Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f8c1bbec2d0a4fa69c953c9d0f576bb90eca02c37091685c3b647e4a1273ac6
+oid sha256:f8a7b286c7d9d672c21392e019f77110d92015e2a62fe16f7876edc2b0d7f019
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3b90890a22a1f787f2820c2d6436d6da8a03abdd188996419f85d51db3dd627b
+oid sha256:221015e74e31fecf4177f730824fce0d0bb7f28b78a7327a7f510902b8e2504b
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d4f2730e87c3fec905d7ed01d0ee7c24b2c47552e16531a8df91136a38ce4fb9
+oid sha256:26eae5b470dcb095f8eca549a0b6862d7e5df7ee0d304d7f6fd2a540ccd8e46d
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa637701c478cb8b80b3ade0a46f8ff99ff6e05c6233a12551054df34f86acd1
+oid sha256:466d7528300efc91fe3a7fe887d60e328b080520c102b054e4be51ce66b6e70d
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.4973928413610252,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25-rand-smiles/trial-number=10-learning_rate=4.4e-04-warmup_ratio=0.020-num_cycles=5.500/checkpoint-5000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.5305346884666372,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25-rand-smiles/trial-number=11-learning_rate=1.0e-04-warmup_ratio=0.080-num_cycles=3.000/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": 0.5057240724563599,
-      "learning_rate": 0.000312265868621247,
-      "loss": 1.1988,
+      "grad_norm": 1.0087827444076538,
+      "learning_rate": 7.432336402923124e-05,
+      "loss": 1.8581,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 0.19040384888648987,
-      "learning_rate": 0.0003136951242119503,
-      "loss": 0.0688,
+      "grad_norm": 0.4915059804916382,
+      "learning_rate": 9.453407081430295e-05,
+      "loss": 0.1818,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 0.10162537544965744,
-      "learning_rate": 2.98272159149201e-06,
-      "loss": 0.0167,
+      "grad_norm": 0.2095542550086975,
+      "learning_rate": 4.922460243282525e-05,
+      "loss": 0.0556,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 0.16298483312129974,
-      "learning_rate": 0.0002447133294201148,
-      "loss": 0.012,
+      "grad_norm": 0.19221505522727966,
+      "learning_rate": 6.484064135881745e-06,
+      "loss": 0.0311,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 0.16574081778526306,
-      "learning_rate": 0.0004185994767142762,
-      "loss": 0.042,
+      "grad_norm": 0.17544284462928772,
+      "learning_rate": 6.569215571421697e-06,
+      "loss": 0.0257,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 0.06716415286064148,
-      "learning_rate": 7.844446743689375e-05,
-      "loss": 0.016,
+      "grad_norm": 0.2089957296848297,
+      "learning_rate": 4.939985370478998e-05,
+      "loss": 0.0256,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 0.08064843714237213,
-      "learning_rate": 9.709552744129087e-05,
-      "loss": 0.0063,
+      "grad_norm": 0.20172426104545593,
+      "learning_rate": 9.463435501865708e-05,
+      "loss": 0.0263,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 0.2973688542842865,
-      "learning_rate": 0.00042669351739376836,
-      "loss": 0.0201,
+      "grad_norm": 0.1847366988658905,
+      "learning_rate": 9.9666928389484e-05,
+      "loss": 0.0205,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.16792021691799164,
-      "learning_rate": 0.00022148081700229912,
-      "loss": 0.0503,
+      "grad_norm": 0.13189482688903809,
+      "learning_rate": 5.975745811745197e-05,
+      "loss": 0.0124,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.0907050222158432,
-      "learning_rate": 8.038929879251136e-06,
-      "loss": 0.0099,
+      "grad_norm": 0.11474896222352982,
+      "learning_rate": 1.2496157532491428e-05,
+      "loss": 0.0074,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.4973928413610252,
-      "eval_e3_equal": 0.7912505523641185,
+      "eval_all_ligands_equal": 0.5305346884666372,
+      "eval_e3_equal": 0.8043305346884666,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9989394608926204,
-      "eval_e3_heavy_atoms_difference": 0.23694211224038886,
-      "eval_e3_heavy_atoms_difference_norm": -0.0008665462154205574,
