Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3c856a373f7e52df41421237d8420117e67343df19ca689b10a2a31630045ceb
+oid sha256:6a5bd90cb049a5637534373af262a227fef702a16ff9fa0cf6096bd0c0e7f3fa
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:355ac0ebd329fda4def09a6d94df63be9f5d87108dde115142d8ac4d1c1a5227
+oid sha256:95ef8f20141b8c199b433664d10397fdf9bb4bd9423a72c6831f3a7fc3797eca
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dbbcba53f5c8a819a3c9588686ef0e840abc7d4bdfb838d9c20f4814c24e989f
+oid sha256:4f22a7bd0895701c7ec7e10e370282c2f7a85d0bc870c671f538ccc1a7cc4038
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b7b070ada8df300c245e6daa26454039d86686f1bc260ce873c14c0e22ed613
+oid sha256:efc7b8261dc9f0cd1aff039568c605f93169128d3740989947948ba3d1781e01
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.448519664162616,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=17-learning_rate=2.3e-05-warmup_ratio=0.070-num_cycles=5.500/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.5021652673442333,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=18-learning_rate=2.0e-04-warmup_ratio=0.010-num_cycles=2.500/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": 1.2309077978134155,
-      "learning_rate": 1.66041292757749e-05,
-      "loss": 3.0087,
+      "grad_norm": 0.6551521420478821,
+      "learning_rate": 0.00014340124668857647,
+      "loss": 1.5229,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.763126790523529,
-      "learning_rate": 1.6680127157646984e-05,
-      "loss": 0.5683,
+      "grad_norm": 0.33902546763420105,
+      "learning_rate": 0.0001878001438351963,
+      "loss": 0.0906,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.7100871801376343,
-      "learning_rate": 1.586004103408713e-07,
-      "loss": 0.2939,
+      "grad_norm": 0.1665482521057129,
+      "learning_rate": 0.00012191951719773646,
+      "loss": 0.024,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.6781579852104187,
-      "learning_rate": 1.3012154594856692e-05,
-      "loss": 0.2607,
+      "grad_norm": 0.0971766784787178,
+      "learning_rate": 4.1475601830154605e-05,
+      "loss": 0.0107,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.6720414757728577,
-      "learning_rate": 2.2258211750211904e-05,
-      "loss": 0.1925,
+      "grad_norm": 0.1026405617594719,
+      "learning_rate": 5.005669712982349e-07,
+      "loss": 0.0063,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.40102341771125793,
-      "learning_rate": 4.171131747579263e-06,
-      "loss": 0.1259,
+      "grad_norm": 0.13334587216377258,
+      "learning_rate": 2.6516320374474144e-05,
+      "loss": 0.0059,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.33403462171554565,
-      "learning_rate": 5.162865531391757e-06,
-      "loss": 0.1072,
+      "grad_norm": 0.15596552193164825,
+      "learning_rate": 0.00010204873010150722,
+      "loss": 0.008,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.44077345728874207,
-      "learning_rate": 2.2688596596301756e-05,
-      "loss": 0.0983,
+      "grad_norm": 0.13316264748573303,
+      "learning_rate": 0.00017636456261024047,
+      "loss": 0.0177,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.39217662811279297,
-      "learning_rate": 1.1776811003545574e-05,
-      "loss": 0.0716,
+      "grad_norm": 0.20416474342346191,
+      "learning_rate": 0.00019954772889767774,
+      "loss": 0.0124,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.33058542013168335,
-      "learning_rate": 4.274544366418573e-07,
-      "loss": 0.0557,
+      "grad_norm": 0.10144529491662979,
+      "learning_rate": 0.00015602667506056428,
+      "loss": 0.0083,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.1456473707467963,
-      "eval_e3_equal": 0.6572691117984976,
+      "eval_all_ligands_equal": 0.5021652673442333,
+      "eval_e3_equal": 0.7853292090145824,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9968183826778613,
-      "eval_e3_heavy_atoms_difference": -0.38382677861246134,
