Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b9caff680ded8301b1e34260bc544f4f79635509d4edd8e813cff97fe71e7f6
+oid sha256:361ca0022dcbd0d5654b9ba2b3aef8e0ae68184389d7dedca727f54f6f8bfa84
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:671e7bc8640e10f662af56c77d3879268f1045aecda3b15f8fc57a7184f200d7
+oid sha256:44afa5fcf76002bc78cafdc99be6637f55f1c70be2906e6d44a465f7b5f06e35
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d397309a5ed22a28c120b9463f5011728fe60adfb800dac9b6ca1b26956108dc
+oid sha256:417d1e53bedb4ee47348e21b98c7f8adcaf9a166a388a8b6943a6900823b2903
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98090ec77dec52540b56df32c47f3b896bab1a65be66d7c9e0ca547ee4982d94
+oid sha256:b402addbfff013beeab19332ef7bc29a35392e2911e8d4d8915e4d3e63195a56
 size 1192

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.5124171453822359,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_reduce-opt25/trial-number=8-learning_rate=2.0e-05-warmup_ratio=0.020-min_lr=0.000-factor=0.820/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.20972160848431287,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_reduce-opt25/trial-number=9-learning_rate=1.3e-05-warmup_ratio=0.050-min_lr=0.000-factor=0.890/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": 1.0092626810073853,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 1.3581,
+      "grad_norm": 1.0882785320281982,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 1.6004,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.6695255637168884,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.3853,
+      "grad_norm": 0.7874086499214172,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.5219,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.5797414183616638,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.2014,
+      "grad_norm": 0.7524672150611877,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.3065,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.49628201127052307,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.1258,
+      "grad_norm": 0.5783708691596985,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.2066,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.3971586525440216,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0839,
+      "grad_norm": 0.669140100479126,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.151,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.30195024609565735,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0596,
+      "grad_norm": 0.4841144382953644,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.1148,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.2556217610836029,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0446,
+      "grad_norm": 0.34032905101776123,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.0882,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.2777057886123657,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0349,
+      "grad_norm": 0.32155144214630127,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.0698,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.246878519654274,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.028,
+      "grad_norm": 0.38137614727020264,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.056,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.27177679538726807,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0231,
+      "grad_norm": 0.39012080430984497,
+      "learning_rate": 1.287188117510188e-05,
+      "loss": 0.0469,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.3915156871409633,
-      "eval_e3_equal": 0.7765797613787009,
+      "eval_all_ligands_equal": 0.20972160848431287,
+      "eval_e3_equal": 0.7072028281042864,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9932832523199293,
-      "eval_e3_heavy_atoms_difference": 0.2896155545735749,
-      "eval_e3_heavy_atoms_difference_norm": 0.0020662380654388832,
+      "eval_e3_has_attachment_point(s)": 0.9947856827220504,
+      "eval_e3_heavy_atoms_difference": 0.09491825011047282,
+      "eval_e3_heavy_atoms_difference_norm": -0.006944252067541642,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9932832523199293,
-      "eval_has_all_attachment_points": 0.9840919133893062,
-      "eval_has_three_substructures": 0.9990278391515687,
-      "eval_heavy_atoms_difference": 8.075386654882898,
-      "eval_heavy_atoms_difference_norm": 0.10478609381959636,
-      "eval_linker_equal": 0.563323022536456,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9960229783473266,
-      "eval_linker_heavy_atoms_difference": 0.9376049491825011,
-      "eval_linker_heavy_atoms_difference_norm": 0.02394921771244689,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9960229783473266,
-      "eval_loss": 0.286180704832077,
-      "eval_num_fragments": 2.9999116217410515,
-      "eval_poi_equal": 0.7011047282368538,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9188687582854618,
-      "eval_poi_heavy_atoms_difference": 2.6504639858594787,
-      "eval_poi_heavy_atoms_difference_norm": 0.0799410695670892,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9188687582854618,
-      "eval_reassembly": 0.39752540874944764,
-      "eval_reassembly_nostereo": 0.41688024745912505,
-      "eval_runtime": 1468.607,
-      "eval_samples_per_second": 7.705,
-      "eval_steps_per_second": 0.121,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9114449845338047,
-      "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.2606898248195648,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0194,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.22691254317760468,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0162,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.17791932821273804,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0138,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.21223701536655426,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0121,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.13272830843925476,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0104,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.4880247459125055,