+      "eval_e3_has_attachment_point(s)": 0.993813521873619,
+      "eval_e3_heavy_atoms_difference": 0.3033141847105612,
+      "eval_e3_heavy_atoms_difference_norm": 0.0031366882148514846,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9989394608926204,
-      "eval_has_all_attachment_points": 0.9981440565620857,
-      "eval_has_three_substructures": 0.9999116217410517,
-      "eval_heavy_atoms_difference": 1.7666813963764914,
-      "eval_heavy_atoms_difference_norm": 0.020141735293666703,
-      "eval_linker_equal": 0.7266460450729121,
+      "eval_e3_valid": 0.993813521873619,
+      "eval_has_all_attachment_points": 0.989041095890411,
+      "eval_has_three_substructures": 0.9995581087052585,
+      "eval_heavy_atoms_difference": 3.9929297392841363,
+      "eval_heavy_atoms_difference_norm": 0.05206019398139567,
+      "eval_linker_equal": 0.762439239946973,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.999116217410517,
-      "eval_linker_heavy_atoms_difference": 0.38329650905877155,
-      "eval_linker_heavy_atoms_difference_norm": -0.008708414904468173,
+      "eval_linker_has_attachment_point(s)": 0.9987627043747238,
+      "eval_linker_heavy_atoms_difference": 0.31524524966858153,
+      "eval_linker_heavy_atoms_difference_norm": -0.003168599779403479,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.999116217410517,
-      "eval_loss": 0.29932019114494324,
-      "eval_num_fragments": 2.9999116217410515,
-      "eval_poi_equal": 0.7504197967300045,
+      "eval_linker_valid": 0.9987627043747238,
+      "eval_loss": 0.2911645174026489,
+      "eval_num_fragments": 3.000265134776845,
+      "eval_poi_equal": 0.7620857269111798,
       "eval_poi_graph_edit_distance": 9.999999999999999e+63,
       "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9897481219619974,
-      "eval_poi_heavy_atoms_difference": 0.4080424215642952,
-      "eval_poi_heavy_atoms_difference_norm": -0.0001784973990885129,
+      "eval_poi_has_attachment_point(s)": 0.9606716747680071,
+      "eval_poi_heavy_atoms_difference": 1.122227132125497,
+      "eval_poi_heavy_atoms_difference_norm": 0.0297152624658196,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9897481219619974,
-      "eval_reassembly": 0.5049049933716305,
-      "eval_reassembly_nostereo": 0.5323906319045515,
-      "eval_runtime": 1523.5108,
-      "eval_samples_per_second": 7.427,
-      "eval_steps_per_second": 0.116,
+      "eval_poi_valid": 0.9606716747680071,
+      "eval_reassembly": 0.5367211665930182,
+      "eval_reassembly_nostereo": 0.5691559876270438,
+      "eval_runtime": 1513.244,
+      "eval_samples_per_second": 7.477,
+      "eval_steps_per_second": 0.117,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.9880689350419797,
+      "eval_valid": 0.9542200618647813,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.17162871360778809,
-      "learning_rate": 0.00033406568194236334,
-      "loss": 0.0091,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.6462278962135315,
-      "learning_rate": 0.00036296734876936724,
-      "loss": 0.0531,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.24894048273563385,
-      "learning_rate": 2.0581448478610224e-05,
-      "loss": 0.0333,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.2341412603855133,
-      "learning_rate": 0.00018547973538572384,
-      "loss": 0.0182,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 1.3385168313980103,
-      "learning_rate": 0.0004345286134851596,
-      "loss": 0.0346,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.02978347326557667,
-      "eval_e3_equal": 0.28678745028722935,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.962439239946973,
-      "eval_e3_heavy_atoms_difference": 5.146884666372072,
-      "eval_e3_heavy_atoms_difference_norm": 0.14751881702611244,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.962439239946973,
-      "eval_has_all_attachment_points": 0.9655324790101635,
-      "eval_has_three_substructures": 0.9856827220503757,
-      "eval_heavy_atoms_difference": 23.890676093680955,
-      "eval_heavy_atoms_difference_norm": 0.2898808104790617,
