From c54ddb88b532be28ca3096e21de405d90163ecfa Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Mon, 29 Jun 2026 12:04:47 -0500 Subject: Package full RRoG GNN project --- logs/local_gpu3_act_ablation_20260624_102042.log | 39 + ...cal_gpu3_act_ablation_retry_20260624_102233.log | 2105 ++++++++++++++++++++ logs/ogbg-molbace_0.log | 683 +++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 166 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ logs/ogbg-molbbbp_0.log | 647 ++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 166 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ ...hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log | 84 + logs/ogbg-molclintox_0.log | 756 +++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 164 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ logs/ogbg-molesol_0.log | 757 +++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 31 + ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 242 +++ logs/ogbg-molfreesolv_0.log | 791 ++++++++ logs/ogbg-molhiv_0.log | 612 ++++++ logs/ogbg-molhiv_T1_ns3_s0.log | 626 ++++++ logs/ogbg-molhiv_T1_ns3_s1.log | 887 +++++++++ logs/ogbg-molhiv_T1_ns3_s2.log | 887 +++++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 152 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ ...hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log | 84 + logs/ogbg-mollipo_0.log | 828 ++++++++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 242 +++ logs/ogbg-molsider_0.log | 791 ++++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 164 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ ...hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log | 84 + logs/ogbg-moltox21_0.log | 720 +++++++ ...am_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log | 164 ++ ..._hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log | 278 +++ logs/zinc_cycle56_0.log | 536 +++++ 33 files changed, 15076 insertions(+) create mode 100644 logs/local_gpu3_act_ablation_20260624_102042.log create mode 100644 logs/local_gpu3_act_ablation_retry_20260624_102233.log create mode 100644 logs/ogbg-molbace_0.log create mode 100644 logs/ogbg-molbace_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molbace_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molbbbp_0.log create mode 100644 logs/ogbg-molbbbp_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molbbbp_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molbbbp_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log create mode 100644 logs/ogbg-molclintox_0.log create mode 100644 logs/ogbg-molclintox_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molclintox_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molesol_0.log create mode 100644 logs/ogbg-molesol_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molesol_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molfreesolv_0.log create mode 100644 logs/ogbg-molhiv_0.log create mode 100644 logs/ogbg-molhiv_T1_ns3_s0.log create mode 100644 logs/ogbg-molhiv_T1_ns3_s1.log create mode 100644 logs/ogbg-molhiv_T1_ns3_s2.log create mode 100644 logs/ogbg-molhiv_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molhiv_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molhiv_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log create mode 100644 logs/ogbg-mollipo_0.log create mode 100644 logs/ogbg-mollipo_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molsider_0.log create mode 100644 logs/ogbg-molsider_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molsider_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-molsider_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw30_e100_s0.log create mode 100644 logs/ogbg-moltox21_0.log create mode 100644 logs/ogbg-moltox21_act_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/ogbg-moltox21_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log create mode 100644 logs/zinc_cycle56_0.log (limited to 'logs') diff --git a/logs/local_gpu3_act_ablation_20260624_102042.log b/logs/local_gpu3_act_ablation_20260624_102042.log new file mode 100644 index 0000000..99d42f7 --- /dev/null +++ b/logs/local_gpu3_act_ablation_20260624_102042.log @@ -0,0 +1,39 @@ +[gpu3-ablation] start Wed Jun 24 10:20:44 CDT 2026 +[gpu3-ablation] phase1 exact target, classification tasks, views=gin gcn sgc tag graphconv appnp pna resgated +[launch] cuda:3: ogbg-molhiv ogbg-molbbbp ogbg-molsider ogbg-molbace ogbg-moltox21 ogbg-molclintox +[launch] cuda:3: ogbg-molesol ogbg-mollipo ogbg-moltox21 ogbg-molclintox +[task] ogbg-molhiv on cuda:3 +[task] ogbg-molesol on cuda:3 +[run] ogbg-molesol view=gin compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +[run] ogbg-molhiv view=gin compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gin --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view gin --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +Downloading http://snap.stanford.edu/ogb/data/graphproppred/csv_mol_download/esol.zip + 0%| | 0/2 [00:00 + main() + ~~~~^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 601, in main + act_state, train_metrics = train_epoch( + ~~~~~~~~~~~^ + model, train_loader, opt, dev, args, act_state, ema_state, ep + 1, metric) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 507, in train_epoch + act_state, m = act_train_step(model, act_state, batch, opt, dev, args, epoch, metric) + ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 412, in act_train_step + raise ValueError(args.halt_target) +ValueError: exact diff --git a/logs/local_gpu3_act_ablation_retry_20260624_102233.log b/logs/local_gpu3_act_ablation_retry_20260624_102233.log new file mode 100644 index 0000000..52adc5f --- /dev/null +++ b/logs/local_gpu3_act_ablation_retry_20260624_102233.log @@ -0,0 +1,2105 @@ +[gpu3-ablation] start Wed Jun 24 10:22:35 CDT 2026 +[gpu3-ablation] phase1 exact target, classification tasks, views=gin gcn sgc tag graphconv appnp pna resgated +[launch] cuda:3: ogbg-molhiv ogbg-molbbbp ogbg-molsider ogbg-molbace ogbg-moltox21 ogbg-molclintox +[task] ogbg-molhiv on cuda:3 +[run] ogbg-molhiv view=gin compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view gin --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.57684 val_adapt_rocauc=0.58405 adapt_steps=2.00 halt=0.47 train_steps=1.59 +ep20 val_rocauc=0.60452 val_adapt_rocauc=0.59144 adapt_steps=2.01 halt=0.47 train_steps=1.59 +ep30 val_rocauc=0.71542 val_adapt_rocauc=0.72246 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep40 val_rocauc=0.70211 val_adapt_rocauc=0.68285 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep50 val_rocauc=0.59821 val_adapt_rocauc=0.72487 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.64796 val_adapt_rocauc=0.71730 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep70 val_rocauc=0.70277 val_adapt_rocauc=0.73025 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.68902 val_adapt_rocauc=0.72341 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.72762 val_adapt_rocauc=0.74793 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.74125 val_adapt_rocauc=0.75760 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_gin_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=100 val={'rocauc': 0.7412490201842054} test={'rocauc': 0.7254234342107805} adaptive={'rocauc': 0.7567005929044593} steps=2.00024313153416 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_gin_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=gcn compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view gcn --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.67913 val_adapt_rocauc=0.65745 adapt_steps=2.00 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.53086 val_adapt_rocauc=0.70728 adapt_steps=2.02 halt=0.47 train_steps=1.59 +ep30 val_rocauc=0.73614 val_adapt_rocauc=0.73784 adapt_steps=2.04 halt=0.48 train_steps=1.57 +ep40 val_rocauc=0.74062 val_adapt_rocauc=0.74468 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep50 val_rocauc=0.64917 val_adapt_rocauc=0.74024 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.74111 val_adapt_rocauc=0.74943 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep70 val_rocauc=0.69073 val_adapt_rocauc=0.75506 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.75837 val_adapt_rocauc=0.74771 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.77126 val_adapt_rocauc=0.75543 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.76237 val_adapt_rocauc=0.74920 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_gcn_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rocauc': 0.7712620027434841} test={'rocauc': 0.7476081036713725} adaptive={'rocauc': 0.7387782691824871} steps=2.012156576707999 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_gcn_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=sgc compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view sgc --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.56968 val_adapt_rocauc=0.68573 adapt_steps=2.01 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.65960 val_adapt_rocauc=0.64533 adapt_steps=2.00 halt=0.47 train_steps=1.59 +ep30 val_rocauc=0.73782 val_adapt_rocauc=0.74603 adapt_steps=2.03 halt=0.48 train_steps=1.57 +ep40 val_rocauc=0.70390 val_adapt_rocauc=0.72870 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep50 val_rocauc=0.68717 val_adapt_rocauc=0.73036 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.68431 val_adapt_rocauc=0.72508 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep70 val_rocauc=0.69653 val_adapt_rocauc=0.73800 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.70493 val_adapt_rocauc=0.74137 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.69423 val_adapt_rocauc=0.73412 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.69464 val_adapt_rocauc=0.74486 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_sgc_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=30 val={'rocauc': 0.7378227268273565} test={'rocauc': 0.7035709457502076} adaptive={'rocauc': 0.724386334228162} steps=2.0269876002917577 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_sgc_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=tag compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view tag --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.64438 val_adapt_rocauc=0.62773 adapt_steps=2.01 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.64326 val_adapt_rocauc=0.64524 adapt_steps=2.01 halt=0.47 train_steps=1.61 +ep30 val_rocauc=0.68766 val_adapt_rocauc=0.66943 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep40 val_rocauc=0.63871 val_adapt_rocauc=0.72422 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep50 val_rocauc=0.63257 val_adapt_rocauc=0.73165 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.68678 