il_mine.yaml

program: imitate_episodes_pmi.py
method: grid

parameters:
  task_name:
    value: "sim_insertion_human"
    # value: "sim_transfer_cube_scripted"
  policy_class:
    value: "ACT"
  kl_weight:
    value: 10
  decay_rate:
    values: [1]
  chunk_size:
    values: [100]
  ckpt_dir:
    # value: "/localdata/yy/zzzzworkspace/act/ckpt/sim_transfer_cube_scripted_run3"
    # value: "./ckpt/sim_insertion_scripted/Testchunk"
    # value: "/localdata/yy/zzzzworkspace/act/ckpt/sim_insertion_human_chunk100"
    # value: "/localdata/yy/zzzzworkspace/act/ckpt/sim_transfer_cube_scripted_run2"
    value: "/home/torchuser/adaptiveAC/wandb_ckpt/sim_insertion_human_newdataset_mine_newnewarch_test5_chunk100_selfnormalized "
  mine_ckpt_dir:
    value: "/home/torchuser/adaptiveAC/wandb_ckpt/sim_transfer_cube_human_newdataset_MINEnewnewarch_lr1e-4_test1_meansqreg"
  load_mine_ckpt_path:
    # value: "/home/torchuser/adaptiveAC/wandb_ckpt/sim_transfer_cube_human_newdataset_MINEnewnewarch_lr1e-4_test1_meansqreg_seed_1/mine_model_best.ckpt"
    value: '/home/torchuser/adaptiveAC/wandb_ckpt/sim_insertion_human_newdataset_MINEnewnewarch_lr1e-4_test1_meansqreg_seed_1/mine_model_best.ckpt'
  hidden_dim:
    value: 512
  batch_size:
    value: 8
  dim_feedforward:
    value: 3200
  query_freq:
    values: [100]
  temporal_agg:
    value: false
  num_epochs:
    value: 2000
  num_epochs_mine:
    value: 10000
  lr:
    value: 1e-5 # 7e-6
  seed:
    values: [2, 3]
  eval:
    value: false
  onscreen_render:
    value: false
  use_predict_model:
    value: false
  state_dim:
    value: 14
  action_dim:
    value: 14
  reweight:
    value: true
  load_mine:
    value: true
  lr_mine:
    value: 1e-4
  lr_backbone:
    value: 5e-6
  mine_batch_size:
    value: 50
  self_normalize_weight:
    value: true
  weight_clip:
    value: false

command:
  - python3
  - ${program}
  - ${args_no_boolean_flags}