Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ca080cb5810c4f1cb337e9a9ffc0b33793c0454bee48575739f405720f00b69
+oid sha256:f2f89d1cbe59dcb8a26470d236c79015c4fdad3f279aea064d414df70f55c710
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b68ddf0126470330e730e5f70d28d6bc73b78a80406821de9c617810992e4cc
+oid sha256:7ef919e935701cb810544ea4cf3e62aa266e513dee82d825a19bc15e3b1ee773
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f05568c9474927fa3f00b14190be9bec9b14406698bef27ced0d260778a147e1
+oid sha256:72baddb0147b97af8f16407b09e69da9a38c23b42d021be1313508923c4cec32
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:34d3c16f5dc10554a9e6d40225c10e4c32c37143fd15f582a82747dbcc32005b
+oid sha256:74048130960f2f12910df5eefabdc42e977282738c39c3f79fd993ee3cba9210
 size 1192

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_reduce-opt25-rand-smiles/trial-number=3-learning_rate=3.1e-04-warmup_ratio=0.080-min_lr=0.000-factor=0.350/checkpoint-5000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 8.837825894829872e-05,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_reduce-opt25-rand-smiles/trial-number=4-learning_rate=4.2e-06-warmup_ratio=0.030-min_lr=0.000-factor=0.800/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": 0.7187104225158691,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5961,
+      "grad_norm": 1.2561712265014648,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 2.4831,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.8111324310302734,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5685,
+      "grad_norm": 1.2166751623153687,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 1.0063,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.7726447582244873,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5707,
+      "grad_norm": 1.246788501739502,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.7292,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.9721251130104065,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5688,
+      "grad_norm": 1.0463918447494507,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.5705,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 1.0270103216171265,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5661,
+      "grad_norm": 1.0265284776687622,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.4664,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.9803557991981506,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5645,
+      "grad_norm": 1.0784770250320435,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.3926,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.9630672931671143,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5611,
+      "grad_norm": 0.7869899272918701,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.3353,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.9206050038337708,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.6062,
+      "grad_norm": 1.0476751327514648,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.2925,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 1.0815049409866333,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5614,
+      "grad_norm": 0.7107057571411133,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.2574,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 1.0433939695358276,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5604,
+      "grad_norm": 0.6933206915855408,
+      "learning_rate": 4.170090767711237e-06,
+      "loss": 0.2316,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.0,
+      "eval_all_ligands_equal": 8.837825894829872e-05,
+      "eval_e3_equal": 0.19681838267786125,
       "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)": 0.9943437914273089,
+      "eval_e3_heavy_atoms_difference": -1.1870967741935483,
+      "eval_e3_heavy_atoms_difference_norm": -0.068259312401578,
       "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": 0.9943437914273089,
+      "eval_has_all_attachment_points": 0.940521431727795,
+      "eval_has_three_substructures": 0.9947856827220504,
+      "eval_heavy_atoms_difference": 24.551922227132124,
+      "eval_heavy_atoms_difference_norm": 0.31476917674862775,
+      "eval_linker_equal": 0.005125939019001326,
       "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)": 0.9897481219619974,
+      "eval_linker_heavy_atoms_difference": 5.2481661511268225,
+      "eval_linker_heavy_atoms_difference_norm": 0.21898110136881632,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.0,
-      "eval_loss": 3.5865700244903564,
-      "eval_num_fragments": 1.0033583738400353,
-      "eval_poi_equal": 0.0,
+      "eval_linker_valid": 0.9897481219619974,
+      "eval_loss": 0.386488139629364,
+      "eval_num_fragments": 2.99531595227574,
+      "eval_poi_equal": 0.1725143614670791,
       "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)": 0.748210340256297,
+      "eval_poi_heavy_atoms_difference": 9.451701281484755,
+      "eval_poi_heavy_atoms_difference_norm": 0.26897859855195805,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.0,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 5682.8651,
-      "eval_samples_per_second": 1.991,
-      "eval_steps_per_second": 0.031,
+      "eval_poi_valid": 0.748210340256297,
+      "eval_reassembly": 0.00017675651789659743,
+      "eval_reassembly_nostereo": 0.00035351303579319487,
+      "eval_runtime": 1509.5004,
+      "eval_samples_per_second": 7.496,
+      "eval_steps_per_second": 0.117,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
+      "eval_valid": 0.7445868316394167,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 1.1409233808517456,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.559,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 1.2372825145721436,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5594,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 1.2818654775619507,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5575,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 1.2530715465545654,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.557,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 1.2819572687149048,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5562,
-      "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": 3.578969955444336,
-      "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": 5664.1464,
-      "eval_samples_per_second": 1.998,
-      "eval_steps_per_second": 0.031,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 1.2820355892181396,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5566,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 1.3644280433654785,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5561,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 1.2222177982330322,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5558,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 1.3301723003387451,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5551,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 1.2963240146636963,
-      "learning_rate": 0.0003143431054003243,
-      "loss": 3.5558,
-      "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": 3.576150894165039,
-      "eval_num_fragments": 1.0025629695095006,
-      "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": 5655.7487,
-      "eval_samples_per_second": 2.001,
-      "eval_steps_per_second": 0.031,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.0,
-      "step": 10000
     }
   ],
   "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": 8.729673191023411e+16,
+  "total_flos": 4.364959997072179e+16,
   "train_batch_size": 128,
-  "trial_name": "trial-number=3-learning_rate=3.1e-04-warmup_ratio=0.080-min_lr=0.000-factor=0.350",
+  "trial_name": "trial-number=4-learning_rate=4.2e-06-warmup_ratio=0.030-min_lr=0.000-factor=0.800",
   "trial_params": {
-    "factor": 0.35,
-    "learning_rate": 0.0003143431054003243,
-    "min_lr": 7.783062873657726e-11,
-    "warmup_ratio": 0.08
+    "factor": 0.8,
+    "learning_rate": 4.170090767711237e-06,
+    "min_lr": 6.0724410942321634e-12,
+    "warmup_ratio": 0.03
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3fb53e5e8cd5ad3dc5f594c48bdfd22a6fa924f20a1441dc26739379ff2bf120
+oid sha256:bb3a31938db14c7e84935e0b50cb36c1f4b49136a4342b4b9c8e65d937342068
 size 7544