DRL-based algorithms for solving SFC Embedding Problems
- Apply Advantage Actor-Critic DRL algorithm to solve the SFC embedding problem
- Run "AC_SFC_Main_01.05.py" for training/testing the A2C RL-agent
- Usage: AC_SFC_Main_01.05.py [-h] [--net_arch NET_ARCH] [--epochs EPOCHS] [--n_workers N_WORKERS] [--sfc_spec SFC_SPEC] [--data_folder DATA_FOLDER] [--model_dir MODEL_DIR] [--model_name MODEL_NAME] [--train_freq TRAIN_FREQ] [--adv_style ADV_STYLE] [--lr LR] [--gamma GAMMA] [--tau TAU] [--critic_factor CRITIC_FACTOR] [--actor_factor ACTOR_FACTOR] [--entropy_factor ENTROPY_FACTOR] [--entropy_decay_val ENTROPY_DECAY_VAL] [--entropy_decay_freq ENTROPY_DECAY_FREQ] [--entropy_min ENTROPY_MIN] [--betas BETAS [BETAS ...]] [--big_rwd BIG_RWD] [--n_steps N_STEPS] [--max_moves MAX_MOVES] [--rsc_scaler RSC_SCALER] [--is_binary_state IS_BINARY_STATE] [--state_noise_scale STATE_NOISE_SCALE] [--opt_lr OPT_LR] [--opt_epsilon OPT_EPSILON] [--opt_weight_decay OPT_WEIGHT_DECAY] [--opt_alpha OPT_ALPHA] [--opt_momentum OPT_MOMENTUM] [--opt_centered OPT_CENTERED] [--test_size TEST_SIZE] [--traffic_type TRAFFIC_TYPE] [--en_log EN_LOG] mode
- Apply DQN RL algorithm to solve the SFC embedding problem
- Run "DQ_SFC_Main_01.08.py" for training the DQN RL-agent
- Run "test_model_random_req_order_v01.08.py" for testing the DQN RL-agent
-
Using the Dataset created by Huy Duong in the SOF-WP1 Project
-
Notes
- The "reordered_traffic_500000_slots_1_con.tra" file cannot be committed to GitHub due to larger than 100 MB of size. This file will be uploaded to another cloud repo.
- Re-organize each algorithm's source code into a separated folder