-      "eval_e3_equal": 0.7960229783473266,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9928413610251878,
-      "eval_e3_heavy_atoms_difference": 0.21272646928855501,
-      "eval_e3_heavy_atoms_difference_norm": -0.0006436019849305715,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9928413610251878,
-      "eval_has_all_attachment_points": 0.981352187361909,
+      "eval_e3_valid": 0.9947856827220504,
+      "eval_has_all_attachment_points": 0.9820592134334953,
       "eval_has_three_substructures": 0.9984975695978789,
-      "eval_heavy_atoms_difference": 7.0852850198851085,
-      "eval_heavy_atoms_difference_norm": 0.09377076700849872,
-      "eval_linker_equal": 0.6921785240830756,
+      "eval_heavy_atoms_difference": 9.720901458241272,
+      "eval_heavy_atoms_difference_norm": 0.12639386628820087,
+      "eval_linker_equal": 0.33548387096774196,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9941670349094123,
-      "eval_linker_heavy_atoms_difference": 0.3749005744586832,
-      "eval_linker_heavy_atoms_difference_norm": -0.0022863167643088684,
+      "eval_linker_has_attachment_point(s)": 0.9950508174988952,
+      "eval_linker_heavy_atoms_difference": 1.5590808661069377,
+      "eval_linker_heavy_atoms_difference_norm": 0.05308081409891352,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9941670349094123,
-      "eval_loss": 0.29602476954460144,
-      "eval_num_fragments": 3.0004418912947415,
-      "eval_poi_equal": 0.7327441449403447,
+      "eval_linker_valid": 0.9950508174988952,
+      "eval_loss": 0.2932600677013397,
+      "eval_num_fragments": 2.999381352187362,
+      "eval_poi_equal": 0.636411842686699,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9255855059655325,
-      "eval_poi_heavy_atoms_difference": 2.2584180291648255,
-      "eval_poi_heavy_atoms_difference_norm": 0.07375030609713656,
+      "eval_poi_has_attachment_point(s)": 0.9011047282368537,
+      "eval_poi_heavy_atoms_difference": 3.2722934158197083,
+      "eval_poi_heavy_atoms_difference_norm": 0.1006642870975325,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9255855059655325,
-      "eval_reassembly": 0.4942996022978347,
-      "eval_reassembly_nostereo": 0.5196641626159965,
-      "eval_runtime": 1488.7706,
-      "eval_samples_per_second": 7.6,
-      "eval_steps_per_second": 0.119,
+      "eval_poi_valid": 0.9011047282368537,
+      "eval_reassembly": 0.2139637649138312,
+      "eval_reassembly_nostereo": 0.23932832523199293,
+      "eval_runtime": 1512.9908,
+      "eval_samples_per_second": 7.479,
+      "eval_steps_per_second": 0.117,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9170128148475475,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.1938847005367279,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0098,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.1654110550880432,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0082,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.20306991040706635,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0075,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.18328730762004852,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0068,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.19016805291175842,
-      "learning_rate": 1.9500595004156286e-05,
-      "loss": 0.0062,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.5124171453822359,
-      "eval_e3_equal": 0.7961997348652231,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9913389306230668,
-      "eval_e3_heavy_atoms_difference": 0.4334069818824569,
-      "eval_e3_heavy_atoms_difference_norm": 0.008561304108695537,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9913389306230668,
-      "eval_has_all_attachment_points": 0.9805567830313743,
-      "eval_has_three_substructures": 0.9982324348210341,
-      "eval_heavy_atoms_difference": 8.227308882015024,
-      "eval_heavy_atoms_difference_norm": 0.1080935669489637,
-      "eval_linker_equal": 0.7361908970393283,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9947856827220504,
-      "eval_linker_heavy_atoms_difference": 0.6195315952275741,
-      "eval_linker_heavy_atoms_difference_norm": 0.019850974836285937,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9947856827220504,
-      "eval_loss": 0.30660033226013184,
-      "eval_num_fragments": 2.99946973044631,
-      "eval_poi_equal": 0.7485638532920902,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9164825452938578,
-      "eval_poi_heavy_atoms_difference": 2.7237295625276183,
-      "eval_poi_heavy_atoms_difference_norm": 0.08457533862065081,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9164825452938578,
-      "eval_reassembly": 0.5210782147591693,
-      "eval_reassembly_nostereo": 0.5454706142288996,
-      "eval_runtime": 1474.9305,
-      "eval_samples_per_second": 7.672,
-      "eval_steps_per_second": 0.12,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9076447193990278,
-      "step": 10000
+      "eval_valid": 0.8955368979231109,
+      "step": 5000
     }
   ],
   "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=8-learning_rate=2.0e-05-warmup_ratio=0.020-min_lr=0.000-factor=0.820",
+  "trial_name": "trial-number=9-learning_rate=1.3e-05-warmup_ratio=0.050-min_lr=0.000-factor=0.890",
   "trial_params": {
-    "factor": 0.8200000000000001,
-    "learning_rate": 1.9500595004156286e-05,
-    "min_lr": 1.163109026324453e-09,
-    "warmup_ratio": 0.02
+    "factor": 0.89,
+    "learning_rate": 1.287188117510188e-05,
+    "min_lr": 7.953202892241938e-10,
+    "warmup_ratio": 0.05
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:041278698216885e2a0cc9e60e59109a8f1a4d27ade63a2779a5251e3e4da60c
+oid sha256:8f14039046a82de020163ad9a293b54d8c44c1c489589672206495ade799b622
 size 7480