Feature/implement ray

.env.docker

@@ -1,4 +1,3 @@
 # .env.docker  (container)
 CONTAINER_WORKSPACE=/app
-PYTHONPATH_CONTAINER=/app/src
-NFL_HOME_CONTAINER=/app
+PYTHONPATH_CONTAINER=/app/src

.github/workflows/pr-ci.yml

@@ -37,16 +37,14 @@ jobs:
           version: "latest"
 
       - name: Sync uv
-        run: |
-          uv sync --locked
+        run: uv sync --locked
 
       - name: Create features
-        run: |
-          source environment.sh
-          uv run src/create_features.py
+        run: uv run src/create_features.py
 
       - name: Train transformer
-        run: |
-          source environment.sh
-          uv run src/train_transformer.py
+        run: uv run src/train_transformer.py
+
+      - name: Make week 9 predictions
+        run: uv run src/generate_predictions.py
 

.gitignore

@@ -187,6 +187,10 @@ output/**/*.txt
 
 # MLflow
 mlruns/
+mlflow.db
+
+# Ray
+log/
 
 # Artifacts
 *.pkl

.vscode/launch.json

@@ -12,7 +12,6 @@
       "justMyCode": false,
       "subProcess": true,
       "env": {
-        "NFL_HOME": "${env:NFL_HOME}",
         "PYTHONPATH": "${workspaceFolder}"
       },
       "envFile": "${workspaceFolder}/.env",
@@ -24,14 +23,12 @@
       "request": "launch",
       "program": "${workspaceFolder}/src/main.py",
       "args": [
-        "--data-root", "${env:NFL_HOME}"
       ],
       "console": "integratedTerminal",
       "cwd": "${workspaceFolder}",
       "justMyCode": false,
       "subProcess": true,
       "env": {
-        "NFL_HOME": "${env:NFL_HOME}",
         "PYTHONPATH": "${workspaceFolder}"
       },
       "envFile": "${workspaceFolder}/.env",
@@ -48,7 +45,6 @@
       "justMyCode": false,
       "subProcess": true,
       "env": {
-        "NFL_HOME": "${env:NFL_HOME}",
         "PYTHONPATH": "${workspaceFolder}"
       },
       "envFile": "${workspaceFolder}/.env",
@@ -63,7 +59,6 @@
       "cwd": "${workspaceFolder}",
       "justMyCode": false,
       "env": {
-        "NFL_HOME": "${env:NFL_HOME}",
         "PYTHONPATH": "${workspaceFolder}"
       },
       "envFile": "${workspaceFolder}/.env",

Dockerfile

@@ -30,8 +30,5 @@ RUN uv sync --frozen --no-dev --no-install-project
 # Copy the rest of your repo
 COPY . .
 
-# Default envs helpful for your project
-ENV NFL_HOME=/app
-
 # Set a sensible default; override in docker-compose.yml
 CMD ["bash"]

README.md

@@ -103,34 +103,34 @@ curl -LsSf https://astral.sh/uv/install.sh | sh
 To create and activate the virtual environment, run the following:
 ```bash
 uv sync --locked
-source .venv/bin/activate
 ```
 
 ## Run instructions
 ```bash
-# Source the environmental variables
-source environment.sh
 
 # Digest the raw data and create features
-uv src/create_features.py
+uv run src/create_features.py
 
 # Train the model
-uv src/train_transformer.py
+uv run src/train_transformer.py
 
 # Run inference on week 9
-uv src/generate_predictions.py
+uv run src/generate_predictions.py
 
 # From this point you can explore the notebooks/predictions.ipynb, adjust the data path as neccesary and see the animations and plots
 ```
 
 ## Activating MLflow
 This project uses MLflow to manage ML models and track experiments from hyper paremter optimization. To start MLflow simply:
 ```bash
-# Make sure your virtual environment is started if you opened a new terminal
+# Make sure your virtual environment is sourced so the terminal recognizes mlflow
 source .venv/bin/activate
 
 # Start the tracking server
-mlflow ui --backend-store-uri ./mlruns
+mlflow ui \
+  --backend-store-uri sqlite:///mlflow.db \
+  --host 0.0.0.0 \
+  --port 5000
 ```
 
