Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e03ef9e412f92413e6a7dd5df46e7fb1ef8fa8edb6784ada1b83b5230dd58fc7
+oid sha256:bd54f18ba95abb3e83b5c69be48574a67f5dc638c380c159eedae3b0850b1908
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0a6e26e4183f0a1c6a19ed4badbca47f4b0afabe21f1832e958e8d92eede3ea5
+oid sha256:07eaba784767c039b87a56c8bbc5e441e0f1527176e173143bbd9f414b63b54d
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28316fc68eee2ab3cc8969d4fe7b6a25e6e236fabe423ef412772900166c1026
+oid sha256:ad3f63352a251f31d1ec97c81494799c7e42ae3c2a487cb00350f0dafde0e065
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f33d61e7faf50026c6df62b1820a4e5e991131d1b2a9f07f2595d907464c4ed2
+oid sha256:d6f447d613c8f23872e44f681809793d7bb3f543c2932bfa3bd7ce8811321e46
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.00017675651789659743,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=12-learning_rate=2.2e-06-warmup_ratio=0.050-num_cycles=10.000/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 8.837825894829872e-05,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=13-learning_rate=4.2e-06-warmup_ratio=0.100-num_cycles=6.500/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": 4.174542427062988,
-      "learning_rate": 1.5613799171014363e-06,
-      "loss": 6.0021,
+      "grad_norm": 2.3850045204162598,
+      "learning_rate": 2.989305882911821e-06,
+      "loss": 5.0844,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 1.8601213693618774,
-      "learning_rate": 6.043608340530767e-07,
-      "loss": 1.9802,
+      "grad_norm": 1.3937734365463257,
+      "learning_rate": 2.60466910705453e-06,
+      "loss": 1.378,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 1.5528665781021118,
-      "learning_rate": 1.7934007606188836e-06,
-      "loss": 1.4405,
+      "grad_norm": 1.3762049674987793,
+      "learning_rate": 1.4566522299079576e-07,
+      "loss": 1.0026,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 1.4923285245895386,
-      "learning_rate": 2.134062284657905e-07,
-      "loss": 1.2124,
+      "grad_norm": 1.3970109224319458,
+      "learning_rate": 3.848142336404206e-06,
+      "loss": 0.9337,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 1.4196209907531738,
-      "learning_rate": 2.0966891974916694e-06,
-      "loss": 1.1001,
+      "grad_norm": 1.310513973236084,
+      "learning_rate": 1.9763030136461577e-06,
+      "loss": 0.7344,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 1.3385555744171143,
-      "learning_rate": 1.4611952262523042e-08,
-      "loss": 0.9606,
+      "grad_norm": 1.279549479484558,
+      "learning_rate": 4.633445297631418e-07,
+      "loss": 0.6374,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 1.3339844942092896,
-      "learning_rate": 2.1799591909569794e-06,
-      "loss": 0.8973,
+      "grad_norm": 1.3507317304611206,
+      "learning_rate": 4.104914163179973e-06,
+      "loss": 0.6003,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 1.3039367198944092,
-      "learning_rate": 5.1486486192622186e-08,
-      "loss": 0.7968,
+      "grad_norm": 1.0677459239959717,
+      "learning_rate": 1.3582738363327274e-06,
+      "loss": 0.5024,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 1.4882336854934692,
-      "learning_rate": 2.024986117163073e-06,
-      "loss": 0.7516,
+      "grad_norm": 1.069346308708191,
+      "learning_rate": 9.294907338615516e-07,
+      "loss": 0.4581,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 1.4356786012649536,
-      "learning_rate": 3.159594019707261e-07,
-      "loss": 0.6797,
+      "grad_norm": 1.2829123735427856,
+      "learning_rate": 4.177685109431936e-06,
+      "loss": 0.4336,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.0,
-      "eval_e3_equal": 0.04021210782147592,
+      "eval_all_ligands_equal": 8.837825894829872e-05,
+      "eval_e3_equal": 0.11179849756959788,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9433495360141405,
-      "eval_e3_heavy_atoms_difference": 1.9438798055678304,
