Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c68f9cc8640a1f2750b15328a4cb174ace0f1e564e2297a8c2b4f5a0e74c737
+oid sha256:c8fc98e3feff45adbfecc20c1e25564c6fab000d9f738e8e9feeeb22937177f8
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:236d0d0e1d3d3bd11456cd16c52bec8bb6636a8bb5401ecc809d9803cd2d91f7
+oid sha256:dfa2f99fcfea5b880b3cdff501b6e446867748078e92f01efcf8350774632a4e
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffe5e77a84d9bf1e48157505bdf2cce2d248a92b34e9ba74ee38f9fb52ce2ac0
+oid sha256:b011a31d42c1446fe5cc11f0c3078bae2b04dae6c0d5ab945e989a90dfcde070
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:53dcf2f4afbc4e86edee9525058b1f129fb67afb5b9d175ca9d0d6ef33304a5b
+oid sha256:418250b00e984c16e542a3795ab0a0ea5ca42ee5c239316091a0ed085900733e
 size 1192

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.5588157313300928,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_reduce-opt25/trial-number=19-learning_rate=8.0e-05-warmup_ratio=0.060-min_lr=0.000-factor=0.970/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.026513477684489615,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_reduce-opt25/trial-number=20-learning_rate=8.1e-06-warmup_ratio=0.020-min_lr=0.000-factor=0.870/checkpoint-5000",
+  "epoch": 0.49319392385085814,
   "eval_steps": 2500,
-  "global_step": 10000,
+  "global_step": 5000,
   "is_hyper_param_search": true,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.049319392385085814,
-      "grad_norm": 0.48272669315338135,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.8167,
+      "grad_norm": 1.1766011714935303,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 1.9236,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.35677370429039,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.1099,
+      "grad_norm": 1.0651462078094482,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.7026,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.28323549032211304,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0399,
+      "grad_norm": 0.940447986125946,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.4596,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.18444126844406128,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.022,
+      "grad_norm": 0.8968250751495361,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.3322,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.1738751381635666,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0141,
+      "grad_norm": 0.859994113445282,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.2537,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.14796946942806244,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0107,
+      "grad_norm": 0.6769737005233765,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.2037,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.13383328914642334,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0089,
+      "grad_norm": 0.3946596682071686,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.1682,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.10535630583763123,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0073,
+      "grad_norm": 0.5438293218612671,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.1438,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.11387604475021362,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0062,
+      "grad_norm": 0.48131752014160156,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.1234,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.11770522594451904,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0059,
+      "grad_norm": 0.442853718996048,
+      "learning_rate": 8.081437055445882e-06,
+      "loss": 0.1065,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.5199292973928413,
-      "eval_e3_equal": 0.7980556783031374,
+      "eval_all_ligands_equal": 0.026513477684489615,
+      "eval_e3_equal": 0.5127706584180292,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.976049491825011,
-      "eval_e3_heavy_atoms_difference": 0.9379584622182943,
-      "eval_e3_heavy_atoms_difference_norm": 0.02296914272883557,
+      "eval_e3_has_attachment_point(s)": 0.9961997348652232,
+      "eval_e3_heavy_atoms_difference": -0.4975695978789218,
+      "eval_e3_heavy_atoms_difference_norm": -0.03454149824526644,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.976049491825011,
-      "eval_has_all_attachment_points": 0.9775519222271322,
-      "eval_has_three_substructures": 0.9992045956694653,
-      "eval_heavy_atoms_difference": 8.915245249668581,
-      "eval_heavy_atoms_difference_norm": 0.11359321330061124,
-      "eval_linker_equal": 0.7530711444984534,
+      "eval_e3_valid": 0.9961997348652232,
+      "eval_has_all_attachment_points": 0.9762262483429076,
+      "eval_has_three_substructures": 0.9978789217852408,
+      "eval_heavy_atoms_difference": 12.464074237737517,
+      "eval_heavy_atoms_difference_norm": 0.1621929702774956,
+      "eval_linker_equal": 0.05983208130799823,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9946973044631021,
-      "eval_linker_heavy_atoms_difference": 0.3158638974812196,
-      "eval_linker_heavy_atoms_difference_norm": -0.005294061039033342,
+      "eval_linker_has_attachment_point(s)": 0.9939902783915157,
+      "eval_linker_heavy_atoms_difference": 3.101193106495802,
+      "eval_linker_heavy_atoms_difference_norm": 0.126932032645133,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9946973044631021,
-      "eval_loss": 0.29772570729255676,
-      "eval_num_fragments": 2.9995581087052585,
-      "eval_poi_equal": 0.7513919575784357,
+      "eval_linker_valid": 0.9939902783915157,
+      "eval_loss": 0.32072311639785767,
+      "eval_num_fragments": 2.9984091913389306,
+      "eval_poi_equal": 0.5090587715422006,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9204595669465312,
-      "eval_poi_heavy_atoms_difference": 2.561201944321697,
-      "eval_poi_heavy_atoms_difference_norm": 0.07901894293442624,
+      "eval_poi_has_attachment_point(s)": 0.8727353071144498,
+      "eval_poi_heavy_atoms_difference": 4.03844454264251,
+      "eval_poi_heavy_atoms_difference_norm": 0.12411017698653612,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9204595669465312,