val_adapt_rocauc=0.75939 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep70 val_rocauc=0.73265 val_adapt_rocauc=0.76437 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.71019 val_adapt_rocauc=0.75081 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.71181 val_adapt_rocauc=0.75327 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.72871 val_adapt_rocauc=0.76052 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_tag_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=70 val={'rocauc': 0.7326450127376054} test={'rocauc': 0.759611039224396} adaptive={'rocauc': 0.7786419204696885} steps=2.0 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_tag_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=graphconv compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view graphconv --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.62212 val_adapt_rocauc=0.62817 adapt_steps=2.01 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.61445 val_adapt_rocauc=0.69759 adapt_steps=2.00 halt=0.47 train_steps=1.59 +ep30 val_rocauc=0.68557 val_adapt_rocauc=0.71732 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep40 val_rocauc=0.47925 val_adapt_rocauc=0.71316 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep50 val_rocauc=0.61766 val_adapt_rocauc=0.69094 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.73183 val_adapt_rocauc=0.73302 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep70 val_rocauc=0.72858 val_adapt_rocauc=0.73825 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.74850 val_adapt_rocauc=0.74077 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.73303 val_adapt_rocauc=0.74217 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.73150 val_adapt_rocauc=0.73659 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_graphconv_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rocauc': 0.7484965951401137} test={'rocauc': 0.722140249908264} adaptive={'rocauc': 0.7285405280132873} steps=2.00024313153416 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_graphconv_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=appnp compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view appnp --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.70638 val_adapt_rocauc=0.68842 adapt_steps=2.01 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.64695 val_adapt_rocauc=0.67193 adapt_steps=2.00 halt=0.47 train_steps=1.59 +ep30 val_rocauc=0.71509 val_adapt_rocauc=0.75189 adapt_steps=2.01 halt=0.48 train_steps=1.58 +ep40 val_rocauc=0.65625 val_adapt_rocauc=0.68931 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep50 val_rocauc=0.73236 val_adapt_rocauc=0.74043 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.68610 val_adapt_rocauc=0.70942 adapt_steps=2.01 halt=0.48 train_steps=1.58 +ep70 val_rocauc=0.71282 val_adapt_rocauc=0.71800 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep80 val_rocauc=0.70420 val_adapt_rocauc=0.71488 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.71668 val_adapt_rocauc=0.72406 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.71284 val_adapt_rocauc=0.72607 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_appnp_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=50 val={'rocauc': 0.7323571918479326} test={'rocauc': 0.7487070047702736} adaptive={'rocauc': 0.7287433129260897} steps=2.0063214198881596 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_appnp_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=pna compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view pna --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep10 val_rocauc=0.52661 val_adapt_rocauc=0.57540 adapt_steps=2.02 halt=0.47 train_steps=1.60 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep20 val_rocauc=0.53124 val_adapt_rocauc=0.54271 adapt_steps=2.00 halt=0.48 train_steps=1.58 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep30 val_rocauc=0.64086 val_adapt_rocauc=0.71386 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep40 val_rocauc=0.66964 val_adapt_rocauc=0.69129 adapt_steps=2.00 halt=0.48 train_steps=1.58 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep50 val_rocauc=0.63916 val_adapt_rocauc=0.69673 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep60 val_rocauc=0.73238 val_adapt_rocauc=0.74287 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep70 val_rocauc=0.73849 val_adapt_rocauc=0.76782 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep80 val_rocauc=0.72881 val_adapt_rocauc=0.78390 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep90 val_rocauc=0.73387 val_adapt_rocauc=0.77480 adapt_steps=2.00 halt=0.48 train_steps=1.57 +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/home/yurenh2/miniconda3/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep100 val_rocauc=0.73341 val_adapt_rocauc=0.77625 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_pna_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=70 val={'rocauc': 0.7384932882618068} test={'rocauc': 0.7185075030417737} adaptive={'rocauc': 0.7421078043222156} steps=2.0 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_pna_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molhiv view=resgated compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molhiv --view resgated --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +ep10 val_rocauc=0.50096 