-      "eval_e3_heavy_atoms_difference_norm": 0.01739639034499783,
+      "eval_e3_has_attachment_point(s)": 0.983738400353513,
+      "eval_e3_heavy_atoms_difference": -0.5517454706142289,
+      "eval_e3_heavy_atoms_difference_norm": -0.052895150883789134,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9433495360141405,
-      "eval_has_all_attachment_points": 0.6479893946089262,
-      "eval_has_three_substructures": 0.9459125055236411,
-      "eval_heavy_atoms_difference": 65.59602297834732,
-      "eval_heavy_atoms_difference_norm": 0.8529760282186839,
-      "eval_linker_equal": 0.001237295625276182,
+      "eval_e3_valid": 0.983738400353513,
+      "eval_has_all_attachment_points": 0.848519664162616,
+      "eval_has_three_substructures": 0.986389748121962,
+      "eval_heavy_atoms_difference": 47.190366769774634,
+      "eval_heavy_atoms_difference_norm": 0.6028372354818079,
+      "eval_linker_equal": 0.002209456473707468,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9062306672558551,
-      "eval_linker_heavy_atoms_difference": 9.314980114891737,
-      "eval_linker_heavy_atoms_difference_norm": 0.47730779217018066,
+      "eval_linker_has_attachment_point(s)": 0.9752540874944764,
+      "eval_linker_heavy_atoms_difference": 7.183826778612461,
+      "eval_linker_heavy_atoms_difference_norm": 0.3251208261071983,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9062306672558551,
-      "eval_loss": 0.7388172745704651,
-      "eval_num_fragments": 3.039416703490941,
-      "eval_poi_equal": 0.00017675651789659743,
+      "eval_linker_valid": 0.9752540874944764,
+      "eval_loss": 0.5234163999557495,
+      "eval_num_fragments": 2.9943437914273088,
+      "eval_poi_equal": 0.02492266902342024,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.2838709677419355,
-      "eval_poi_heavy_atoms_difference": 26.241979673000444,
-      "eval_poi_heavy_atoms_difference_norm": 0.8095363094685946,
+      "eval_poi_has_attachment_point(s)": 0.48519664162615994,
+      "eval_poi_heavy_atoms_difference": 19.781528943879806,
+      "eval_poi_heavy_atoms_difference_norm": 0.5782018706950223,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.2838709677419355,
-      "eval_reassembly": 0.0,
-      "eval_reassembly_nostereo": 0.0,
-      "eval_runtime": 1458.6362,
-      "eval_samples_per_second": 7.757,
+      "eval_poi_valid": 0.48519664162615994,
+      "eval_reassembly": 8.837825894829872e-05,
+      "eval_reassembly_nostereo": 8.837825894829872e-05,
+      "eval_runtime": 1459.5769,
+      "eval_samples_per_second": 7.752,
       "eval_steps_per_second": 0.121,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.2685815289438798,
+      "eval_valid": 0.47971718957136544,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 1.1125599145889282,
-      "learning_rate": 1.6656876679399406e-06,
-      "loss": 0.6443,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 1.166053295135498,
-      "learning_rate": 7.501476710478152e-07,
-      "loss": 0.5931,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 1.236864447593689,
-      "learning_rate": 1.1807005624117962e-06,
-      "loss": 0.561,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 1.3960652351379395,
-      "learning_rate": 1.2590240506504755e-06,
-      "loss": 0.5254,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.9353835582733154,
-      "learning_rate": 6.761699700853001e-07,
-      "loss": 0.4944,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.00017675651789659743,
-      "eval_e3_equal": 0.09067609368095449,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9836500220945648,
-      "eval_e3_heavy_atoms_difference": -0.5509500662836943,
-      "eval_e3_heavy_atoms_difference_norm": -0.0561423282470494,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9836500220945648,
-      "eval_has_all_attachment_points": 0.8178524083075563,
-      "eval_has_three_substructures": 0.984887317719841,
-      "eval_heavy_atoms_difference": 55.044542642509946,
-      "eval_heavy_atoms_difference_norm": 0.7070090754939828,
