Skip to content

README.md

Latest commit

 

History

History
39 lines (27 loc) · 1.15 KB

README.md

File metadata and controls

39 lines (27 loc) · 1.15 KB
Raw

Adaptive Cache

A learning-based cache eviction simulator, dataset pipeline, and model training stack.

Repo layout

  • src/: Rust simulator, policies, workloads, and dataset builders
  • pytorch_model/: Python training, sweep, and evaluation scripts
  • analysis_outputs/: demo CLI and presentation/demo helpers
  • artifacts/models/: offline checkpoints and normalization metadata
  • artifacts/datasets/: generated train/validation/test CSVs
  • artifacts/traces/: sampled real-world traces used by the simulator and builders
  • scripts/: one-off data preparation helpers

Common commands

Generate training data:

cargo run --bin pairwise_training_dataset_builder -- artifacts/traces/wiki_sampled_train_trace.tsv

Generate validation and test data:

cargo run --bin pairwise_evaluation_datasets_builder -- val:artifacts/traces/wiki_sampled_val_trace.tsv test:artifacts/traces/wiki_sampled_test_trace.tsv

Train the offline model:

python3 pytorch_model/model.py

Run the simulator demo CLI:

python3 analysis_outputs/demo_cli.py race --workload mixed --mixed-mode interleaved --cache-capacity 64 --include-learned