val_adapt_rocauc=0.63103 adapt_steps=2.00 halt=0.47 train_steps=1.60 +ep20 val_rocauc=0.65405 val_adapt_rocauc=0.70279 adapt_steps=2.00 halt=0.47 train_steps=1.58 +ep30 val_rocauc=0.68276 val_adapt_rocauc=0.65325 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep40 val_rocauc=0.80097 val_adapt_rocauc=0.78965 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep50 val_rocauc=0.77943 val_adapt_rocauc=0.71584 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep60 val_rocauc=0.77564 val_adapt_rocauc=0.75634 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep70 val_rocauc=0.77493 val_adapt_rocauc=0.78545 adapt_steps=2.00 halt=0.48 train_steps=1.58 +ep80 val_rocauc=0.77393 val_adapt_rocauc=0.77547 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep90 val_rocauc=0.76533 val_adapt_rocauc=0.76729 adapt_steps=2.00 halt=0.48 train_steps=1.57 +ep100 val_rocauc=0.76010 val_adapt_rocauc=0.76356 adapt_steps=2.00 halt=0.48 train_steps=1.57 +[ogbg-molhiv_resgated_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=40 val={'rocauc': 0.8009657309425828} test={'rocauc': 0.7787017903010873} adaptive={'rocauc': 0.7875702504876494} steps=2.0 + wrote /home/yurenh2/rrog-gnn-runner/runs/ogbg-molhiv_resgated_rrog-act_T1_ns3_stream_hm8_hmin2_exact_lq0.1_hex0.1_qw0_h128_e100_s0.json +[task] ogbg-molbbbp on cuda:3 +[run] ogbg-molbbbp view=gin compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:3 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molbbbp --view gin --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target exact --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:3 --num_workers 0 +Downloading http://snap.stanford.edu/ogb/data/graphproppred/csv_mol_download/bbbp.zip + 0%| | 0/2 [00:00 + main() + ~~~~^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 601, in main + act_state, train_metrics = train_epoch( + ~~~~~~~~~~~^ + model, train_loader, opt, dev, args, act_state, ema_state, ep + 1, metric) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 507, in train_epoch + act_state, m = act_train_step(model, act_state, batch, opt, dev, args, epoch, metric) + ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/yurenh2/rrog-gnn-runner/rrog/train_ogb_graphprop.py", line 412, in act_train_step + raise ValueError(args.halt_target) +ValueError: exact diff --git a/logs/ogbg-molesol_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log b/logs/ogbg-molesol_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log new file mode 100644 index 0000000..cd5a0e1 --- /dev/null +++ b/logs/ogbg-molesol_act_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_e100_s0.log @@ -0,0 +1,242 @@ +[run] ogbg-molesol view=gin compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gin --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.63809 val_adapt_rmse=2.39395 adapt_steps=6.48 halt=0.19 train_steps=3.84 +ep20 val_rmse=4.26988 val_adapt_rmse=3.11447 adapt_steps=3.99 halt=0.16 train_steps=4.33 +ep30 val_rmse=4.14634 val_adapt_rmse=3.52731 adapt_steps=4.76 halt=0.16 train_steps=4.14 +ep40 val_rmse=2.34153 val_adapt_rmse=1.55824 adapt_steps=2.54 halt=0.29 train_steps=2.64 +ep50 val_rmse=2.10126 val_adapt_rmse=1.29124 adapt_steps=2.93 halt=0.34 train_steps=2.23 +ep60 val_rmse=1.97711 val_adapt_rmse=1.38502 adapt_steps=3.55 halt=0.29 train_steps=2.55 +ep70 val_rmse=1.85198 val_adapt_rmse=1.17093 adapt_steps=2.89 halt=0.35 train_steps=2.24 +ep80 val_rmse=1.40467 val_adapt_rmse=0.94661 adapt_steps=2.88 halt=0.37 train_steps=2.11 +ep90 val_rmse=1.51603 val_adapt_rmse=0.96581 adapt_steps=2.71 halt=0.39 train_steps=1.94 +ep100 val_rmse=1.55791 val_adapt_rmse=0.98643 adapt_steps=2.61 halt=0.40 train_steps=1.96 +[ogbg-molesol_gin_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rmse': np.float32(1.4046698)} test={'rmse': np.float32(2.0651195)} adaptive={'rmse': np.float32(0.96633816)} steps=2.982300884955752 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_gin_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=gine compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gine --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=3.38193 val_adapt_rmse=3.24861 adapt_steps=5.82 halt=0.18 train_steps=4.23 +ep20 val_rmse=2.14610 val_adapt_rmse=1.81691 adapt_steps=4.67 halt=0.18 train_steps=3.98 +ep30 val_rmse=2.42302 val_adapt_rmse=2.12907 adapt_steps=4.80 halt=0.19 train_steps=3.43 +ep40 val_rmse=1.72178 val_adapt_rmse=1.35344 adapt_steps=4.30 halt=0.24 train_steps=3.14 +ep50 val_rmse=2.54499 val_adapt_rmse=1.50567 adapt_steps=2.35 halt=0.30 train_steps=2.66 +ep60 val_rmse=1.52869 val_adapt_rmse=1.10370 adapt_steps=3.33 halt=0.34 train_steps=2.66 +ep70 val_rmse=1.72160 val_adapt_rmse=1.01718 adapt_steps=3.65 halt=0.35 train_steps=2.32 +ep80 val_rmse=1.64223 val_adapt_rmse=1.07042 adapt_steps=2.33 halt=0.38 train_steps=2.24 +ep90 val_rmse=1.31103 val_adapt_rmse=1.01651 adapt_steps=2.51 halt=0.41 train_steps=2.03 +ep100 val_rmse=1.36548 val_adapt_rmse=1.01212 adapt_steps=2.47 halt=0.40 train_steps=2.11 +[ogbg-molesol_gine_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rmse': np.float32(1.3110343)} test={'rmse': np.float32(2.0394285)} adaptive={'rmse': np.float32(1.0992771)} steps=2.8849557522123894 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_gine_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=gcn compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gcn --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.02916 val_adapt_rmse=1.75337 adapt_steps=4.50 halt=0.23 train_steps=3.50 +ep20 val_rmse=1.54387 val_adapt_rmse=1.38205 adapt_steps=4.77 halt=0.22 train_steps=3.63 +ep30 val_rmse=1.69653 val_adapt_rmse=1.49022 adapt_steps=4.99 halt=0.22 train_steps=3.20 +ep40 val_rmse=2.92037 val_adapt_rmse=1.90808 adapt_steps=3.73 halt=0.25 train_steps=3.16 +ep50 val_rmse=1.03997 val_adapt_rmse=0.90930 adapt_steps=3.62 halt=0.29 train_steps=2.79 +ep60 val_rmse=1.07837 val_adapt_rmse=1.00303 adapt_steps=3.17 halt=0.39 train_steps=2.10 +ep70 val_rmse=1.16818 val_adapt_rmse=0.92152 adapt_steps=3.02 halt=0.38 train_steps=2.08 +ep80 val_rmse=1.08067 val_adapt_rmse=0.92960 adapt_steps=3.04 halt=0.38 train_steps=2.18 +ep90 val_rmse=1.02939 val_adapt_rmse=0.87992 adapt_steps=2.79 halt=0.40 train_steps=2.00 +ep100 val_rmse=1.04713 val_adapt_rmse=0.88052 adapt_steps=2.70 halt=0.40 train_steps=1.98 +[ogbg-molesol_gcn_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rmse': np.float32(1.0293915)} test={'rmse': np.float32(1.2572919)} adaptive={'rmse': np.float32(0.92217094)} steps=2.7168141592920354 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_gcn_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=graphsage compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view graphsage --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.09992 val_adapt_rmse=1.91741 adapt_steps=5.06 halt=0.15 train_steps=4.24 +ep20 val_rmse=1.34611 val_adapt_rmse=1.17463 adapt_steps=3.27 halt=0.20 train_steps=3.81 +ep30 val_rmse=2.05686 val_adapt_rmse=1.52317 adapt_steps=3.48 halt=0.24 train_steps=3.01 +ep40 val_rmse=1.58216 val_adapt_rmse=1.15681 adapt_steps=2.60 halt=0.34 train_steps=2.19 +ep50 val_rmse=1.32595 val_adapt_rmse=1.11669 adapt_steps=3.81 halt=0.24 train_steps=2.77 +ep60 val_rmse=1.68484 val_adapt_rmse=1.33655 adapt_steps=2.79 halt=0.38 train_steps=2.03 +ep70 val_rmse=1.46531 val_adapt_rmse=1.02634 adapt_steps=2.50 halt=0.36 train_steps=2.18 +ep80 val_rmse=1.21621 val_adapt_rmse=0.92526 adapt_steps=2.48 halt=0.40 train_steps=1.95 +ep90 val_rmse=1.15871 val_adapt_rmse=0.92033 adapt_steps=2.41 halt=0.43 train_steps=1.76 +ep100 val_rmse=1.19358 val_adapt_rmse=0.92401 adapt_steps=2.24 halt=0.44 train_steps=1.82 +[ogbg-molesol_graphsage_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rmse': np.float32(1.1587086)} test={'rmse': np.float32(1.3228503)} adaptive={'rmse': np.float32(0.9323515)} steps=2.398230088495575 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_graphsage_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=gatv2 compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gatv2 --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=1.71115 val_adapt_rmse=1.59292 adapt_steps=5.16 halt=0.19 train_steps=3.79 +ep20 val_rmse=2.04041 val_adapt_rmse=1.48254 adapt_steps=3.27 halt=0.19 train_steps=3.61 +ep30 val_rmse=2.19314 val_adapt_rmse=1.41846 adapt_steps=2.98 halt=0.26 train_steps=3.01 +ep40 val_rmse=1.26104 val_adapt_rmse=1.11810 adapt_steps=2.53 halt=0.31 train_steps=2.64 +ep50 val_rmse=1.35004 val_adapt_rmse=1.06584 adapt_steps=3.13 halt=0.32 train_steps=2.50 +ep60 val_rmse=1.14151 val_adapt_rmse=0.91810 adapt_steps=2.64 halt=0.39 train_steps=2.00 +ep70 val_rmse=1.08552 val_adapt_rmse=0.98888 adapt_steps=3.42 halt=0.35 train_steps=2.30 +ep80 val_rmse=1.01822 val_adapt_rmse=0.95246 adapt_steps=2.61 halt=0.44 train_steps=1.82 +ep90 val_rmse=0.98423 val_adapt_rmse=0.94151 adapt_steps=2.59 halt=0.44 train_steps=1.75 +ep100 val_rmse=1.02002 val_adapt_rmse=0.97264 adapt_steps=2.60 halt=0.46 train_steps=1.75 +[ogbg-molesol_gatv2_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rmse': np.float32(0.9842344)} test={'rmse': np.float32(0.95342916)} adaptive={'rmse': np.float32(0.88059396)} steps=2.4867256637168142 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_gatv2_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=graphconv compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view graphconv --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=3.24691 val_adapt_rmse=3.06485 adapt_steps=5.43 halt=0.16 train_steps=4.14 +ep20 val_rmse=2.19544 val_adapt_rmse=1.92386 adapt_steps=4.11 halt=0.22 train_steps=3.14 +ep30 val_rmse=1.85038 val_adapt_rmse=1.84516 adapt_steps=5.91 halt=0.20 train_steps=3.80 +ep40 val_rmse=1.80320 val_adapt_rmse=1.53707 adapt_steps=4.28 halt=0.24 train_steps=3.32 +ep50 val_rmse=1.61943 val_adapt_rmse=1.16471 adapt_steps=4.05 halt=0.34 train_steps=2.62 +ep60 val_rmse=1.38128 val_adapt_rmse=1.12648 adapt_steps=4.63 halt=0.30 train_steps=2.80 +ep70 val_rmse=1.15362 val_adapt_rmse=0.98844 adapt_steps=2.86 halt=0.38 train_steps=2.03 +ep80 val_rmse=1.14805 val_adapt_rmse=1.00274 adapt_steps=3.15 halt=0.43 train_steps=1.82 +ep90 val_rmse=1.14935 val_adapt_rmse=1.02412 adapt_steps=3.17 halt=0.42 train_steps=1.87 +ep100 val_rmse=1.15233 val_adapt_rmse=1.03384 adapt_steps=3.10 halt=0.43 train_steps=1.84 +[ogbg-molesol_graphconv_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rmse': np.float32(1.148049)} test={'rmse': np.float32(1.2080547)} adaptive={'rmse': np.float32(1.0046481)} steps=2.743362831858407 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_graphconv_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=transformer compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view transformer --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=1.80556 val_adapt_rmse=1.57940 adapt_steps=5.16 halt=0.21 train_steps=3.72 +ep20 val_rmse=2.04585 val_adapt_rmse=1.53384 adapt_steps=4.59 halt=0.19 train_steps=3.46 +ep30 val_rmse=2.20692 val_adapt_rmse=1.76335 adapt_steps=2.60 halt=0.24 train_steps=3.17 +ep40 val_rmse=1.06643 val_adapt_rmse=1.01122 adapt_steps=2.19 halt=0.37 train_steps=2.20 +ep50 val_rmse=1.26673 val_adapt_rmse=1.02385 adapt_steps=2.27 halt=0.35 train_steps=2.07 +ep60 val_rmse=1.27758 val_adapt_rmse=1.04181 adapt_steps=2.62 halt=0.34 train_steps=2.26 +ep70 val_rmse=1.27827 val_adapt_rmse=1.01072 adapt_steps=2.54 halt=0.36 train_steps=2.05 +ep80 val_rmse=1.20543 val_adapt_rmse=1.03738 adapt_steps=2.62 halt=0.41 train_steps=1.93 +ep90 val_rmse=1.26838 val_adapt_rmse=1.02284 adapt_steps=2.44 halt=0.44 train_steps=1.75 +ep100 val_rmse=1.23701 val_adapt_rmse=1.02901 adapt_steps=2.45 halt=0.44 train_steps=1.76 +[ogbg-molesol_transformer_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=40 val={'rmse': np.float32(1.0664288)} test={'rmse': np.float32(0.95339394)} adaptive={'rmse': np.float32(0.8505217)} steps=2.2831858407079646 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_transformer_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=pna compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view pna --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +/orion/u/oscarwan/rrog-gnn-runner/.venv/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='min')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +/orion/u/oscarwan/rrog-gnn-runner/.venv/lib/python3.13/site-packages/torch_geometric/utils/_scatter.py:91: UserWarning: The usage of `scatter(reduce='max')` can be accelerated via the 'torch-scatter' package, but it was not found + warnings.warn( +ep10 val_rmse=2.43831 val_adapt_rmse=2.32531 adapt_steps=5.19 halt=0.16 train_steps=4.00 +ep20 val_rmse=2.56767 val_adapt_rmse=1.59439 adapt_steps=3.97 halt=0.17 train_steps=3.98 +ep30 val_rmse=1.46321 val_adapt_rmse=1.22048 adapt_steps=3.73 halt=0.24 train_steps=3.12 +ep40 val_rmse=1.91382 val_adapt_rmse=1.86215 adapt_steps=3.05 halt=0.22 train_steps=3.11 +ep50 val_rmse=1.55782 val_adapt_rmse=1.32158 adapt_steps=2.92 halt=0.32 train_steps=2.34 +ep60 val_rmse=1.68320 val_adapt_rmse=1.18630 adapt_steps=2.27 halt=0.35 train_steps=2.29 +ep70 val_rmse=1.38794 val_adapt_rmse=1.15712 adapt_steps=2.46 halt=0.33 train_steps=2.29 +ep80 val_rmse=1.28100 val_adapt_rmse=1.02706 adapt_steps=2.46 halt=0.35 train_steps=2.23 +ep90 val_rmse=1.47195 val_adapt_rmse=1.10808 adapt_steps=2.43 halt=0.35 train_steps=2.19 +ep100 val_rmse=1.45332 val_adapt_rmse=1.10590 adapt_steps=2.44 halt=0.35 train_steps=2.21 +[ogbg-molesol_pna_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rmse': np.float32(1.2809987)} test={'rmse': np.float32(1.4175682)} adaptive={'rmse': np.float32(0.98294556)} steps=2.601769911504425 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_pna_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=gen compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view gen --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=1.68554 val_adapt_rmse=1.40872 adapt_steps=4.66 halt=0.17 train_steps=4.16 +ep20 val_rmse=1.78903 val_adapt_rmse=1.34417 adapt_steps=3.88 halt=0.27 train_steps=2.88 +ep30 val_rmse=1.43062 val_adapt_rmse=1.26301 adapt_steps=2.82 halt=0.29 train_steps=2.68 +ep40 val_rmse=1.26387 val_adapt_rmse=1.08550 adapt_steps=2.36 halt=0.32 train_steps=2.40 +ep50 val_rmse=1.43121 val_adapt_rmse=1.00136 adapt_steps=2.40 halt=0.32 train_steps=2.47 +ep60 val_rmse=1.16224 val_adapt_rmse=1.00159 adapt_steps=2.69 halt=0.32 train_steps=2.49 +ep70 val_rmse=1.36256 val_adapt_rmse=1.13936 adapt_steps=2.18 halt=0.39 train_steps=2.13 +ep80 val_rmse=1.19902 val_adapt_rmse=0.99500 adapt_steps=2.50 halt=0.39 train_steps=1.97 +ep90 val_rmse=1.19996 val_adapt_rmse=1.00030 adapt_steps=2.23 halt=0.41 train_steps=1.93 +ep100 val_rmse=1.16268 val_adapt_rmse=0.98398 adapt_steps=2.26 halt=0.40 train_steps=2.02 +[ogbg-molesol_gen_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=60 val={'rmse': np.float32(1.1622369)} test={'rmse': np.float32(1.4459299)} adaptive={'rmse': np.float32(1.0918992)} steps=2.6106194690265485 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_gen_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=film compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view film --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.15953 val_adapt_rmse=1.69788 adapt_steps=5.27 halt=0.13 train_steps=4.40 +ep20 val_rmse=1.76413 val_adapt_rmse=1.78195 adapt_steps=5.78 halt=0.17 train_steps=4.04 +ep30 val_rmse=1.71072 val_adapt_rmse=1.54898 adapt_steps=4.97 halt=0.15 train_steps=3.94 +ep40 val_rmse=1.73032 val_adapt_rmse=1.23604 adapt_steps=4.69 halt=0.19 train_steps=3.30 +ep50 val_rmse=1.80233 val_adapt_rmse=1.54453 adapt_steps=4.03 halt=0.23 train_steps=3.17 +ep60 val_rmse=2.03690 val_adapt_rmse=1.28646 adapt_steps=3.64 halt=0.22 train_steps=3.11 +ep70 val_rmse=1.81923 val_adapt_rmse=1.44603 adapt_steps=4.35 halt=0.20 train_steps=3.30 +ep80 val_rmse=1.36738 val_adapt_rmse=1.27507 adapt_steps=4.00 halt=0.28 train_steps=2.75 +ep90 val_rmse=1.47914 val_adapt_rmse=1.26687 adapt_steps=3.88 halt=0.27 train_steps=2.67 +ep100 val_rmse=1.56815 