Skip to content

MBH Catalyst discovery #9

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
MiquelAngelPerezPuigdo wants to merge 4 commits into
base: master
Choose a base branch
from
Open

MBH Catalyst discovery #9

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f6f8f52
MBH Catalyst discovery
MiquelAngelPerezPuigdo Jun 2, 2026
096aab1
API key out
MiquelAngelPerezPuigdo Jun 2, 2026
7691ef2
Add agentic MBH catalyst oracle, plots; update README; remove MBH_tri…
MiquelAngelPerezPuigdo Jun 16, 2026
65e4a02
Remove butenolide catalyst oracle references
MiquelAngelPerezPuigdo Jun 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
105 changes: 105 additions & 0 deletions MBH/MBH_trial.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,105 @@
{
"logging": {
"logging_frequency": 100,
"logging_path": "/home/miquelaperez/saturn/MBH/mbh_catalyst_run/log.log",
"model_checkpoints_dir": "/home/miquelaperez/saturn/MBH/mbh_catalyst_run"
},
"oracle": {
"budget": 1500,
"allow_oracle_repeats": false,
"aggregator": "product",
"components": [
{
"name": "MBH_catalyst_score",
"weight": 1,
"preliminary_check": false,
"specific_parameters": {
"api_key": "",
"model_name": "gemini-3.5-flash",
"num_calls": 3,
"batch_size": 16,
"rate_limit_delay": 1.0
},
"reward_shaping_function_parameters": {
"transformation_function": "sigmoid",
"parameters": {
"low": 40,
"high": 90,
"k": 0.15
}
}
},
{
"name": "mw",
"weight": 1,
"preliminary_check": false,
"specific_parameters": {},
"reward_shaping_function_parameters": {
"transformation_function": "double_sigmoid",
"parameters": {
"low": 50,
"high": 400,
"coef_div": 500,
"coef_si": 250,
"coef_se": 250
}
}
}
]
},
"goal_directed_generation": {
"reinforcement_learning": {
"prior": "/home/miquelaperez/saturn/experimental_reproduction/checkpoint_models/zinc-250k-aizynth-purged-epoch-100.prior",
"agent": "/home/miquelaperez/saturn/experimental_reproduction/checkpoint_models/zinc-250k-aizynth-purged-epoch-100.prior",
"batch_size": 16,
"learning_rate": 0.0001,
"sigma": 128.0,
"augmented_memory": true,
"augmentation_rounds": 10,
"selective_memory_purge": true
},
"experience_replay": {
"memory_size": 100,
"sample_size": 10,
"smiles": []
},
"diversity_filter": {
"name": "IdenticalMurckoScaffold",
"bucket_size": 10
},
"hallucinated_memory": {
"execute_hallucinated_memory": false,
"hallucination_method": "ga",
"num_hallucinations": 100,
"num_selected": 5,
"selection_criterion": "random"
},
"beam_enumeration": {
"execute_beam_enumeration": false,
"beam_k": 2,
"beam_steps": 18,
"substructure_type": "structure",
"structure_min_size": 15,
"pool_size": 4,
"pool_saving_frequency": 1000,
"patience": 5,
"token_sampling_method": "topk",
"filter_patience_limit": 100000
}
},
"distribution_learning": {
"parameters": {
"agent": "<unused>",
"training_steps": 20,
"batch_size": 512,
"learning_rate": 0.0001,
"training_dataset_path": "cleaned-chembl-33.smi",
"train_with_randomization": true,
"transfer_learning": false
}
},
"running_mode": "goal_directed_generation",
"model_architecture": "mamba",
"device": "cuda",
"seed": 0
}
Binary file added MBH/mbh_catalyst_run/best_agent.ckpt
View file
Open in desktop
Binary file not shown.
72 changes: 72 additions & 0 deletions MBH/mbh_catalyst_run/log.log
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,72 @@
2026-05-26 14:43:53,278 - INFO - Starting RL generative experiment with oracle budget: 1500
2026-05-26 14:43:55,296 - INFO - Oracle calls: 0/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:43:57,102 - INFO - Oracle calls: 15/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:43:57,820 - INFO - Oracle calls: 30/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:43:58,529 - INFO - Oracle calls: 42/1500 (Batch Validity: 81.25%, 13 / 16)
2026-05-26 14:43:59,344 - INFO - Oracle calls: 55/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:44:00,266 - INFO - Oracle calls: 69/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:44:01,302 - INFO - Oracle calls: 84/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:44:23,982 - INFO - Oracle calls: 98/1500 (Batch Validity: 81.25%, 13 / 16)
2026-05-26 14:44:42,099 - INFO - Logging intermediate results at 111 oracle calls.
2026-05-26 14:44:42,304 - INFO - Oracle calls: 111/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:44:43,381 - INFO - Oracle calls: 127/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:45:01,617 - INFO - Oracle calls: 141/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:45:08,954 - INFO - Oracle calls: 157/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:45:10,008 - INFO - Oracle calls: 170/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:45:32,628 - INFO - Oracle calls: 185/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:45:33,702 - INFO - Oracle calls: 199/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:45:34,621 - INFO - Logging intermediate results at 214 oracle calls.
