE_NAME = "model.pth"
 
-# ASTRO_TWILIGHT_DEGS = -18
+MODEL_STATE_DICT_FILE_NAME = "model.pth"
 
 OPEN_METEO_BASE_URL = "https://api.open-meteo.com"
 MAX_OKTAS = 8

ctts/prediction/prediction.py

@@ -11,7 +11,7 @@
     MODEL_STATE_DICT_FILE_NAME,
     LOGFILE_KEY,
 )
-from .meteo import OpenMeteoClient
+from .open_meteo_client import OpenMeteoClient
 from .nn import NeuralNetwork
 from .observer_site import ObserverSite