Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:234a8b06c42bde1a16b91986a0335e1b78fd275814369d9824f7ebaabc8dddc1
+oid sha256:fd74243970fd33c036b75de8b23581a75b6c1135a8123b00df009cbf2780a1f0
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:56f1043273a1a9627c84c2cd50f9a793dc0dc10e66d508f97f1ebbaba479cc68
+oid sha256:a8c091ad07b517cab64ddda2ae4ad75e998c210a02c5f6d4c662a63ba4c24284
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:00fb925f04b19eca323662b6c4791ee1ebc5cdca8cedad3c713371a868bf334f
+oid sha256:31ce09e4561ec38c52fc773f8f49a8d4fbf72d72199f6c9a6b5bba444e9dd3ec
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:27b31a871d1ea80cfdcae77c01bb81a288da556fb8a8ae455581eab8347024a3
+oid sha256:6111ba44dda4526fc2296485b4467d1f9d05727481377d4654e5bc82a72dd135
 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_restarts-opt25/trial-number=23-learning_rate=5.9e-04-warmup_ratio=0.070-num_cycles=5/checkpoint-5000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.006098099867432612,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts-opt25/trial-number=24-learning_rate=1.2e-05-warmup_ratio=0.020-num_cycles=9/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": 0.26378336548805237,
-      "learning_rate": 0.0005319819238555916,
-      "loss": 0.8781,
+      "grad_norm": 1.1682956218719482,
+      "learning_rate": 8.609489056175177e-06,
+      "loss": 2.1752,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.26860642433166504,
-      "learning_rate": 0.0003369078003427231,
-      "loss": 0.1831,
+      "grad_norm": 1.1533381938934326,
+      "learning_rate": 9.656008727064383e-07,
+      "loss": 0.672,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.3114512264728546,
-      "learning_rate": 0.00011653163666206531,
-      "loss": 0.1642,
+      "grad_norm": 0.9731566905975342,
+      "learning_rate": 1.0065956095929134e-05,
+      "loss": 0.5459,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.4044862985610962,
-      "learning_rate": 2.429070832855221e-06,
-      "loss": 0.1558,
+      "grad_norm": 0.7106562256813049,
+      "learning_rate": 2.0993649965972538e-06,
+      "loss": 0.3736,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.5458288788795471,
-      "learning_rate": 0.0005271959652135703,
-      "loss": 2.7241,
+      "grad_norm": 0.980603039264679,
+      "learning_rate": 1.1199720219819944e-05,
+      "loss": 0.3217,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.7785220146179199,
-      "learning_rate": 0.0003290315839343549,
-      "loss": 3.5665,
+      "grad_norm": 0.6568108201026917,
+      "learning_rate": 3.555832036351209e-06,
+      "loss": 0.2471,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 1.143982172012329,
-      "learning_rate": 0.00011026765926286528,
-      "loss": 3.5575,
+      "grad_norm": 0.583015501499176,
+      "learning_rate": 1.1918930601284844e-05,
+      "loss": 0.2145,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 1.1621668338775635,
-      "learning_rate": 1.5172415558257588e-06,
-      "loss": 3.5531,
+      "grad_norm": 0.5566324591636658,
+      "learning_rate": 5.2170076943504535e-06,
+      "loss": 0.1804,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.5300591588020325,
-      "learning_rate": 0.0005213489853416686,
-      "loss": 3.5607,
+      "grad_norm": 0.5157020092010498,
+      "learning_rate": 1.2165321092526382e-05,
+      "loss": 0.155,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.8344257473945618,
-      "learning_rate": 0.000319733743673733,
-      "loss": 3.5623,
+      "grad_norm": 0.516018271446228,
+      "learning_rate": 6.948313398175921e-06,
+      "loss": 0.1403,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.0,
+      "eval_all_ligands_equal": 0.006098099867432612,
+      "eval_e3_equal": 0.43163941670349093,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.0,
-      "eval_e3_heavy_atoms_difference": 28.086433937251435,
-      "eval_e3_heavy_atoms_difference_norm": 1.0,
+      "eval_e3_has_attachment_point(s)": 0.9950508174988952,
+      "eval_e3_heavy_atoms_difference": -0.6882015024304021,
+      "eval_e3_heavy_atoms_difference_norm": -0.043577084852665905,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.0,
-      "eval_has_all_attachment_points": 0.0,
-      "eval_has_three_substructures": 0.0,
-      "eval_heavy_atoms_difference": 75.67370746796288,
-      "eval_heavy_atoms_difference_norm": 1.0,
-      "eval_linker_equal": 0.0,
+      "eval_e3_valid": 0.9950508174988952,
+      "eval_has_all_attachment_points": 0.9637649138311976,
+      "eval_has_three_substructures": 0.996464869642068,
+      "eval_heavy_atoms_difference": 16.120282810428634,
+      "eval_heavy_atoms_difference_norm": 0.20922444422385186,
+      "eval_linker_equal": 0.023862129916040652,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.0,
-      "eval_linker_heavy_atoms_difference": 16.430402121078213,
-      "eval_linker_heavy_atoms_difference_norm": 1.0,
+      "eval_linker_has_attachment_point(s)": 0.9878921785240831,
+      "eval_linker_heavy_atoms_difference": 3.8430402121078213,
+      "eval_linker_heavy_atoms_difference_norm": 0.15478254398680125,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.0,
-      "eval_loss": 3.5808053016662598,
-      "eval_num_fragments": 1.0018559434379142,
-      "eval_poi_equal": 0.0,
+      "eval_linker_valid": 0.9878921785240831,
+      "eval_loss": 0.33513495326042175,
+      "eval_num_fragments": 2.9966416261599647,
