Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d42eb3cc3676aa5ab3aba53c2f41a55a65a12a81571a83830360833de78681cb
+oid sha256:f5145c2cc7d97ae3d9a124fc843bd98b3ed8f6f8784b198e6ff821ea49461e95
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d857951e70a8a0628eb3e8bc8d0d7413567222c40804a0c5602d7980cd82395e
+oid sha256:a7abd2543c0f4161003949ca3b0a09f32ee52189eb6cf1ce1b0d339218db56fb
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc2928b365c0ef1120237fc40be202f53b478c2918d560736b898e24618aee8b
+oid sha256:56116cd4c05fa65ee5b68ac09ebd53a9cecc2c132b0229f70f547f950f24eb44
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5442f58e6023536ebb399e7159e58225a0952bc1b64060575c79ec6328ce14bb
+oid sha256:b8cec5d697af01fd16026abee650745e0dfef8562ef22eba2f073c03e205449e
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.5703049049933716,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts-opt25/trial-number=11-learning_rate=1.0e-04-warmup_ratio=0.080-num_cycles=3/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.0,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts-opt25/trial-number=12-learning_rate=2.2e-06-warmup_ratio=0.050-num_cycles=10/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": 0.6575852632522583,
-      "learning_rate": 0.000100182631515627,
-      "loss": 1.1484,
+      "grad_norm": 2.3763999938964844,
+      "learning_rate": 1.4154648142539406e-06,
+      "loss": 3.9782,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.343759149312973,
-      "learning_rate": 8.597337690594066e-05,
-      "loss": 0.1449,
+      "grad_norm": 2.259150743484497,
+      "learning_rate": 4.420955339237543e-08,
+      "loss": 1.6474,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.2601257562637329,
-      "learning_rate": 6.420018709015348e-05,
-      "loss": 0.0439,
+      "grad_norm": 1.417884111404419,
+      "learning_rate": 1.3822352979831102e-06,
+      "loss": 1.4116,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.13981187343597412,
-      "learning_rate": 3.9703875822711815e-05,
-      "loss": 0.0221,
+      "grad_norm": 1.4686042070388794,
+      "learning_rate": 3.498093351576883e-08,
+      "loss": 1.1769,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.1797427237033844,
-      "learning_rate": 1.793068600692463e-05,
-      "loss": 0.0141,
+      "grad_norm": 1.374851942062378,
+      "learning_rate": 1.348712939813865e-06,
+      "loss": 1.07,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.13262967765331268,
-      "learning_rate": 3.721431397238288e-06,
-      "loss": 0.011,
+      "grad_norm": 1.3333460092544556,
+      "learning_rate": 2.681604268390353e-08,
+      "loss": 0.9419,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.24220901727676392,
-      "learning_rate": 0.00010366882009194501,
-      "loss": 0.0124,
+      "grad_norm": 1.3361984491348267,
+      "learning_rate": 1.3149314939193131e-06,
+      "loss": 0.8707,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.15627698600292206,
-      "learning_rate": 9.565686145691327e-05,
-      "loss": 0.0201,
+      "grad_norm": 1.3137452602386475,
+      "learning_rate": 1.97231022497391e-08,
+      "loss": 0.7844,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.12300624698400497,
-      "learning_rate": 7.792804718464899e-05,
-      "loss": 0.0093,
+      "grad_norm": 1.387094259262085,
+      "learning_rate": 1.2809249753519381e-06,
+      "loss": 0.7308,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.09846580028533936,
-      "learning_rate": 5.442400864406504e-05,
-      "loss": 0.006,
+      "grad_norm": 1.45970618724823,
+      "learning_rate": 1.3709254203049984e-08,
+      "loss": 0.6691,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.5189571365444101,
-      "eval_e3_equal": 0.7959346000883782,
+      "eval_all_ligands_equal": 0.0,
+      "eval_e3_equal": 0.043747238179407864,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9897481219619974,
-      "eval_e3_heavy_atoms_difference": 0.609191338930623,