-      "eval_linker_equal": 0.16208572691117984,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9818824569155987,
-      "eval_linker_heavy_atoms_difference": 4.136367653557225,
-      "eval_linker_heavy_atoms_difference_norm": 0.1652414940413321,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9818824569155987,
-      "eval_loss": 0.4170914888381958,
-      "eval_num_fragments": 2.979849756959788,
-      "eval_poi_equal": 0.29155987627043745,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.8293415819708352,
-      "eval_poi_heavy_atoms_difference": 7.791515687140963,
-      "eval_poi_heavy_atoms_difference_norm": 0.19881068969693025,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.8293415819708352,
-      "eval_reassembly": 0.030844012372956253,
-      "eval_reassembly_nostereo": 0.03561643835616438,
-      "eval_runtime": 1296.9734,
-      "eval_samples_per_second": 8.724,
-      "eval_steps_per_second": 0.136,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.8080424215642952,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.3832992911338806,
-      "learning_rate": 0.00012866172649195304,
-      "loss": 0.0432,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.3067149817943573,
-      "learning_rate": 5.274202368198822e-05,
-      "loss": 0.0177,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 1.9233143329620361,
-      "learning_rate": 0.00040158157654602796,
-      "loss": 0.0211,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.3631303012371063,
-      "learning_rate": 0.0002800483785552586,
-      "loss": 0.0567,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.31977763772010803,
-      "learning_rate": 1.5065967452169716e-09,
-      "loss": 0.0235,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.3239946973044631,
-      "eval_e3_equal": 0.67220503756076,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9982324348210341,
-      "eval_e3_heavy_atoms_difference": 0.16783031374281926,
-      "eval_e3_heavy_atoms_difference_norm": -0.011426042102501486,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9982324348210341,
-      "eval_has_all_attachment_points": 0.9994697304463102,
-      "eval_has_three_substructures": 1.0,
-      "eval_heavy_atoms_difference": 0.8422448077772868,
-      "eval_heavy_atoms_difference_norm": 0.005544369955712167,
-      "eval_linker_equal": 0.531064958020327,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9996464869642068,
-      "eval_linker_heavy_atoms_difference": 0.36906760936809546,
-      "eval_linker_heavy_atoms_difference_norm": -0.041618462345356454,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9996464869642068,
-      "eval_loss": 0.34475502371788025,
-      "eval_num_fragments": 3.0,
-      "eval_poi_equal": 0.6563853292090146,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.9961997348652232,
-      "eval_poi_heavy_atoms_difference": -0.12010605391073796,
-      "eval_poi_heavy_atoms_difference_norm": -0.025590422637123363,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.9961997348652232,
-      "eval_reassembly": 0.32841361025187804,
-      "eval_reassembly_nostereo": 0.35024304021210784,
-      "eval_runtime": 1619.6426,
-      "eval_samples_per_second": 6.986,
-      "eval_steps_per_second": 0.109,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.994078656650464,
-      "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": 8.729673191023411e+16,
+  "total_flos": 4.364959997072179e+16,
   "train_batch_size": 128,
-  "trial_name": "trial-number=10-learning_rate=4.4e-04-warmup_ratio=0.020-num_cycles=5.500",
+  "trial_name": "trial-number=11-learning_rate=1.0e-04-warmup_ratio=0.080-num_cycles=3.000",
   "trial_params": {
-    "learning_rate": 0.00043654768433250334,
-    "num_cycles": 5.5,
-    "warmup_ratio": 0.02
+    "learning_rate": 0.00010390406291286528,
+    "num_cycles": 3.0,
+    "warmup_ratio": 0.08
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e14b6bd110aad7baabf4d5d6d4532bfbd9b62a621f30d72e77a3bf09db70d9c
+oid sha256:df253bfc107f1530386928af53b53aa5d613d5622ef461f94f7df0220f62cfd2
 size 7480