val_adapt_rmse=1.29753 adapt_steps=3.90 halt=0.28 train_steps=2.80 +[ogbg-molesol_film_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rmse': np.float32(1.3673769)} test={'rmse': np.float32(1.364793)} adaptive={'rmse': np.float32(1.12864)} steps=3.6106194690265485 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_film_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=resgated compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view resgated --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.01199 val_adapt_rmse=1.99778 adapt_steps=6.58 halt=0.17 train_steps=4.21 +ep20 val_rmse=2.81354 val_adapt_rmse=1.57554 adapt_steps=3.77 halt=0.24 train_steps=3.39 +ep30 val_rmse=2.27715 val_adapt_rmse=1.78467 adapt_steps=2.41 halt=0.31 train_steps=2.96 +ep40 val_rmse=2.06149 val_adapt_rmse=1.59902 adapt_steps=4.17 halt=0.25 train_steps=2.81 +ep50 val_rmse=1.25669 val_adapt_rmse=1.02207 adapt_steps=3.41 halt=0.33 train_steps=2.60 +ep60 val_rmse=1.24545 val_adapt_rmse=1.02109 adapt_steps=3.96 halt=0.38 train_steps=2.21 +ep70 val_rmse=1.11263 val_adapt_rmse=0.97278 adapt_steps=2.63 halt=0.47 train_steps=1.62 +ep80 val_rmse=1.21134 val_adapt_rmse=1.06008 adapt_steps=3.45 halt=0.41 train_steps=1.89 +ep90 val_rmse=1.09389 val_adapt_rmse=0.97993 adapt_steps=3.27 halt=0.41 train_steps=1.90 +ep100 val_rmse=1.14293 val_adapt_rmse=1.00867 adapt_steps=3.06 halt=0.44 train_steps=1.83 +[ogbg-molesol_resgated_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=90 val={'rmse': np.float32(1.0938909)} test={'rmse': np.float32(1.0833443)} adaptive={'rmse': np.float32(0.9181242)} steps=3.230088495575221 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_resgated_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=tag compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view tag --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=3.47158 val_adapt_rmse=2.85814 adapt_steps=6.48 halt=0.15 train_steps=4.28 +ep20 val_rmse=2.59229 val_adapt_rmse=2.29301 adapt_steps=4.50 halt=0.19 train_steps=3.54 +ep30 val_rmse=1.38104 val_adapt_rmse=1.23529 adapt_steps=3.79 halt=0.17 train_steps=4.02 +ep40 val_rmse=1.62549 val_adapt_rmse=1.31054 adapt_steps=3.39 halt=0.24 train_steps=3.04 +ep50 val_rmse=1.17882 val_adapt_rmse=1.00373 adapt_steps=3.59 halt=0.26 train_steps=2.89 +ep60 val_rmse=1.11478 val_adapt_rmse=1.01010 adapt_steps=2.65 halt=0.36 train_steps=2.17 +ep70 val_rmse=1.21310 val_adapt_rmse=1.04888 adapt_steps=2.49 halt=0.35 train_steps=2.14 +ep80 val_rmse=1.13492 val_adapt_rmse=1.00575 adapt_steps=2.39 halt=0.40 train_steps=1.92 +ep90 val_rmse=1.10848 val_adapt_rmse=0.98178 adapt_steps=2.41 halt=0.39 train_steps=1.99 +ep100 val_rmse=1.08712 val_adapt_rmse=0.96400 adapt_steps=2.39 halt=0.44 train_steps=1.82 +[ogbg-molesol_tag_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=100 val={'rmse': np.float32(1.087122)} test={'rmse': np.float32(1.2508167)} adaptive={'rmse': np.float32(1.0189501)} steps=2.353982300884956 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_tag_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=sgc compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view sgc --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=2.13452 val_adapt_rmse=1.54890 adapt_steps=4.80 halt=0.18 train_steps=4.20 +ep20 val_rmse=1.80887 val_adapt_rmse=1.55135 adapt_steps=4.19 halt=0.20 train_steps=3.79 +ep30 val_rmse=1.56727 val_adapt_rmse=1.38135 adapt_steps=3.77 halt=0.28 train_steps=3.18 +ep40 val_rmse=1.44096 val_adapt_rmse=1.19056 adapt_steps=5.21 halt=0.21 train_steps=3.29 +ep50 val_rmse=1.27696 val_adapt_rmse=1.20518 adapt_steps=2.98 halt=0.28 train_steps=2.83 +ep60 val_rmse=1.03453 val_adapt_rmse=1.03118 adapt_steps=3.41 halt=0.29 train_steps=2.72 +ep70 val_rmse=1.36152 val_adapt_rmse=1.19328 adapt_steps=2.70 halt=0.37 train_steps=2.20 +ep80 val_rmse=1.10044 val_adapt_rmse=1.00095 adapt_steps=2.53 halt=0.41 train_steps=1.89 +ep90 val_rmse=1.09772 val_adapt_rmse=1.00221 adapt_steps=2.59 halt=0.43 train_steps=1.80 +ep100 val_rmse=1.08154 val_adapt_rmse=1.01459 adapt_steps=2.57 halt=0.42 train_steps=1.90 +[ogbg-molesol_sgc_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=60 val={'rmse': np.float32(1.0345254)} test={'rmse': np.float32(1.0465622)} adaptive={'rmse': np.float32(0.97806746)} steps=3.2831858407079646 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_sgc_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=cheb compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view cheb --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=1.61304 val_adapt_rmse=1.41832 adapt_steps=3.95 halt=0.23 train_steps=3.80 +ep20 val_rmse=2.75789 val_adapt_rmse=2.14843 adapt_steps=3.42 halt=0.25 train_steps=3.13 +ep30 val_rmse=1.83964 val_adapt_rmse=1.52535 adapt_steps=3.04 halt=0.22 train_steps=3.33 +ep40 val_rmse=1.78527 val_adapt_rmse=1.47363 adapt_steps=2.77 halt=0.34 train_steps=2.29 +ep50 val_rmse=1.79619 val_adapt_rmse=1.39923 adapt_steps=2.56 halt=0.46 train_steps=1.75 +ep60 val_rmse=1.55892 val_adapt_rmse=1.29826 adapt_steps=2.79 halt=0.40 train_steps=2.14 +ep70 val_rmse=1.21742 val_adapt_rmse=1.01859 adapt_steps=2.35 halt=0.43 train_steps=1.82 +ep80 val_rmse=1.06434 val_adapt_rmse=0.91884 adapt_steps=2.42 halt=0.44 train_steps=1.73 +ep90 val_rmse=1.08489 val_adapt_rmse=0.95548 adapt_steps=2.42 halt=0.46 train_steps=1.66 +ep100 val_rmse=1.08888 