-      "eval_e3_heavy_atoms_difference_norm": -0.029623813342479082,
+      "eval_e3_has_attachment_point(s)": 0.9720724701723376,
+      "eval_e3_heavy_atoms_difference": 1.070967741935484,
+      "eval_e3_heavy_atoms_difference_norm": 0.027758721897316797,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9968183826778613,
-      "eval_has_all_attachment_points": 0.9814405656208572,
-      "eval_has_three_substructures": 0.9990278391515687,
-      "eval_heavy_atoms_difference": 8.372691117984976,
-      "eval_heavy_atoms_difference_norm": 0.11094699556330354,
-      "eval_linker_equal": 0.2407423773751657,
+      "eval_e3_valid": 0.9720724701723376,
+      "eval_has_all_attachment_points": 0.9681838267786125,
+      "eval_has_three_substructures": 0.9967300044189129,
+      "eval_heavy_atoms_difference": 10.814228899690676,
+      "eval_heavy_atoms_difference_norm": 0.13911716976370936,
+      "eval_linker_equal": 0.7400795404330535,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.99584622182943,
-      "eval_linker_heavy_atoms_difference": 1.0983650022094564,
-      "eval_linker_heavy_atoms_difference_norm": 0.01812954549559559,
+      "eval_linker_has_attachment_point(s)": 0.9931064958020327,
+      "eval_linker_heavy_atoms_difference": 0.5992929739284136,
+      "eval_linker_heavy_atoms_difference_norm": 0.010982995551670276,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.99584622182943,
-      "eval_loss": 0.2961718440055847,
-      "eval_num_fragments": 2.999381352187362,
-      "eval_poi_equal": 0.623950508174989,
+      "eval_linker_valid": 0.9931064958020327,
+      "eval_loss": 0.28404873609542847,
+      "eval_num_fragments": 2.9961113566062747,
+      "eval_poi_equal": 0.7383119752540875,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9027839151568714,
-      "eval_poi_heavy_atoms_difference": 2.9172779496243924,
-      "eval_poi_heavy_atoms_difference_norm": 0.09385926668544316,
+      "eval_poi_has_attachment_point(s)": 0.9011047282368537,
+      "eval_poi_heavy_atoms_difference": 3.4031816173221388,
+      "eval_poi_heavy_atoms_difference_norm": 0.10832716160390982,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9027839151568714,
-      "eval_reassembly": 0.14794520547945206,
-      "eval_reassembly_nostereo": 0.1665046398585948,
-      "eval_runtime": 1567.3459,
-      "eval_samples_per_second": 7.219,
-      "eval_steps_per_second": 0.113,
+      "eval_poi_valid": 0.9011047282368537,
+      "eval_reassembly": 0.5109147149801149,
+      "eval_reassembly_nostereo": 0.5356606274856386,
+      "eval_runtime": 1588.2721,
+      "eval_samples_per_second": 7.124,
+      "eval_steps_per_second": 0.111,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8985417587273531,
+      "eval_valid": 0.8764471939902784,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.4289386570453644,
-      "learning_rate": 1.776329188349048e-05,
-      "loss": 0.0531,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.2785295844078064,
-      "learning_rate": 1.9300081717101823e-05,
-      "loss": 0.046,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.22967936098575592,
-      "learning_rate": 1.0943784305675876e-06,
-      "loss": 0.0343,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.2513010501861572,
-      "learning_rate": 9.862523618028005e-06,
-      "loss": 0.0309,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.29722490906715393,
-      "learning_rate": 2.3105212568338626e-05,
-      "loss": 0.0302,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.3163057887759611,
-      "eval_e3_equal": 0.7402562969509501,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9941670349094123,
-      "eval_e3_heavy_atoms_difference": 0.19036676977463543,
-      "eval_e3_heavy_atoms_difference_norm": -0.002474182119390328,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9941670349094123,
-      "eval_has_all_attachment_points": 0.9791427308882015,
-      "eval_has_three_substructures": 0.9985859478568272,
-      "eval_heavy_atoms_difference": 8.547768448961556,
