params.yaml

env_name: ASTEnv
seed: 1
cuda: true
project_name: assistant_rl
cuda_deterministic: true
num_processes: 12
num_steps: 128
log_interval: 1
eval_interval: 0 #1000
save_interval: 1000
num_env_steps: 1.0e+8
use_linear_lr_decay: true
resume_carryover:
  - seed
  - env
  - eval
  - base
  - ppo
  - make_tests
  - return
env:
  get_from1: data/auto_curricula/2_basic/param_snippet.yaml   
  get_from2: data/auto_curricula/3_basic/param_snippet.yaml    
  ds_ratio: [0.4,0.6]
  perturbation: 0
eval:
    assignment_dir: /RL_env/data/auto_curricula_frozen/3vars_basic/curriculum/test
    code_per_assignment:
    - 0
    cursor_start_pos:
    - 0 
    done_action: true
    max_episode_steps: 0
    num_assignments: 0
    perturbation: 0
base:
    hidden_size: 256
    num_layers: 10
    embedding_size: 250
    heads: 16
    use_qkv: false
    gnn_type: GATv2
ppo:
    lr: 1.0e-05
    eps: 1.0e-05
    value_loss_coef: 0.5
    max_grad_norm: 0.5
    ppo_epoch: 4
    num_mini_batch: 4
    clip_param: 0.1
    entropy_coef: 0.015
    entropy_coeff_decay: True 
make_tests:
    # command: "echo-frozen tests"
    command: python3 scripts/gen_template_funcs.py
    args:
      base:
      - '2'
      - data/auto_curricula/2_basic
    #   - -v 
      mns_correction: 1.6
      curriculum: 
      variations: false  
      permutations: false
      test_split: 0.5
    args2:
      base:
      - '3'
      - data/auto_curricula/3_basic
      # - -v 
      mns_correction: 1.6
      curriculum: 
      variations: false  
      permutations: false
      # test_split: 0.5
return:
    gamma: 0.99
    use_gae: true
    gae_lambda: 0.95
    use_proper_time_limits: false