Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:955661c6c98d738e6d93e4baefe143e26050501513c94695dcac2d6a24c97b24
+oid sha256:ec346ca69d4812dbdf7f98771ee671825861eeeebc9002629664edc74b4ae44a
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:135742922f90c9aa6caf4db4ac73fc43f31b802a5457a7aac2f5d4c29703a917
+oid sha256:e1e1d2a38d4c84663c4d57cb115803a52fb08d896333fc9a59322173bebf08c9
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:201f0edb7cc68926f375d0324afde9eee7d61f206f8cfb247c057e05f32ca508
+oid sha256:abc995446869d12e0af72f243ebb9dda0e042250caa7ad7c6497c840ba43f934
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c9a7f72cf8bfa17a6359f7c182fea3f4c09b3e10e107c5a971789a8549c593ba
+oid sha256:d925c5fe16ac296f1170b611f090ec75101b814c00f93070017194f62dd64053
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.0,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=7-learning_rate=7.8e-04-warmup_ratio=0.090-num_cycles=2.000/checkpoint-5000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.0002651347768448962,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=8-learning_rate=6.7e-06-warmup_ratio=0.040-num_cycles=6.500/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": 0.5237686038017273,
-      "learning_rate": 0.0005572089662714916,
-      "loss": 1.0925,
+      "grad_norm": 1.609957218170166,
+      "learning_rate": 4.764287234716251e-06,
+      "loss": 4.427,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.3377338945865631,
-      "learning_rate": 0.0007472119550121175,
-      "loss": 0.2098,
+      "grad_norm": 1.3062211275100708,
+      "learning_rate": 4.151261952929334e-06,
+      "loss": 1.1114,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.20445334911346436,
-      "learning_rate": 0.0005723056787112626,
-      "loss": 0.1766,
+      "grad_norm": 1.2533338069915771,
+      "learning_rate": 2.321578953844436e-07,
+      "loss": 0.7793,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.22173060476779938,
-      "learning_rate": 0.00031709545642021487,
-      "loss": 0.1621,
+      "grad_norm": 1.1933494806289673,
+      "learning_rate": 6.133081099363204e-06,
+      "loss": 0.7138,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.3483358323574066,
-      "learning_rate": 9.368482054050638e-05,
-      "loss": 0.1547,
+      "grad_norm": 1.1922088861465454,
+      "learning_rate": 3.1497864683804228e-06,
+      "loss": 0.5344,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.3770233988761902,
-      "learning_rate": 2.0903083311035306e-07,
-      "loss": 0.1498,
+      "grad_norm": 1.0372626781463623,
+      "learning_rate": 7.384678968603421e-07,
+      "loss": 0.451,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.35748037695884705,
-      "learning_rate": 7.772822245308111e-05,
-      "loss": 0.1489,
+      "grad_norm": 0.9294341206550598,
+      "learning_rate": 6.542318154539015e-06,
+      "loss": 0.4202,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.23404386639595032,
-      "learning_rate": 0.0002921913486535547,
-      "loss": 0.1537,
+      "grad_norm": 0.7656099796295166,
+      "learning_rate": 2.164785723897086e-06,
+      "loss": 0.3395,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.2566510736942291,
-      "learning_rate": 0.0005493934289242663,
-      "loss": 0.1764,
+      "grad_norm": 0.7425525188446045,
+      "learning_rate": 1.481401038093183e-06,
+      "loss": 0.3034,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.5120010375976562,
-      "learning_rate": 0.0007351513096037483,
-      "loss": 0.3178,
+      "grad_norm": 0.9094532132148743,
+      "learning_rate": 6.658298821578928e-06,
+      "loss": 0.2848,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.01794078656650464,
+      "eval_all_ligands_equal": 0.0002651347768448962,
+      "eval_e3_equal": 0.17516570923552807,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9992045956694653,
-      "eval_e3_heavy_atoms_difference": 9.66911179849757,
-      "eval_e3_heavy_atoms_difference_norm": 0.27981822246010174,
+      "eval_e3_has_attachment_point(s)": 0.9882456915598763,
+      "eval_e3_heavy_atoms_difference": -0.993813521873619,
+      "eval_e3_heavy_atoms_difference_norm": -0.06273710988010914,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9992045956694653,
-      "eval_has_all_attachment_points": 0.6350861688024746,
-      "eval_has_three_substructures": 0.9995581087052585,
-      "eval_heavy_atoms_difference": 47.50543526292532,
-      "eval_heavy_atoms_difference_norm": 0.6127405949371868,
-      "eval_linker_equal": 0.0,
+      "eval_e3_valid": 0.9882456915598763,
+      "eval_has_all_attachment_points": 0.8872293415819709,
+      "eval_has_three_substructures": 0.9905435262925321,
+      "eval_heavy_atoms_difference": 36.26442775077331,
+      "eval_heavy_atoms_difference_norm": 0.46364946102044235,
+      "eval_linker_equal": 0.0036235086168802475,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.6439239946973044,
-      "eval_linker_heavy_atoms_difference": 11.617764030048608,
-      "eval_linker_heavy_atoms_difference_norm": 0.6387309032815724,
+      "eval_linker_has_attachment_point(s)": 0.979584622182943,
+      "eval_linker_heavy_atoms_difference": 6.370835174547062,
+      "eval_linker_heavy_atoms_difference_norm": 0.2822560976043188,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.6439239946973044,
-      "eval_loss": 0.8429178595542908,
-      "eval_num_fragments": 2.9995581087052585,
-      "eval_poi_equal": 0.0010605391073795847,
+      "eval_linker_valid": 0.979584622182943,
+      "eval_loss": 0.42363283038139343,
+      "eval_num_fragments": 2.9927529827662394,
+      "eval_poi_equal": 0.10720282810428634,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.8980114891736632,
-      "eval_poi_heavy_atoms_difference": 12.632434821034026,
