Training in progress, step 5000, checkpoint

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8c23742e474292a7c55d08fa35da5d386b6dc94e75b0758af539411b0a6300b0
+oid sha256:8c0489f4d8a0feab5d3e38c38764efb3065a72c0df1166c52788a4fb8b5db167
 size 409608164

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bcbac88bb42d64b5113d8b7784064778f08387dc103774db1d3e75a7d51df86b
+oid sha256:10a21a5a2f88c53aa60c7d7a14adc335d7340e642ca9d190920e4e42bed1705e
 size 814647162

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:551f4e1407f546402384ab496e542d88dbab87683febd9da9183e2d8deccce45
+oid sha256:12f5435a54f71ddfe7351c79962b19289d71b582fb8054a3e03ea18859c1e069
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98a089bb4a69e00a7e8a8814929d1de493c84d2bd6273a6f081d3b7a9500a40d
+oid sha256:a7c3ad366a455d36d790316ad7de46df5551299c0b50500acfaef90eebc7bbcf
 size 1064

last-checkpoint/trainer_state.json

@@ -1,278 +1,124 @@
 {
-  "best_metric": 0.0009721608484312859,
-  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=20-learning_rate=4.8e-06-warmup_ratio=0.030-num_cycles=3.000/checkpoint-10000",
-  "epoch": 0.9863878477017163,
+  "best_metric": 0.0,
+  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/trial-number=21-learning_rate=1.6e-06-warmup_ratio=0.080-num_cycles=1.500/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": 2.0794825553894043,
-      "learning_rate": 3.4057266290417822e-06,
-      "loss": 4.898,
+      "grad_norm": 4.178253650665283,
+      "learning_rate": 1.1556051821104526e-06,
+      "loss": 6.3981,
       "step": 500
     },
     {
       "epoch": 0.09863878477017163,
-      "grad_norm": 1.3309763669967651,
-      "learning_rate": 4.331843781954863e-06,
-      "loss": 1.2931,
+      "grad_norm": 1.9749269485473633,
+      "learning_rate": 1.5782537021726088e-06,
+      "loss": 2.3749,
       "step": 1000
     },
     {
       "epoch": 0.14795817715525744,
-      "grad_norm": 1.2491517066955566,
-      "learning_rate": 2.2556236723022006e-06,
-      "loss": 0.8498,
+      "grad_norm": 1.4726980924606323,
+      "learning_rate": 1.3637510062898264e-06,
+      "loss": 1.4006,
       "step": 1500
     },
     {
       "epoch": 0.19727756954034326,
-      "grad_norm": 1.1954443454742432,
-      "learning_rate": 2.9711989198042844e-07,
-      "loss": 0.7115,
+      "grad_norm": 1.4031742811203003,
+      "learning_rate": 1.0095554213221014e-06,
+      "loss": 1.1642,
       "step": 2000
     },
     {
       "epoch": 0.24659696192542907,
-      "grad_norm": 1.1868677139282227,
-      "learning_rate": 3.0102179436748257e-07,
-      "loss": 0.6854,
+      "grad_norm": 1.421210527420044,
+      "learning_rate": 6.046599668215422e-07,
+      "loss": 1.0397,
       "step": 2500
     },
     {
       "epoch": 0.2959163543105149,
-      "grad_norm": 1.359408974647522,
-      "learning_rate": 2.263654228124041e-06,
-      "loss": 0.6669,
+      "grad_norm": 1.4061120748519897,
+      "learning_rate": 2.5079620192718024e-07,
+      "loss": 0.9784,
       "step": 3000
     },
     {
       "epoch": 0.3452357466956007,
-      "grad_norm": 1.3000223636627197,
-      "learning_rate": 4.336439114656795e-06,
-      "loss": 0.5823,
+      "grad_norm": 1.4008420705795288,
+      "learning_rate": 3.687377510596774e-08,
+      "loss": 0.9519,
       "step": 3500
     },
     {
       "epoch": 0.3945551390806865,
-      "grad_norm": 1.0815534591674805,
-      "learning_rate": 4.567047206277767e-06,
-      "loss": 0.4713,
+      "grad_norm": 1.39585280418396,
+      "learning_rate": 1.664152886385891e-08,
+      "loss": 0.9482,
       "step": 4000
     },
     {
       "epoch": 0.4438745314657723,
-      "grad_norm": 0.9021003246307373,
-      "learning_rate": 2.7382717272379126e-06,
-      "loss": 0.385,
+      "grad_norm": 1.3957849740982056,
+      "learning_rate": 1.9518289383020806e-07,
+      "loss": 0.9436,
       "step": 4500
     },
     {
       "epoch": 0.49319392385085814,
-      "grad_norm": 0.8889977335929871,
-      "learning_rate": 5.726126235670513e-07,
-      "loss": 0.3445,
+      "grad_norm": 1.4641075134277344,
+      "learning_rate": 5.276386570231308e-07,
+      "loss": 0.9277,
       "step": 5000
     },
     {
       "epoch": 0.49319392385085814,
-      "eval_all_ligands_equal": 0.00017675651789659743,
-      "eval_e3_equal": 0.13239063190455147,
+      "eval_all_ligands_equal": 0.0,
+      "eval_e3_equal": 0.023243482103402564,
       "eval_e3_graph_edit_distance": 9.999999999999999e+63,
       "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9877154220061864,
-      "eval_e3_heavy_atoms_difference": -1.3631462660185594,
-      "eval_e3_heavy_atoms_difference_norm": -0.07969778182157912,
+      "eval_e3_has_attachment_point(s)": 0.7854175872735307,
+      "eval_e3_heavy_atoms_difference": 7.005744586831639,
+      "eval_e3_heavy_atoms_difference_norm": 0.19202268846027856,
       "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9877154220061864,
-      "eval_has_all_attachment_points": 0.8787450287229341,
-      "eval_has_three_substructures": 0.9895713654441007,