-      "eval_e3_heavy_atoms_difference_norm": 0.013221863982469433,
+      "eval_e3_has_attachment_point(s)": 0.9451171011931065,
+      "eval_e3_heavy_atoms_difference": 1.780733539549271,
+      "eval_e3_heavy_atoms_difference_norm": 0.011796785596124824,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9897481219619974,
-      "eval_has_all_attachment_points": 0.987273530711445,
-      "eval_has_three_substructures": 0.9997348652231551,
-      "eval_heavy_atoms_difference": 3.68519664162616,
-      "eval_heavy_atoms_difference_norm": 0.047260073323281075,
-      "eval_linker_equal": 0.7496243923994698,
+      "eval_e3_valid": 0.9451171011931065,
+      "eval_has_all_attachment_points": 0.653557224922669,
+      "eval_has_three_substructures": 0.9476800707026072,
+      "eval_heavy_atoms_difference": 66.00486080424216,
+      "eval_heavy_atoms_difference_norm": 0.8581384495257799,
+      "eval_linker_equal": 0.001237295625276182,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9977021652673442,
-      "eval_linker_heavy_atoms_difference": 0.32470172337604947,
-      "eval_linker_heavy_atoms_difference_norm": -0.001965493580026269,
+      "eval_linker_has_attachment_point(s)": 0.907998232434821,
+      "eval_linker_heavy_atoms_difference": 9.123287671232877,
+      "eval_linker_heavy_atoms_difference_norm": 0.4624936678575905,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9977021652673442,
-      "eval_loss": 0.3125883936882019,
-      "eval_num_fragments": 3.0000883782589485,
-      "eval_poi_equal": 0.7557224922669024,
+      "eval_linker_valid": 0.907998232434821,
+      "eval_loss": 0.729906439781189,
+      "eval_num_fragments": 3.038886433937251,
+      "eval_poi_equal": 8.837825894829872e-05,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9706584180291649,
-      "eval_poi_heavy_atoms_difference": 0.7122403888643394,
-      "eval_poi_heavy_atoms_difference_norm": 0.010501165767226064,
+      "eval_poi_has_attachment_point(s)": 0.26619531595227575,
+      "eval_poi_heavy_atoms_difference": 26.38851082633672,
+      "eval_poi_heavy_atoms_difference_norm": 0.8104800494182971,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9706584180291649,
-      "eval_reassembly": 0.5268228015908086,
-      "eval_reassembly_nostereo": 0.552364118426867,
-      "eval_runtime": 1535.8187,
-      "eval_samples_per_second": 7.367,
-      "eval_steps_per_second": 0.115,
+      "eval_poi_valid": 0.26619531595227575,
+      "eval_reassembly": 0.0,
+      "eval_reassembly_nostereo": 0.0,
+      "eval_runtime": 1464.416,
+      "eval_samples_per_second": 7.727,
+      "eval_steps_per_second": 0.121,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9590808661069377,
+      "eval_valid": 0.24984533804684048,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.07069461047649384,
-      "learning_rate": 3.037037763348429e-05,
-      "loss": 0.004,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.05461514741182327,
-      "learning_rate": 1.1114976351929146e-05,
-      "loss": 0.003,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.06773870438337326,
-      "learning_rate": 9.388408882665784e-07,
-      "loss": 0.0026,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.17225702106952667,
-      "learning_rate": 0.00010179964060387911,
-      "loss": 0.0115,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.0747770220041275,
-      "learning_rate": 8.955148497015824e-05,
-      "loss": 0.0072,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.5173663278833407,
-      "eval_e3_equal": 0.8022094564737074,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9894829871851525,
-      "eval_e3_heavy_atoms_difference": 0.6038886433937252,
-      "eval_e3_heavy_atoms_difference_norm": 0.012404143305947432,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9894829871851525,
-      "eval_has_all_attachment_points": 0.9878921785240831,
-      "eval_has_three_substructures": 0.9992045956694653,
-      "eval_heavy_atoms_difference": 4.957578435704817,