-      "eval_heavy_atoms_difference_norm": 0.11291736666343977,
-      "eval_linker_equal": 0.4749447635881573,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9948740609809986,
-      "eval_linker_heavy_atoms_difference": 0.5648254529385771,
-      "eval_linker_heavy_atoms_difference_norm": -0.003980535978264268,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9948740609809986,
-      "eval_loss": 0.2942982017993927,
-      "eval_num_fragments": 2.9990278391515686,
-      "eval_poi_equal": 0.6655766681396377,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9026071586389748,
-      "eval_poi_heavy_atoms_difference": 3.056738842244808,
-      "eval_poi_heavy_atoms_difference_norm": 0.09512481066347839,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9026071586389748,
-      "eval_reassembly": 0.32081307998232433,
-      "eval_reassembly_nostereo": 0.3444100751215201,
-      "eval_runtime": 1578.1937,
-      "eval_samples_per_second": 7.17,
-      "eval_steps_per_second": 0.112,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8960671674768007,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.20609797537326813,
-      "learning_rate": 6.84133667553644e-06,
-      "loss": 0.0245,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.23419219255447388,
-      "learning_rate": 2.8044543687991375e-06,
-      "loss": 0.0197,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.28124216198921204,
-      "learning_rate": 2.135331805932135e-05,
-      "loss": 0.02,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.21178261935710907,
-      "learning_rate": 1.4891027000593144e-05,
-      "loss": 0.0194,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.23398344218730927,
-      "learning_rate": 0.0,
-      "loss": 0.0141,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.448519664162616,
-      "eval_e3_equal": 0.7790543526292532,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9925762262483429,
-      "eval_e3_heavy_atoms_difference": 0.1948740609809987,
-      "eval_e3_heavy_atoms_difference_norm": -0.0020766324563693556,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9925762262483429,
-      "eval_has_all_attachment_points": 0.9817057003977022,
-      "eval_has_three_substructures": 0.9988510826336722,
-      "eval_heavy_atoms_difference": 7.299425541316836,
-      "eval_heavy_atoms_difference_norm": 0.09865363120279987,
-      "eval_linker_equal": 0.6330534688466637,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9959346000883783,
-      "eval_linker_heavy_atoms_difference": 0.44418912947414935,
-      "eval_linker_heavy_atoms_difference_norm": 0.001766423395292503,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9959346000883783,
-      "eval_loss": 0.2937150001525879,
-      "eval_num_fragments": 3.0000883782589485,
-      "eval_poi_equal": 0.7224038886433938,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9182501104728237,
-      "eval_poi_heavy_atoms_difference": 2.356164383561644,
-      "eval_poi_heavy_atoms_difference_norm": 0.07590865194375644,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9182501104728237,
-      "eval_reassembly": 0.4549712770658418,
-      "eval_reassembly_nostereo": 0.47724259832081306,
-      "eval_runtime": 1505.5772,
-      "eval_samples_per_second": 7.515,
-      "eval_steps_per_second": 0.118,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9097657976137871,
-      "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=17-learning_rate=2.3e-05-warmup_ratio=0.070-num_cycles=5.500",
+  "trial_name": "trial-number=18-learning_rate=2.0e-04-warmup_ratio=0.010-num_cycles=2.500",
   "trial_params": {
-    "learning_rate": 2.321257272753331e-05,
-    "num_cycles": 5.5,
-    "warmup_ratio": 0.06999999999999999
+    "learning_rate": 0.0002004749428706299,
+    "num_cycles": 2.5,
+    "warmup_ratio": 0.01
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f210b06f27913c2673bfd7d2d682546273265dd19b3943552ec0e3bed293fbd
+oid sha256:eb77addbddf6855fb55635ab344a8dba8622decc4d1782413855cf464e45aa31
 size 7416