forked from LTL2Action/LTL2Action
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathexperiments_paper_rgcn_dfa.sh
executable file
·32 lines (25 loc) · 5.68 KB
/
experiments_paper_rgcn_dfa.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
seed=$1
# No pretraining
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalGeneralDFA --epochs 4 --lr 0.0003 --seed ${seed} --dfa &> outputs/dfa_train_CompositionalGeneralDFA_RGCN_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoid_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa &> outputs/dfa_train_CompositionalReachAvoid_1_2_1_3_RGCN_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoidFix_1_2_1_2 --epochs 4 --lr 0.0003 --seed ${seed} --dfa &> outputs/dfa_train_CompositionalReachAvoidFix_1_2_1_2_RGCN_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalParity_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa &> outputs/dfa_train_CompositionalParity_1_2_1_3_RGCN_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalEventually_1_4_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa &> outputs/dfa_train_CompositionalEventually_1_4_1_3_RGCN_seed_${seed}.txt &
# Pretraining
python train_agent.py --algo ppo --env Simple-LTL-Env-v0 --log-interval 1 --save-interval 20 --frames-per-proc 512 --batch-size 1024 --frames 10000000 --dumb-ac --discount 0.9 --ltl-sampler CompositionalGeneralDFA --lr 0.001 --clip-eps 0.1 --gae-lambda 0.5 --epochs 2 --seed ${seed} --dfa &> outputs/dfa_pretrain_RGCN_seed_${seed}.txt;
cp -r storage/RGCN_8x32_ROOT_SHARED-dumb_ac_CompositionalGeneralDFA_Simple-LTL-Env-v0_seed:${seed}_epochs:2_bs:1024_fpp:512_dsc:0.9_lr:0.001_ent:0.01_clip:0.1_prog:full_dfa:True symbol-storage/
# General DFA
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalGeneralDFA --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA --freeze-ltl &> outputs/dfa_train_CompositionalGeneralDFA_pretrained_RGCN_frozen_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalGeneralDFA --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA &> outputs/dfa_train_CompositionalGeneralDFA_pretrained_RGCN_seed_${seed}.txt &
# Reach-Avoid
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoid_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA --freeze-ltl &> outputs/dfa_train_CompositionalReachAvoid_1_2_1_3_pretrained_RGCN_frozen_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoid_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA &> outputs/dfa_train_CompositionalReachAvoid_1_2_1_3_pretrained_RGCN_seed_${seed}.txt &
# Reach-Avoid-Fix
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoidFix_1_2_1_2 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA --freeze-ltl &> outputs/dfa_train_CompositionalReachAvoidFix_1_2_1_2_pretrained_RGCN_frozen_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalReachAvoidFix_1_2_1_2 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA &> outputs/dfa_train_CompositionalReachAvoidFix_1_2_1_2_pretrained_RGCN_seed_${seed}.txt &
# Parity
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalParity_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA --freeze-ltl &> outputs/dfa_train_CompositionalParity_1_2_1_3_pretrained_RGCN_frozen_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalParity_1_2_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA &> outputs/dfa_train_CompositionalParity_1_2_1_3_pretrained_RGCN_seed_${seed}.txt &
# Partially-Ordered
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalEventually_1_4_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA --freeze-ltl &> outputs/dfa_train_CompositionalEventually_1_4_1_3_pretrained_RGCN_frozen_seed_${seed}.txt &
python train_agent.py --algo ppo --env Letter-7x7-v3 --log-interval 1 --save-interval 20 --frames 20000000 --discount 0.94 --ltl-sampler CompositionalEventually_1_4_1_3 --epochs 4 --lr 0.0003 --seed ${seed} --dfa --pretrained-gnn --pretrained-gnn-sampler CompositionalGeneralDFA &> outputs/dfa_train_CompositionalEventually_1_4_1_3_pretrained_RGCN_seed_${seed}.txt &