2026-05-26 14:45:34,794 - INFO - Oracle calls: 214/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:45:35,856 - INFO - Oracle calls: 229/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:45:43,798 - INFO - Oracle calls: 243/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:45:44,858 - INFO - Oracle calls: 259/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:46:06,420 - INFO - Oracle calls: 274/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:46:27,506 - INFO - Oracle calls: 287/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:46:43,118 - INFO - Logging intermediate results at 302 oracle calls.
2026-05-26 14:46:43,319 - INFO - Oracle calls: 302/1500 (Batch Validity: 81.25%, 13 / 16)
2026-05-26 14:46:55,424 - INFO - Oracle calls: 314/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:47:15,288 - INFO - Oracle calls: 330/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:47:32,474 - INFO - Oracle calls: 340/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:48:24,464 - INFO - Oracle calls: 353/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:49:02,077 - INFO - Oracle calls: 367/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:49:03,099 - INFO - Oracle calls: 381/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:50:05,367 - INFO - Oracle calls: 394/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:50:33,046 - INFO - Logging intermediate results at 404 oracle calls.
2026-05-26 14:50:33,184 - INFO - Oracle calls: 404/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:51:09,529 - INFO - Oracle calls: 415/1500 (Batch Validity: 100.0%, 15 / 15)
2026-05-26 14:52:12,112 - INFO - Oracle calls: 425/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:53:03,378 - INFO - Oracle calls: 436/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:54:05,809 - INFO - Oracle calls: 442/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:55:13,607 - INFO - Oracle calls: 450/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:55:47,990 - INFO - Oracle calls: 456/1500 (Batch Validity: 81.25%, 13 / 16)
2026-05-26 14:56:19,982 - INFO - Oracle calls: 458/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 14:56:49,604 - INFO - Oracle calls: 459/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:57:50,046 - INFO - Oracle calls: 465/1500 (Batch Validity: 75.0%, 12 / 16)
2026-05-26 14:58:31,885 - INFO - Oracle calls: 469/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:59:21,403 - INFO - Oracle calls: 472/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 14:59:47,043 - INFO - Oracle calls: 475/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 14:59:47,847 - INFO - Oracle calls: 476/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:00:15,608 - INFO - Oracle calls: 479/1500 (Batch Validity: 100.0%, 15 / 15)
2026-05-26 15:00:37,562 - INFO - Oracle calls: 480/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:00:38,286 - INFO - Oracle calls: 480/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:01:03,499 - INFO - Oracle calls: 481/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:01:29,321 - INFO - Oracle calls: 484/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:02:04,416 - INFO - Oracle calls: 488/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:02:18,495 - INFO - Oracle calls: 489/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:02:41,666 - INFO - Oracle calls: 491/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:03:19,625 - INFO - Oracle calls: 492/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:04:21,521 - INFO - Oracle calls: 497/1500 (Batch Validity: 93.75%, 15 / 16)
2026-05-26 15:05:03,912 - INFO - Oracle calls: 499/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:05:38,765 - INFO - Logging intermediate results at 501 oracle calls.
2026-05-26 15:05:38,885 - INFO - Oracle calls: 501/1500 (Batch Validity: 87.5%, 14 / 16)
2026-05-26 15:06:39,769 - INFO - Oracle calls: 508/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:07:16,890 - INFO - Oracle calls: 516/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:10:04,290 - INFO - Oracle calls: 526/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:11:50,432 - INFO - Oracle calls: 537/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:13:40,653 - INFO - Oracle calls: 547/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:14:25,384 - INFO - Oracle calls: 553/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:15:10,944 - INFO - Oracle calls: 556/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:16:43,319 - INFO - Oracle calls: 565/1500 (Batch Validity: 93.33%, 14 / 15)
2026-05-26 15:17:35,943 - INFO - Oracle calls: 570/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:18:42,964 - INFO - Oracle calls: 579/1500 (Batch Validity: 100.0%, 16 / 16)
2026-05-26 15:20:24,586 - INFO - Oracle calls: 586/1500 (Batch Validity: 100.0%, 16 / 16)
Binary file added MBH/mbh_catalyst_run/mamba_111_agent.ckpt
View file
Open in desktop
Binary file not shown.
Binary file added MBH/mbh_catalyst_run/mamba_214_agent.ckpt
View file
Open in desktop
Binary file not shown.
Binary file added MBH/mbh_catalyst_run/mamba_302_agent.ckpt
View file
Open in desktop
Binary file not shown.
Binary file added MBH/mbh_catalyst_run/mamba_404_agent.ckpt
View file
Open in desktop
Binary file not shown.
Binary file added MBH/mbh_catalyst_run/mamba_501_agent.ckpt
View file
Open in desktop
Binary file not shown.
Loading