Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e4ecf920570cb1c80e8a356cc61da35af23f9765deba1e395998a941feaf60d6
+oid sha256:b6ddc8f2a3677e5e317b80ecd5acf9d1491b0668d5f53b8d9bd9e0d1abbfa2b5
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:763e99e54c0fb4e53f57ab4797450ffab298e7f15107cba1ecac33a76c37f15a
+oid sha256:6d76ad6ce78a4b0037233c4839e647df1579c5d586d5c6854be18638b73cb4cb
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5958f51e039b5dc0df43a65dd50d3d71d5c285cbc46ab6ee2ee402a8334afe3c
+oid sha256:c85698974fe6e7526ea4561e96c9ff34251353ffb22a51e97eb7da83eed5ae17
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:179ad92d70333cf19aaae75def41701f7fcc152b11f9793c71f23796091ee380
+oid sha256:db54f45be63f2a31cb673e60b81aae6b0e3b4b54ab7baae14cb41e91340d7eb8
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.4817498895271763,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=14-learning_rate=3.5e-05-warmup_ratio=0.040-num_cycles=1.000/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.41140079540433055,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=15-learning_rate=2.3e-04-warmup_ratio=0.070-num_cycles=8.500/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": 1.0934492349624634,
-      "learning_rate": 2.5209540295945244e-05,
-      "loss": 2.6469,
+      "grad_norm": 0.5925800800323486,
+      "learning_rate": 0.00016371366993372984,
+      "loss": 1.4564,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.8319031000137329,
-      "learning_rate": 3.487990219078287e-05,
-      "loss": 0.4237,
+      "grad_norm": 0.2535039484500885,
+      "learning_rate": 9.64787889193764e-05,
+      "loss": 0.08,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.5890763998031616,
-      "learning_rate": 3.272552148958062e-05,
-      "loss": 0.1634,
+      "grad_norm": 0.18813341856002808,
+      "learning_rate": 0.00010153394284957266,
+      "loss": 0.0206,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.4209933280944824,
-      "learning_rate": 2.8864268479814478e-05,
-      "loss": 0.0896,
+      "grad_norm": 0.15169543027877808,
+      "learning_rate": 0.00015725808546261587,
+      "loss": 0.0263,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.297393798828125,
-      "learning_rate": 2.3732493981763435e-05,
-      "loss": 0.0571,
+      "grad_norm": 0.10389164090156555,
+      "learning_rate": 4.4808780348991996e-05,
+      "loss": 0.0091,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.27099844813346863,
-      "learning_rate": 1.7910127361891612e-05,
-      "loss": 0.0419,
+      "grad_norm": 0.17846329510211945,
+      "learning_rate": 0.00020580215204037155,
+      "loss": 0.0157,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.2135707437992096,
-      "learning_rate": 1.20551401242989e-05,
-      "loss": 0.0323,
+      "grad_norm": 0.08230599761009216,
+      "learning_rate": 7.977567091364175e-06,
+      "loss": 0.0088,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.22257240116596222,
-      "learning_rate": 6.829190149829874e-06,
-      "loss": 0.0274,
+      "grad_norm": 0.1894439160823822,
+      "learning_rate": 0.00022824090725901466,
+      "loss": 0.0083,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.23932363092899323,
-      "learning_rate": 2.822849346704189e-06,
-      "loss": 0.0242,
+      "grad_norm": 0.061632536351680756,
+      "learning_rate": 1.5637709631532583e-06,
+      "loss": 0.0094,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.1828487515449524,
-      "learning_rate": 4.888645541834838e-07,
-      "loss": 0.023,
+      "grad_norm": 0.1319124698638916,
+      "learning_rate": 0.00021816311730162138,
+      "loss": 0.0049,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.3802916482545294,
-      "eval_e3_equal": 0.7616438356164383,
+      "eval_all_ligands_equal": 0.41140079540433055,
+      "eval_e3_equal": 0.7638532920901459,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.993194874060981,
-      "eval_e3_heavy_atoms_difference": 0.16783031374281926,