-      "eval_heavy_atoms_difference_norm": 0.06262525333255187,
-      "eval_linker_equal": 0.7567830313742819,
-      "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": 0.40980998674326113,
-      "eval_linker_heavy_atoms_difference_norm": 0.006682600227663089,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.996464869642068,
-      "eval_loss": 0.30381327867507935,
-      "eval_num_fragments": 2.999381352187362,
-      "eval_poi_equal": 0.7467962881131242,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9570481661511269,
-      "eval_poi_heavy_atoms_difference": 1.3366327883340698,
-      "eval_poi_heavy_atoms_difference_norm": 0.031869690411330115,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9570481661511269,
-      "eval_reassembly": 0.5249668581528943,
-      "eval_reassembly_nostereo": 0.5500662836942112,
-      "eval_runtime": 1535.106,
-      "eval_samples_per_second": 7.371,
-      "eval_steps_per_second": 0.115,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9453822359699514,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.06744291633367538,
-      "learning_rate": 6.894387657509068e-05,
-      "loss": 0.004,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.0616152361035347,
-      "learning_rate": 4.455848650174181e-05,
-      "loss": 0.0028,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.030910490080714226,
-      "learning_rate": 2.1816896643868615e-05,
-      "loss": 0.002,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.05898826941847801,
-      "learning_rate": 5.775224268319631e-06,
-      "loss": 0.0015,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.060594432055950165,
-      "learning_rate": 0.0,
-      "loss": 0.0013,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.5703049049933716,
-      "eval_e3_equal": 0.8189129474149359,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9942554131683606,
-      "eval_e3_heavy_atoms_difference": 0.36783031374281927,
-      "eval_e3_heavy_atoms_difference_norm": 0.006249267556050742,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9942554131683606,
-      "eval_has_all_attachment_points": 0.9908086610693769,
-      "eval_has_three_substructures": 0.9999116217410517,
-      "eval_heavy_atoms_difference": 2.9059655324790103,
-      "eval_heavy_atoms_difference_norm": 0.03736833824404718,
-      "eval_linker_equal": 0.8147591692443659,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9983208130799823,
-      "eval_linker_heavy_atoms_difference": 0.28608042421564295,
-      "eval_linker_heavy_atoms_difference_norm": 0.00016462975909777643,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9983208130799823,
-      "eval_loss": 0.31623375415802,
-      "eval_num_fragments": 2.9999116217410515,
-      "eval_poi_equal": 0.7785240830755634,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9743703049049933,
-      "eval_poi_heavy_atoms_difference": 0.5845338046840477,
-      "eval_poi_heavy_atoms_difference_norm": 0.008440998215488436,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9743703049049933,
-      "eval_reassembly": 0.5772867874502873,
-      "eval_reassembly_nostereo": 0.6049491825011047,
-      "eval_runtime": 1518.7606,
-      "eval_samples_per_second": 7.45,
-      "eval_steps_per_second": 0.117,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9672116659301812,
-      "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=11-learning_rate=1.0e-04-warmup_ratio=0.080-num_cycles=3",
+  "trial_name": "trial-number=12-learning_rate=2.2e-06-warmup_ratio=0.050-num_cycles=10",
   "trial_params": {
-    "learning_rate": 0.00010390406291286528,
-    "num_cycles": 3,
-    "warmup_ratio": 0.08
+    "learning_rate": 2.182809124107808e-06,
+    "num_cycles": 10,
+    "warmup_ratio": 0.05
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:79e8dd2f501b00cb907081399791da6f3e67acafeecf3f0d96b43e4e21e6e9c2
+oid sha256:e2d81e53661da53b9a9ee62f9627e098efc364b9e2a928206d40d7422e054944
 size 7480