 The last command will start the MLflow GUI at your local host loopback on port 5000: http://127.0.0.1:5000

data/training/model_params.json

@@ -0,0 +1,9 @@
+{
+    "epochs": 50,
+    "model_dim": 64,
+    "num_heads": 2,
+    "num_layers": 4,
+    "dropout": 0.1,
+    "batch_size": 64,
+    "checkpoint_freq": 10
+}

notebooks/eda.ipynb

@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "id": "0fe17705",
    "metadata": {},
    "outputs": [],
@@ -15,7 +15,6 @@
     "\n",
     "from load_data import RawDataLoader\n",
     "\n",
-    "os.environ[\"NFL_HOME\"] = \"/home/sam/repos/hobby-repos/nfl/\"\n",
     "rawLoader = RawDataLoader()\n",
     "games_df, plays_df, players_df, location_data_df = rawLoader.get_data(weeks=[i for i in range(1, 10)])"
    ]

notebooks/train_lstm.ipynb

@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "id": "1df8838e",
    "metadata": {},
    "outputs": [
@@ -54,8 +54,6 @@
     "#################################################################\n",
     "\n",
     "import os\n",
-    "os.environ[\"NFL_HOME\"] = \"/home/sam/repos/hobby-repos/nfl/\"\n",
-    "\n",
     "from common.data_loader import RawDataLoader\n",
     "\n",
     "# Get raw data\n",

pyproject.toml

@@ -37,42 +37,17 @@ dependencies = [
   "pandas>=2.3.2",
   "pip-tools>=7.5.0",
   "polars>=1.35.1",
+  "ray[train,tune]>=2.51.1",
   "scikit-learn>=1.7.2",
   "seaborn>=0.13.2",
   "torch>=2.8.0",
   "tqdm>=4.67.1",
   "xgboost>=3.0.5",
 ]
 
-[project.optional-dependencies]
-dev = [
-  "pytest>=7.0",
-  "black>=24.0",
-  "flake8>=6.0",
-  "mypy>=1.0"
-]
-# you can define other "extras", e.g.:
-# docs = ["sphinx", "sphinx-rtd-theme"]
-
 [project.scripts]
 # command = "module:function"
 myproject-cli = "myproject.cli:main"
 
 [tool.setuptools.packages.find]
 where = ["src"]
-
-# Tool-specific configuration examples:
-
-[tool.black]
-line-length = 88
-target-version = ["py39", "py310", "py311", "py312"]
-
-[tool.isort]
-profile = "black"
-
-[tool.flake8]
-max-line-length = 88
-extend-ignore = ["E203"]
-
-[tool.mypy]
-strict = true

src/common/args.py

@@ -0,0 +1,21 @@
+#!/usr/bin/env python
+
+"""
+Parses input arguments to then act on in code
+"""
+
+import argparse
+
+def parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--tune",
+        action="store_true",
+        help="Use Ray Tune to search hyperparameters instead of a single training run",
+    )
+    parser.add_argument(
+        "--profile",
+        action="store_true",
+        help="Enable @time_fcn timing decorators for profiling",
+    )
+    return parser.parse_args()

src/common/csv_to_parquet.py

@@ -4,9 +4,9 @@
 Transforms raw csv data to raw parquet data
 
 Requires:
-- Raw location tracking data in NFL_HOME/data/parquet
+- Raw location tracking data in nfl/data/parquet
 
-Outputs parquet data in NFL_HOME/data/parquet that is stored in git LFS
+Outputs parquet data in nfl/data/parquet that is stored in git LFS
 """
 
 import sys

src/common/decorators.py

@@ -4,14 +4,34 @@
 Contains common use decorators
 """
 
-def time_fcn(func):
-    import time
-    from functools import wraps
+import time
+import functools
+import logging
 
-    @wraps(func)
+_TIME_DECORATORS_ENABLED = False
+
+def set_time_decorators_enabled(enabled: bool) -> None:
+    """
+    Enable/disable time_fcn decorators globally, called at startup
+    """
+    global _TIME_DECORATORS_ENABLED
+    _TIME_DECORATORS_ENABLED = enabled
+
+def time_fcn(fn):
+    """
+    Decorator that measures wall-clock time of a function and logs it.
+    When _TIME_DECORATORS_ENABLED is False, it becomes a simple passthrough.
+    """
+    @functools.wraps(fn)
     def wrapper(*args, **kwargs):
+        # Fast exit when disabled (minimal overhead)
+        if not _TIME_DECORATORS_ENABLED:
+            return fn(*args, **kwargs)
+
         start = time.perf_counter()
-        result = func(*args, **kwargs)
-        print(f"{func.__name__} took {time.perf_counter() - start:.4f}s")
-        return result
-    return wrapper
+        try:
+            return fn(*args, **kwargs)
+        finally:
+            elapsed = time.perf_counter() - start
+            logging.info("Function %s took %.4f s", fn.__name__, elapsed)
+    return wrapper

src/common/paths.py

@@ -0,0 +1,12 @@
+from pathlib import Path
+
+def project_root() -> Path:
+    # 2 since this is at src/commmon/path.py
+    return Path(__file__).resolve().parents[2]
+
+# Set project root
+PROJECT_ROOT = project_root()
+
+# Create processing dirs
+SAVE_DIR = PROJECT_ROOT / "data" / "processed"
+SAVE_DIR.mkdir(parents=True, exist_ok=True)