-      "eval_e3_heavy_atoms_difference_norm": -0.0030251272862322418,
+      "eval_e3_has_attachment_point(s)": 0.9593460008837826,
+      "eval_e3_heavy_atoms_difference": 1.5961997348652233,
+      "eval_e3_heavy_atoms_difference_norm": 0.039123254458077565,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.993194874060981,
-      "eval_has_all_attachment_points": 0.9793194874060981,
-      "eval_has_three_substructures": 0.9988510826336722,
-      "eval_heavy_atoms_difference": 7.607865665046399,
-      "eval_heavy_atoms_difference_norm": 0.1017810445617049,
-      "eval_linker_equal": 0.5477684489615554,
+      "eval_e3_valid": 0.9593460008837826,
+      "eval_has_all_attachment_points": 0.9722492266902342,
+      "eval_has_three_substructures": 0.9957578435704817,
+      "eval_heavy_atoms_difference": 16.26027397260274,
+      "eval_heavy_atoms_difference_norm": 0.20647141395181035,
+      "eval_linker_equal": 0.6685815289438798,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9951391957578436,
-      "eval_linker_heavy_atoms_difference": 0.5447635881573133,
-      "eval_linker_heavy_atoms_difference_norm": 0.0015117630138471627,
+      "eval_linker_has_attachment_point(s)": 0.9886875828546178,
+      "eval_linker_heavy_atoms_difference": 1.2433937251436147,
+      "eval_linker_heavy_atoms_difference_norm": 0.04663004433191857,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9951391957578436,
-      "eval_loss": 0.2833138108253479,
-      "eval_num_fragments": 3.000353513035793,
-      "eval_poi_equal": 0.7028722934158197,
+      "eval_linker_valid": 0.9886875828546178,
+      "eval_loss": 0.29912540316581726,
+      "eval_num_fragments": 2.9957578435704817,
+      "eval_poi_equal": 0.6661953159522758,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9144498453380469,
-      "eval_poi_heavy_atoms_difference": 2.439946973044631,
-      "eval_poi_heavy_atoms_difference_norm": 0.0796284265989636,
+      "eval_poi_has_attachment_point(s)": 0.852938577110031,
+      "eval_poi_heavy_atoms_difference": 5.4296067167476805,
+      "eval_poi_heavy_atoms_difference_norm": 0.16879143962252643,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9144498453380469,
-      "eval_reassembly": 0.3852408307556341,
-      "eval_reassembly_nostereo": 0.4114891736632788,
-      "eval_runtime": 1513.2312,
-      "eval_samples_per_second": 7.477,
-      "eval_steps_per_second": 0.117,
+      "eval_poi_valid": 0.852938577110031,
+      "eval_reassembly": 0.4201502430402121,
+      "eval_reassembly_nostereo": 0.4426866990720283,
+      "eval_runtime": 1668.2452,
+      "eval_samples_per_second": 6.783,
+      "eval_steps_per_second": 0.106,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9066725585505966,
+      "eval_valid": 0.8162615996464869,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.2371991127729416,
-      "learning_rate": 9.09937223837378e-08,
-      "loss": 0.0225,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.2061084508895874,
-      "learning_rate": 1.6741992672310223e-06,
-      "loss": 0.0224,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.24754801392555237,
-      "learning_rate": 5.059566977090652e-06,
-      "loss": 0.0221,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.216526061296463,
-      "learning_rate": 9.864524673085213e-06,
-      "loss": 0.0215,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.23388701677322388,
-      "learning_rate": 1.554607576142763e-05,
-      "loss": 0.0199,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.39876270437472383,
-      "eval_e3_equal": 0.771011931064958,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.991427308882015,
-      "eval_e3_heavy_atoms_difference": 0.3308882015024304,
-      "eval_e3_heavy_atoms_difference_norm": 0.003835468501410631,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.991427308882015,
-      "eval_has_all_attachment_points": 0.9794962439239947,
-      "eval_has_three_substructures": 0.999116217410517,
-      "eval_heavy_atoms_difference": 8.459301811754308,
