-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathselfnote.txt
More file actions
41 lines (23 loc) · 4.17 KB
/
Copy pathselfnote.txt
File metadata and controls
41 lines (23 loc) · 4.17 KB
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
33
34
35
36
37
38
39
40
41
procedural demo figures: python generate_paper_figures.py --demo
custom model paths: python generate_paper_figures.py --model --zelda-model path/to/zelda_model.zip --sokoban-model path/to/sokoban_model.zip
D:\Work\thesis\RAPCG-MetaRL\pcg_env\Scripts\python.exe train.py --game zelda --timesteps 10000 --checkpoint-freq 2000 --device auto
D:\Work\thesis\RAPCG-MetaRL\pcg_env\Scripts\python.exe train.py --game sokoban --algorithm A2C --timesteps 10000 --checkpoint-freq 2000 --device auto
D:\Work\thesis\RAPCG-MetaRL\pcg_env\Scripts\python.exe train.py --game zelda --algorithm A2C --timesteps 10000 --checkpoint-freq 2000 --device auto
D:\Work\thesis\RAPCG-MetaRL\pcg_env\Scripts\python.exe train.py --game sokoban --algorithm SAC --timesteps 10000 --checkpoint-freq 2000 --device auto
D:\Work\thesis\RAPCG-MetaRL\pcg_env\Scripts\python.exe inference_timed.py checkpoints/sokoban_PPO_20260129_205715/final_model.zip --game sokoban --n-levels 3 --max-steps 1000 --save-dir generated_levels/sokoban_forward --log-file inference_timing_forward.csv --device auto 2>&1 | Select-Object -First 100
python inference_timed.py checkpoints/sokoban_PPO_20260130_162717/best_model.zip --n-levels 5 --save-dir generated_levels/sokoban_forward_test --log-file inference_timing_forward_test.csv --trust-model
python compare_approaches.py --forward-dir generated_levels/sokoban_forward_v2 --backward-dir generated_levels/sokoban_backward_v2
python train.py --game sokoban --algorithm PPO --timesteps 20000 --device cuda --checkpoint-freq 5000 --n-steps 128 --batch-size 64 --lr 2.5e-4
inference_timed.py checkpoints/sokoban_PPO_CUDA_20260220_140931/final_model.zip --game sokoban --algorithm PPO --n-levels 20 --max-steps 1000 --save-dir generated_levels/sokoban_PPO_CUDA_20260220_140931 --log-file inference_timing_sokoban_PPO_CUDA_20260220_140931.csv --device cuda --trust-model
python train.py --game zelda --algorithm PPO --timesteps 20000 --device cuda --checkpoint-freq 5000 --n-steps 128 --batch-size 64 --lr 2.5e-4
inference_timed.py checkpoints\zelda_PPO_CUDA_20260228_113105\final_model.zip --game zelda --algorithm PPO --n-levels 20 --max-steps 1000 --save-dir generated_levels/zelda_PPO_CUDA_20260228_113105 --log-file inference_timing_zelda_PPO_CUDA_20260228_113105.csv --device cuda --trust-model
python rlhf_trainer.py --game sokoban --synthetic --n-levels 5 --n-comparisons 10 --timesteps 1000 --reward-epochs 20 --device cuda
python maml_trainer.py --games sokoban --representations narrow --iterations 2 --meta-batch 1 --inner-steps 1 --n-trajectories 32 --device cuda
python3.10 inference_timed.py checkpoints/zelda_PPO_CUDA_20260228_113105/final_model.zip --game zelda --algorithm PPO --n-levels 20 --max-steps 500 --save-dir generated_levels/zelda_PPO_AMD_20260228_113105 --log-file inference_timing_zelda_PPO_AMD_20260228_113105.csv --device cpu --trust-model
python3.10 inference_timed.py checkpoints/sokoban_PPO_CUDA_20260220_140931/final_model.zip --game sokoban --algorithm PPO --n-levels 20 --max-steps 500 --save-dir generated_levels/sokoban_PPO_AMD_20260220_140931 --log-file inference_timing_sokoban_PPO_AMD_20260220_140931.csv --device cpu --trust-model
python maml_trainer.py --games sokoban --representations narrow --iterations 50 --meta-batch 2 --inner-steps 3 --n-trajectories 64 --device cuda --experiment-name sokoban_MAML_inference
python maml_inference_timed.py checkpoints/sokoban_MAML_inference/best_meta_model.pt --game sokoban --n-levels 40 --max-steps 10000 --device cuda
python maml_trainer.py --games sokoban --representations narrow --iterations 400 --meta-batch 8 --inner-steps 5 --n-trajectories 128 --device cuda --experiment-name sokoban_MAML_production_run
python maml_inference_timed.py checkpoints\sokoban_MAML_production_run\best_meta_model.pt --game sokoban --representation narrow --n-levels 20 --max-steps 500 --adapt-steps 5 --device cuda
python rlhf_trainer.py --game sokoban --representation narrow --n-levels 50 --n-comparisons 50 --reward-epochs 100 --timesteps 50000 --rlhf-weight 0.5 --device cuda --synthetic --experiment-name sokoban_RLHF_cuda_v2
python inference_timed.py checkpoints/sokoban_RLHF_cuda_v2/rlhf_model.zip --game sokoban --n-levels 20 --log-file inference_timing_rlhf.csv --device cuda