How srujanasri2008/ml-project works
This research application trains a SAC policy to allocate power, NOMA splits, and airtime while semantic tasks move through service, completion, or deadline expiry. The main workflow runs from CLI configuration through training, then evaluates SAC against allocation baselines and produces plots; an optional Streamlit dashboard is also documented. The graph includes every indexed core module. Some subsystem internals and dashboard data wiring were not shown, so those links are omitted rather than inferred.
Main entry points: cli.py Experiment CLI · environment.py Gym environment · evaluation.py Policy evaluation · config.py Configuration
Components 17
Experiment entry
- Experiment CLI
src/semantic_noma/cli.py - Configuration
src/semantic_noma/config.py - Default settings
configs/default.json - Smoke settings
configs/smoke_test.json
Wireless simulation
- Gym environment
src/semantic_noma/environment.py - Channel model
src/semantic_noma/channel_model.py - Semantic task lifecycle
src/semantic_noma/task_model.py - Resource allocator
src/semantic_noma/resource_allocator.py - Wireless calculations
src/semantic_noma/wireless_math.py
Policy learning
- SAC training
src/semantic_noma/training.py - Training callbacks
src/semantic_noma/callbacks.py - Allocation baselines
src/semantic_noma/baselines.py
Evaluation and reporting
- Policy evaluation
src/semantic_noma/evaluation.py - Episode metrics
src/semantic_noma/metrics.py - Result plotting
src/semantic_noma/plotting.py
Optional interface
- Interactive dashboard
dashboard/app.py
Other components
- Researcher
Connections 17
- Researcher → Experiment CLI: runs
README.md - Experiment CLI → Configuration: loads
src/semantic_noma/cli.py - Configuration → Default settings: reads
- Configuration → Smoke settings: reads
- Experiment CLI → SAC training: starts training
src/semantic_noma/cli.py - Experiment CLI → Policy evaluation: runs evaluation
src/semantic_noma/cli.py - Experiment CLI → Result plotting: generates plots
src/semantic_noma/cli.py - SAC training → Gym environment: creates
src/semantic_noma/training.py - SAC training → Training callbacks: registers
src/semantic_noma/training.py - Gym environment → Channel model: samples gains
src/semantic_noma/environment.py - Gym environment → Semantic task lifecycle: updates tasks
src/semantic_noma/environment.py - Gym environment → Resource allocator: allocates resources
src/semantic_noma/environment.py - Resource allocator → Wireless calculations: computes rates
src/semantic_noma/resource_allocator.py - Policy evaluation → Gym environment: steps through
src/semantic_noma/evaluation.py - Policy evaluation → Allocation baselines: compares policies
src/semantic_noma/evaluation.py - Policy evaluation → Episode metrics: computes fairness
src/semantic_noma/evaluation.py - Researcher → Interactive dashboard: opens
README.md