-      "eval_heavy_atoms_difference_norm": 0.1126544304512196,
-      "eval_linker_equal": 0.5834732655766681,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.994078656650464,
-      "eval_linker_heavy_atoms_difference": 0.3457357490057446,
-      "eval_linker_heavy_atoms_difference_norm": -0.011478420362631973,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.994078656650464,
-      "eval_loss": 0.2935299277305603,
-      "eval_num_fragments": 3.000883782589483,
-      "eval_poi_equal": 0.692620415377817,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9068493150684932,
-      "eval_poi_heavy_atoms_difference": 2.849138311975254,
-      "eval_poi_heavy_atoms_difference_norm": 0.09139285131698564,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9068493150684932,
-      "eval_reassembly": 0.40459566946531156,
-      "eval_reassembly_nostereo": 0.4296067167476801,
-      "eval_runtime": 1508.1015,
-      "eval_samples_per_second": 7.503,
-      "eval_steps_per_second": 0.117,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8958020326999558,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.22054216265678406,
-      "learning_rate": 2.14621619590723e-05,
-      "loss": 0.0184,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.2767944931983948,
-      "learning_rate": 2.694422074016186e-05,
-      "loss": 0.0165,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.2199663668870926,
-      "learning_rate": 3.137273796255239e-05,
-      "loss": 0.0139,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.1930292844772339,
-      "learning_rate": 3.424725765567624e-05,
-      "loss": 0.0117,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.19016879796981812,
-      "learning_rate": 3.524293733373145e-05,
-      "loss": 0.0107,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.4817498895271763,
-      "eval_e3_equal": 0.7871851524524967,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.993194874060981,
-      "eval_e3_heavy_atoms_difference": 0.185505965532479,
-      "eval_e3_heavy_atoms_difference_norm": -0.0041066829710338,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.993194874060981,
-      "eval_has_all_attachment_points": 0.9818824569155987,
-      "eval_has_three_substructures": 0.9984975695978789,
-      "eval_heavy_atoms_difference": 8.25974370304905,
-      "eval_heavy_atoms_difference_norm": 0.10874206514539499,
-      "eval_linker_equal": 0.6921785240830756,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.993813521873619,
-      "eval_linker_heavy_atoms_difference": 0.5291206363234644,
-      "eval_linker_heavy_atoms_difference_norm": 0.011250318578223506,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.993813521873619,
-      "eval_loss": 0.3034479022026062,
-      "eval_num_fragments": 3.0004418912947415,
-      "eval_poi_equal": 0.7307114449845338,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9071144498453381,
-      "eval_poi_heavy_atoms_difference": 2.7717189571365446,
-      "eval_poi_heavy_atoms_difference_norm": 0.08902196954398302,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9071144498453381,
-      "eval_reassembly": 0.4891736632788334,
-      "eval_reassembly_nostereo": 0.5138311975254087,
-      "eval_runtime": 1507.2599,
-      "eval_samples_per_second": 7.507,
-      "eval_steps_per_second": 0.117,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.898806893504198,
-      "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=14-learning_rate=3.5e-05-warmup_ratio=0.040-num_cycles=1.000",
+  "trial_name": "trial-number=15-learning_rate=2.3e-04-warmup_ratio=0.070-num_cycles=8.500",
   "trial_params": {
-    "learning_rate": 3.524293733373145e-05,
-    "num_cycles": 1.0,
-    "warmup_ratio": 0.04
+    "learning_rate": 0.00022887171056735432,
+    "num_cycles": 8.5,
+    "warmup_ratio": 0.06999999999999999
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:df98ec59c1f28c04630ec1fd68e1ce4551f90a6c990680ba52fb5d3fc63b8693
+oid sha256:a7ab3268adb85b02359ec17e701405643275b8826a21c0e929b050857cbff37e
 size 7416