-      "eval_reassembly": 0.5268228015908086,
-      "eval_reassembly_nostereo": 0.5501546619531595,
-      "eval_runtime": 1487.2402,
-      "eval_samples_per_second": 7.608,
-      "eval_steps_per_second": 0.119,
+      "eval_poi_valid": 0.8727353071144498,
+      "eval_reassembly": 0.027308882015024304,
+      "eval_reassembly_nostereo": 0.033318603623508615,
+      "eval_runtime": 1527.8596,
+      "eval_samples_per_second": 7.406,
+      "eval_steps_per_second": 0.116,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8945647370746797,
+      "eval_valid": 0.8687582854617764,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.07127052545547485,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0051,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.0946563333272934,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.005,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.06068830564618111,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0057,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.08396610617637634,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0034,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.08121811598539352,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0033,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.552364118426867,
-      "eval_e3_equal": 0.8115775519222271,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9831197525408749,
-      "eval_e3_heavy_atoms_difference": 0.7549270879363676,
-      "eval_e3_heavy_atoms_difference_norm": 0.020315382284783225,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9831197525408749,
-      "eval_has_all_attachment_points": 0.9861246133451171,
-      "eval_has_three_substructures": 0.9994697304463102,
-      "eval_heavy_atoms_difference": 8.231462660185594,
-      "eval_heavy_atoms_difference_norm": 0.10942372385980337,
-      "eval_linker_equal": 0.7871851524524967,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.996464869642068,
-      "eval_linker_heavy_atoms_difference": 0.4628369421122404,
-      "eval_linker_heavy_atoms_difference_norm": 0.008046427004901976,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.996464869642068,
-      "eval_loss": 0.3040188252925873,
-      "eval_num_fragments": 2.9998232434821035,
-      "eval_poi_equal": 0.7715422006186479,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9170128148475475,
-      "eval_poi_heavy_atoms_difference": 2.2678745028722935,
-      "eval_poi_heavy_atoms_difference_norm": 0.07297810777653783,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9170128148475475,
-      "eval_reassembly": 0.5604949182501104,
-      "eval_reassembly_nostereo": 0.5799381352187362,
-      "eval_runtime": 1492.6054,
-      "eval_samples_per_second": 7.581,
-      "eval_steps_per_second": 0.119,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8988952717631463,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.07443885505199432,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0032,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.10396549850702286,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0033,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.0768752321600914,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0041,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.07676319032907486,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0031,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.05903030186891556,
-      "learning_rate": 8.027431614642607e-05,
-      "loss": 0.0025,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.5588157313300928,
-      "eval_e3_equal": 0.8097216084843128,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9815289438798056,
-      "eval_e3_heavy_atoms_difference": 0.6401237295625276,
-      "eval_e3_heavy_atoms_difference_norm": 0.017142329262750198,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9815289438798056,
-      "eval_has_all_attachment_points": 0.984003535130358,
-      "eval_has_three_substructures": 0.999116217410517,
-      "eval_heavy_atoms_difference": 7.976314626601856,
-      "eval_heavy_atoms_difference_norm": 0.10663825505650316,
-      "eval_linker_equal": 0.795581087052585,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9961113566062748,
-      "eval_linker_heavy_atoms_difference": 0.3717189571365444,
-      "eval_linker_heavy_atoms_difference_norm": 0.00716642746514771,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9961113566062748,
-      "eval_loss": 0.29810017347335815,
-      "eval_num_fragments": 2.9992929739284135,
-      "eval_poi_equal": 0.7748121961997348,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9217852408307556,
-      "eval_poi_heavy_atoms_difference": 2.4131683605832963,
-      "eval_poi_heavy_atoms_difference_norm": 0.07863735459718353,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9217852408307556,
-      "eval_reassembly": 0.5674768007070261,
-      "eval_reassembly_nostereo": 0.5911621741051701,
-      "eval_runtime": 1506.4595,
-      "eval_samples_per_second": 7.511,
-      "eval_steps_per_second": 0.117,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9027839151568714,
-      "step": 10000
     }
   ],
   "logging_steps": 500,
@@ -287,18 +133,18 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": true
+        "should_training_stop": false
       },
       "attributes": {}
     }
   },
-  "total_flos": 8.663036348367667e+16,
+  "total_flos": 4.33139477262336e+16,
   "train_batch_size": 128,
-  "trial_name": "trial-number=19-learning_rate=8.0e-05-warmup_ratio=0.060-min_lr=0.000-factor=0.970",
+  "trial_name": "trial-number=20-learning_rate=8.1e-06-warmup_ratio=0.020-min_lr=0.000-factor=0.870",
   "trial_params": {
-    "factor": 0.9700000000000001,
-    "learning_rate": 8.027431614642607e-05,
-    "min_lr": 7.655589150964243e-09,
-    "warmup_ratio": 0.06
+    "factor": 0.8700000000000001,
+    "learning_rate": 8.081437055445882e-06,
+    "min_lr": 1.3978253280193875e-10,
+    "warmup_ratio": 0.02
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fd9e4a725db2ab30f7bf911ea9e6be239c00d1942a0bc3faad39f344c30b894
+oid sha256:c93e6ebe2707647fc37e4303390f6b91ca786dcbba0dcd422274b964b6f7a485
 size 7480