-      "eval_linker_equal": 0.001855943437914273,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.968625718073354,
-      "eval_linker_heavy_atoms_difference": 7.189306230667256,
-      "eval_linker_heavy_atoms_difference_norm": 0.32506378644291267,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.968625718073354,
-      "eval_loss": 0.5763289928436279,
-      "eval_num_fragments": 3.001148917366328,
-      "eval_poi_equal": 0.011224038886433937,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.3774635439681838,
-      "eval_poi_heavy_atoms_difference": 23.0909412284578,
-      "eval_poi_heavy_atoms_difference_norm": 0.692810036859358,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.3774635439681838,
-      "eval_reassembly": 0.00017675651789659743,
-      "eval_reassembly_nostereo": 0.00017675651789659743,
-      "eval_runtime": 1415.0139,
-      "eval_samples_per_second": 7.996,
-      "eval_steps_per_second": 0.125,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.3716305788775961,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 1.3179833889007568,
-      "learning_rate": 1.7312149208305083e-06,
-      "loss": 0.4709,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 1.065822958946228,
-      "learning_rate": 2.625183839161817e-07,
-      "loss": 0.4407,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.9656862020492554,
-      "learning_rate": 2.0633757196738724e-06,
-      "loss": 0.4237,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 1.1340328454971313,
-      "learning_rate": 3.027835050239297e-08,
-      "loss": 0.3967,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.869488537311554,
-      "learning_rate": 2.182809124107808e-06,
-      "loss": 0.3832,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.00017675651789659743,
-      "eval_e3_equal": 0.1170128148475475,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.986389748121962,
-      "eval_e3_heavy_atoms_difference": -1.0851082633672116,
-      "eval_e3_heavy_atoms_difference_norm": -0.07018703541920102,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.986389748121962,
-      "eval_has_all_attachment_points": 0.8554131683605833,
-      "eval_has_three_substructures": 0.9875386654882898,
-      "eval_heavy_atoms_difference": 44.42925320371189,
-      "eval_heavy_atoms_difference_norm": 0.5690573662272553,
-      "eval_linker_equal": 0.0022978347326557667,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9771100309323907,
-      "eval_linker_heavy_atoms_difference": 7.132744144940345,
-      "eval_linker_heavy_atoms_difference_norm": 0.3255079732768209,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9771100309323907,
-      "eval_loss": 0.5013580322265625,
-      "eval_num_fragments": 2.9972602739726026,
-      "eval_poi_equal": 0.03216968625718073,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.5223155103844455,
-      "eval_poi_heavy_atoms_difference": 18.642951833848873,
-      "eval_poi_heavy_atoms_difference_norm": 0.5476717726083311,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.5223155103844455,
-      "eval_reassembly": 0.00017675651789659743,
-      "eval_reassembly_nostereo": 0.00017675651789659743,
-      "eval_runtime": 1399.4913,
-      "eval_samples_per_second": 8.085,
-      "eval_steps_per_second": 0.126,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.5179849756959788,
-      "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=12-learning_rate=2.2e-06-warmup_ratio=0.050-num_cycles=10.000",
+  "trial_name": "trial-number=13-learning_rate=4.2e-06-warmup_ratio=0.100-num_cycles=6.500",
   "trial_params": {
-    "learning_rate": 2.182809124107808e-06,
-    "num_cycles": 10.0,
-    "warmup_ratio": 0.05
+    "learning_rate": 4.179049624310726e-06,
+    "num_cycles": 6.5,
+    "warmup_ratio": 0.09999999999999999
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:661d73971ac0bb5024f9fbef00713b62ce5b99c0c8bf12acc425d08c2c9e10a5
+oid sha256:115135d446f1724bca1ef6352bff266776b73202f3f5c92bd526f2d2d263b045
 size 7416