val_adapt_rmse=0.95485 adapt_steps=2.50 halt=0.46 train_steps=1.74 +[ogbg-molesol_cheb_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=80 val={'rmse': np.float32(1.0643402)} test={'rmse': np.float32(1.034204)} adaptive={'rmse': np.float32(0.8497807)} steps=2.601769911504425 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_cheb_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=arma compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view arma --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=4.56663 val_adapt_rmse=3.60617 adapt_steps=2.74 halt=0.19 train_steps=3.89 +ep20 val_rmse=1.79390 val_adapt_rmse=1.59002 adapt_steps=3.36 halt=0.18 train_steps=3.97 +ep30 val_rmse=2.24326 val_adapt_rmse=1.60470 adapt_steps=3.49 halt=0.18 train_steps=3.64 +ep40 val_rmse=1.96715 val_adapt_rmse=1.62786 adapt_steps=2.96 halt=0.23 train_steps=3.09 +ep50 val_rmse=1.30996 val_adapt_rmse=1.07779 adapt_steps=2.49 halt=0.31 train_steps=2.61 +ep60 val_rmse=1.59369 val_adapt_rmse=1.15346 adapt_steps=2.60 halt=0.31 train_steps=2.58 +ep70 val_rmse=1.33600 val_adapt_rmse=1.06852 adapt_steps=2.23 halt=0.39 train_steps=2.12 +ep80 val_rmse=1.51381 val_adapt_rmse=1.08174 adapt_steps=2.51 halt=0.36 train_steps=2.25 +ep90 val_rmse=1.57786 val_adapt_rmse=1.05584 adapt_steps=2.32 halt=0.37 train_steps=2.15 +ep100 val_rmse=1.51730 val_adapt_rmse=1.01467 adapt_steps=2.27 halt=0.39 train_steps=2.00 +[ogbg-molesol_arma_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=50 val={'rmse': np.float32(1.3099641)} test={'rmse': np.float32(1.5309947)} adaptive={'rmse': np.float32(1.0385827)} steps=2.575221238938053 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_arma_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=mf compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view mf --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=3.91043 val_adapt_rmse=3.93609 adapt_steps=7.12 halt=0.15 train_steps=4.32 +ep20 val_rmse=1.56765 val_adapt_rmse=1.40924 adapt_steps=3.93 halt=0.21 train_steps=3.97 +ep30 val_rmse=2.21943 val_adapt_rmse=1.92857 adapt_steps=4.80 halt=0.21 train_steps=3.70 +ep40 val_rmse=1.34838 val_adapt_rmse=1.13183 adapt_steps=2.79 halt=0.24 train_steps=2.98 +ep50 val_rmse=2.04182 val_adapt_rmse=1.18497 adapt_steps=3.55 halt=0.32 train_steps=2.48 +ep60 val_rmse=1.07650 val_adapt_rmse=1.06298 adapt_steps=3.69 halt=0.38 train_steps=2.00 +ep70 val_rmse=1.17131 val_adapt_rmse=1.05686 adapt_steps=3.22 halt=0.40 train_steps=2.01 +ep80 val_rmse=1.16899 val_adapt_rmse=1.03185 adapt_steps=4.14 halt=0.37 train_steps=2.18 +ep90 val_rmse=1.09374 val_adapt_rmse=1.03779 adapt_steps=3.41 halt=0.41 train_steps=1.92 +ep100 val_rmse=1.13886 val_adapt_rmse=1.05515 adapt_steps=3.32 halt=0.42 train_steps=1.92 +[ogbg-molesol_mf_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=60 val={'rmse': np.float32(1.0765038)} test={'rmse': np.float32(1.0843762)} adaptive={'rmse': np.float32(0.95887977)} steps=3.4070796460176993 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_mf_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json +[run] ogbg-molesol view=appnp compute=rrog-act mode=stream T=1 ns=3 seed=0 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molesol --view appnp --compute rrog-act --T 1 --n_sup 3 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --lam_q 0.1 --halt_max_steps 8 --halt_min_steps 2 --halt_target loss --halt_loss_threshold 0.2 --halt_exploration_prob 0.1 --act_train_mode stream --q_warmup_epochs 0 --device cuda:1 --num_workers 0 +ep10 val_rmse=5.21048 val_adapt_rmse=2.38789 adapt_steps=5.34 halt=0.15 train_steps=4.20 +ep20 val_rmse=2.82679 val_adapt_rmse=2.25175 adapt_steps=6.12 halt=0.18 train_steps=3.93 +ep30 val_rmse=1.68909 val_adapt_rmse=1.41500 adapt_steps=4.84 halt=0.19 train_steps=3.72 +ep40 val_rmse=2.08134 val_adapt_rmse=1.83294 adapt_steps=4.53 halt=0.20 train_steps=3.51 +ep50 val_rmse=1.95089 val_adapt_rmse=1.48969 adapt_steps=3.67 halt=0.26 train_steps=2.94 +ep60 val_rmse=1.80235 val_adapt_rmse=1.37428 adapt_steps=3.46 halt=0.33 train_steps=2.30 +ep70 val_rmse=1.62703 val_adapt_rmse=1.25301 adapt_steps=2.74 halt=0.42 train_steps=1.88 +ep80 val_rmse=1.70083 val_adapt_rmse=1.27133 adapt_steps=2.71 halt=0.39 train_steps=2.06 +ep90 val_rmse=1.71529 val_adapt_rmse=1.30814 adapt_steps=2.67 halt=0.40 train_steps=1.94 +ep100 val_rmse=1.69038 val_adapt_rmse=1.31634 adapt_steps=2.68 halt=0.40 train_steps=1.95 +[ogbg-molesol_appnp_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0] best_ep=70 val={'rmse': np.float32(1.6270279)} test={'rmse': np.float32(1.2973164)} adaptive={'rmse': np.float32(1.0733075)} steps=2.814159292035398 + wrote /orion/u/oscarwan/rrog-gnn-runner/runs/ogbg-molesol_appnp_rrog-act_T1_ns3_stream_hm8_hmin2_loss0.2_lq0.1_hex0.1_qw0_h128_e100_s0.json diff --git a/logs/ogbg-molfreesolv_0.log b/logs/ogbg-molfreesolv_0.log new file mode 100644 index 0000000..d163fe6 --- /dev/null +++ b/logs/ogbg-molfreesolv_0.log @@ -0,0 +1,791 @@ +[run] ogbg-molfreesolv view=gin compute=classic T=0 ns=1 device=cuda:1 +python3 rrog/train_ogb_graphprop.py --dataset ogbg-molfreesolv --view gin --compute classic --T 0 --n_sup 1 --hidden 128 --bs 128 --epochs 100 --eval_every 10 --agg_layers 5 --compute_layers 2 --seed 0 --device cuda:1 +Downloading http://snap.stanford.edu/ogb/data/graphproppred/csv_mol_download/freesolv.zip + 0%| | 0/2 [00:00