+      "eval_poi_equal": 0.41555457357490055,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.0,
-      "eval_poi_heavy_atoms_difference": 31.15687140963323,
-      "eval_poi_heavy_atoms_difference_norm": 1.0,
+      "eval_poi_has_attachment_point(s)": 0.8372072470172338,
+      "eval_poi_heavy_atoms_difference": 5.426336721166593,
+      "eval_poi_heavy_atoms_difference_norm": 0.16311688485735315,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.0,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 5643.0408,
-      "eval_samples_per_second": 2.005,
-      "eval_steps_per_second": 0.031,
+      "eval_poi_valid": 0.8372072470172338,
+      "eval_reassembly": 0.0061864781263809105,
+      "eval_reassembly_nostereo": 0.00821917808219178,
+      "eval_runtime": 1598.5131,
+      "eval_samples_per_second": 7.078,
+      "eval_steps_per_second": 0.111,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
+      "eval_valid": 0.830490499337163,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 1.0807377099990845,
-      "learning_rate": 0.00010307023622930463,
-      "loss": 3.5553,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 1.1222482919692993,
-      "learning_rate": 7.174587879642109e-07,
-      "loss": 3.5524,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.7741912603378296,
-      "learning_rate": 0.0005152736173626581,
-      "loss": 3.5622,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.9775688648223877,
-      "learning_rate": 0.00031041052476416946,
-      "loss": 3.5578,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 1.0934219360351562,
-      "learning_rate": 9.606559633302014e-05,
-      "loss": 3.5532,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.0,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.0,
-      "eval_e3_heavy_atoms_difference": 28.086433937251435,
-      "eval_e3_heavy_atoms_difference_norm": 1.0,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.0,
-      "eval_has_all_attachment_points": 0.0,
-      "eval_has_three_substructures": 0.0,
-      "eval_heavy_atoms_difference": 75.67370746796288,
-      "eval_heavy_atoms_difference_norm": 1.0,
-      "eval_linker_equal": 0.0,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.0,
-      "eval_linker_heavy_atoms_difference": 16.430402121078213,
-      "eval_linker_heavy_atoms_difference_norm": 1.0,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.0,
-      "eval_loss": 3.573688507080078,
-      "eval_num_fragments": 1.0022978347326557,
-      "eval_poi_equal": 0.0,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.0,
-      "eval_poi_heavy_atoms_difference": 31.15687140963323,
-      "eval_poi_heavy_atoms_difference_norm": 1.0,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.0,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 5623.3046,
-      "eval_samples_per_second": 2.012,
-      "eval_steps_per_second": 0.031,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 1.0831973552703857,
-      "learning_rate": 2.1351972818891565e-07,
-      "loss": 3.5517,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.8361011743545532,
-      "learning_rate": 0.0005092960425219065,
-      "loss": 3.56,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.929772138595581,
-      "learning_rate": 0.0003015385100469968,
-      "loss": 3.5569,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 1.103515863418579,
-      "learning_rate": 8.95961801188567e-05,
-      "loss": 3.5529,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 1.0811818838119507,
-      "learning_rate": 9.268420650080632e-09,
-      "loss": 3.5518,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.0,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.0,
-      "eval_e3_heavy_atoms_difference": 28.086433937251435,
-      "eval_e3_heavy_atoms_difference_norm": 1.0,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.0,
-      "eval_has_all_attachment_points": 0.0,
-      "eval_has_three_substructures": 0.0,
-      "eval_heavy_atoms_difference": 75.67370746796288,
-      "eval_heavy_atoms_difference_norm": 1.0,
-      "eval_linker_equal": 0.0,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.0,
-      "eval_linker_heavy_atoms_difference": 16.430402121078213,
-      "eval_linker_heavy_atoms_difference_norm": 1.0,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.0,
-      "eval_loss": 3.5727076530456543,
-      "eval_num_fragments": 1.001767565178966,
-      "eval_poi_equal": 0.0,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.0,
-      "eval_poi_heavy_atoms_difference": 31.15687140963323,
-      "eval_poi_heavy_atoms_difference_norm": 1.0,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.0,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 5607.7308,
-      "eval_samples_per_second": 2.018,
-      "eval_steps_per_second": 0.032,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
-      "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=23-learning_rate=5.9e-04-warmup_ratio=0.070-num_cycles=5",
+  "trial_name": "trial-number=24-learning_rate=1.2e-05-warmup_ratio=0.020-num_cycles=9",
   "trial_params": {
-    "learning_rate": 0.0005890587774708618,
-    "num_cycles": 5,
-    "warmup_ratio": 0.06999999999999999
+    "learning_rate": 1.2165321092526382e-05,
+    "num_cycles": 9,
+    "warmup_ratio": 0.02
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ffc23a2ea6db09dedd83ecaab57dfa5dc9d9036df116c7da8ed463ecb098ff40
+oid sha256:18d8632640bbe126885af10bd60f2424fb195aa4218c094577172b919da3113f
 size 7480