-      "eval_poi_heavy_atoms_difference_norm": 0.3178192566767846,
+      "eval_poi_has_attachment_point(s)": 0.6131683605832965,
+      "eval_poi_heavy_atoms_difference": 14.577286787450287,
+      "eval_poi_heavy_atoms_difference_norm": 0.42516346148688866,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.8980114891736632,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 918.5796,
-      "eval_samples_per_second": 12.318,
-      "eval_steps_per_second": 0.193,
+      "eval_poi_valid": 0.6131683605832965,
+      "eval_reassembly": 0.00035351303579319487,
+      "eval_reassembly_nostereo": 0.00035351303579319487,
+      "eval_runtime": 1482.7962,
+      "eval_samples_per_second": 7.631,
+      "eval_steps_per_second": 0.119,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.606805125939019,
+      "eval_valid": 0.6063632346442775,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.4076462388038635,
-      "learning_rate": 0.0007714765562393852,
-      "loss": 0.1758,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.2280195951461792,
-      "learning_rate": 0.0006400101429446236,
-      "loss": 0.1733,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.210994690656662,
-      "learning_rate": 0.00039849994984040654,
-      "loss": 0.1577,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.26513564586639404,
-      "learning_rate": 0.00015303164040412638,
-      "loss": 0.151,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.24093259871006012,
-      "learning_rate": 1.1429518589633372e-05,
-      "loss": 0.1472,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.007335395492708794,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 1.0,
-      "eval_e3_heavy_atoms_difference": -2.7674768007070263,
-      "eval_e3_heavy_atoms_difference_norm": -0.20737054542073902,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 1.0,
-      "eval_has_all_attachment_points": 1.0,
-      "eval_has_three_substructures": 1.0,
-      "eval_heavy_atoms_difference": -17.350950066283694,
-      "eval_heavy_atoms_difference_norm": -0.27758236918999235,
-      "eval_linker_equal": 0.0017675651789659744,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9999116217410517,
-      "eval_linker_heavy_atoms_difference": -1.4151126822801592,
-      "eval_linker_heavy_atoms_difference_norm": -0.3403509540648143,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9999116217410517,
-      "eval_loss": 0.483898401260376,
-      "eval_num_fragments": 3.0,
-      "eval_poi_equal": 0.001590808661069377,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9999116217410517,
-      "eval_poi_heavy_atoms_difference": -13.178877596111356,
-      "eval_poi_heavy_atoms_difference_norm": -0.6323580421785433,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9999116217410517,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 1662.4049,
-      "eval_samples_per_second": 6.806,
-      "eval_steps_per_second": 0.106,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9998232434821034,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.23720437288284302,
-      "learning_rate": 3.589367147985439e-05,
-      "loss": 0.146,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.2926724851131439,
-      "learning_rate": 0.0002156779858055608,
-      "loss": 0.1476,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.48496443033218384,
-      "learning_rate": 0.0004718104810348968,
-      "loss": 0.1646,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 1.0706260204315186,
-      "learning_rate": 0.000691782507552403,
-      "loss": 0.1702,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.436310738325119,
-      "learning_rate": 0.0007789692476655757,
-      "loss": 0.1936,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.009544851966416261,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 1.0,
-      "eval_e3_heavy_atoms_difference": 0.06575342465753424,
-      "eval_e3_heavy_atoms_difference_norm": -0.09551280289451732,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 1.0,
-      "eval_has_all_attachment_points": 1.0,
-      "eval_has_three_substructures": 1.0,
-      "eval_heavy_atoms_difference": 9.107821475916925,
-      "eval_heavy_atoms_difference_norm": 0.0856541453235094,
-      "eval_linker_equal": 0.0017675651789659744,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 1.0,
-      "eval_linker_heavy_atoms_difference": 0.8593018117543084,
-      "eval_linker_heavy_atoms_difference_norm": -0.16998449507864719,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 1.0,
-      "eval_loss": 0.5665163993835449,
-      "eval_num_fragments": 3.0,
-      "eval_poi_equal": 0.0020326999558108703,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.984887317719841,
-      "eval_poi_heavy_atoms_difference": 7.545558992487848,
-      "eval_poi_heavy_atoms_difference_norm": 0.1311353254304406,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.984887317719841,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 1131.7915,
-      "eval_samples_per_second": 9.997,
-      "eval_steps_per_second": 0.156,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.984887317719841,
-      "step": 10000
     }
   ],
   "logging_steps": 500,
@@ -287,17 +133,17 @@
         "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=7-learning_rate=7.8e-04-warmup_ratio=0.090-num_cycles=2.000",
+  "trial_name": "trial-number=8-learning_rate=6.7e-06-warmup_ratio=0.040-num_cycles=6.500",
   "trial_params": {
-    "learning_rate": 0.0007789781348475454,
-    "num_cycles": 2.0,
-    "warmup_ratio": 0.09
+    "learning_rate": 6.660473554133319e-06,
+    "num_cycles": 6.5,
+    "warmup_ratio": 0.04
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3223531eb3089e28f2c8c8e18885635a559b2dd0dc9fd431293475408c45439
+oid sha256:a4cb0b8bc2a80b08ffdf68affeeaab135998229dbd41f6a5196fbc13cd8d217b
 size 7416