-      "eval_heavy_atoms_difference": 40.59840919133893,
-      "eval_heavy_atoms_difference_norm": 0.5186137998425047,
-      "eval_linker_equal": 0.0030048608042421566,
+      "eval_e3_valid": 0.7854175872735307,
+      "eval_has_all_attachment_points": 0.4845779938135219,
+      "eval_has_three_substructures": 0.7926646045072913,
+      "eval_heavy_atoms_difference": 71.70234202386213,
+      "eval_heavy_atoms_difference_norm": 0.9403941983073605,
+      "eval_linker_equal": 0.0007954043305346885,
       "eval_linker_graph_edit_distance": 9.999999999999999e+63,
       "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.97870083959346,
-      "eval_linker_heavy_atoms_difference": 6.881307998232435,
-      "eval_linker_heavy_atoms_difference_norm": 0.3090588569977506,
+      "eval_linker_has_attachment_point(s)": 0.7095006628369421,
+      "eval_linker_heavy_atoms_difference": 11.820945647370747,
+      "eval_linker_heavy_atoms_difference_norm": 0.6561966279132151,
       "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.97870083959346,
-      "eval_loss": 0.4682455062866211,
-      "eval_num_fragments": 2.995581087052585,
-      "eval_poi_equal": 0.053645603181617325,
+      "eval_linker_valid": 0.7095006628369421,
+      "eval_loss": 0.9539049863815308,
+      "eval_num_fragments": 3.1619089703932834,
+      "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)": 0.5663278833406982,
-      "eval_poi_heavy_atoms_difference": 16.751922227132127,
-      "eval_poi_heavy_atoms_difference_norm": 0.4869952919604095,
+      "eval_poi_has_attachment_point(s)": 0.15510384445426426,
+      "eval_poi_heavy_atoms_difference": 29.169686257180732,
+      "eval_poi_heavy_atoms_difference_norm": 0.9225394246917481,
       "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.5663278833406982,
-      "eval_reassembly": 0.00017675651789659743,
-      "eval_reassembly_nostereo": 0.0002651347768448962,
-      "eval_runtime": 1486.8909,
-      "eval_samples_per_second": 7.61,
-      "eval_steps_per_second": 0.119,
+      "eval_poi_valid": 0.15510384445426426,
+      "eval_reassembly": 0.0,
+      "eval_reassembly_nostereo": 0.0,
+      "eval_runtime": 2245.7826,
+      "eval_samples_per_second": 5.038,
+      "eval_steps_per_second": 0.079,
       "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.5612903225806452,
+      "eval_valid": 0.14653115333627928,
       "step": 5000
-    },
-    {
-      "epoch": 0.542513316235944,
-      "grad_norm": 0.7243369817733765,
-      "learning_rate": 1.098761656065621e-07,
-      "loss": 0.3332,
-      "step": 5500
-    },
-    {
-      "epoch": 0.5918327086210298,
-      "grad_norm": 0.7796162962913513,
-      "learning_rate": 1.785907831928506e-06,
-      "loss": 0.3291,
-      "step": 6000
-    },
-    {
-      "epoch": 0.6411521010061156,
-      "grad_norm": 0.794780433177948,
-      "learning_rate": 4.02207509608434e-06,
-      "loss": 0.3073,
-      "step": 6500
-    },
-    {
-      "epoch": 0.6904714933912014,
-      "grad_norm": 0.8074055910110474,
-      "learning_rate": 4.712160964261165e-06,
-      "loss": 0.2687,
-      "step": 7000
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "grad_norm": 0.5401957035064697,
-      "learning_rate": 3.206182491961818e-06,
-      "loss": 0.2322,
-      "step": 7500
-    },
-    {
-      "epoch": 0.7397908857762873,
-      "eval_all_ligands_equal": 0.000618647812638091,
-      "eval_e3_equal": 0.2136986301369863,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.993813521873619,
-      "eval_e3_heavy_atoms_difference": -1.129562527618206,
-      "eval_e3_heavy_atoms_difference_norm": -0.0636057389477739,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.993813521873619,
-      "eval_has_all_attachment_points": 0.9152452496685816,
-      "eval_has_three_substructures": 0.9952275740167918,
-      "eval_heavy_atoms_difference": 29.648608042421564,
-      "eval_heavy_atoms_difference_norm": 0.3796774462658793,
-      "eval_linker_equal": 0.0057445868316394165,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9870083959346001,
-      "eval_linker_heavy_atoms_difference": 5.789659743703049,
-      "eval_linker_heavy_atoms_difference_norm": 0.2506175164012725,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9870083959346001,
-      "eval_loss": 0.3937079608440399,
-      "eval_num_fragments": 2.9978789217852406,
-      "eval_poi_equal": 0.16871409633230225,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.691117984975696,
-      "eval_poi_heavy_atoms_difference": 11.474149359257623,
-      "eval_poi_heavy_atoms_difference_norm": 0.33367563586799615,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.691117984975696,
-      "eval_reassembly": 0.000618647812638091,
-      "eval_reassembly_nostereo": 0.0007954043305346885,
-      "eval_runtime": 1476.2637,
-      "eval_samples_per_second": 7.665,
-      "eval_steps_per_second": 0.12,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.6858152894387981,
-      "step": 7500
-    },
-    {
-      "epoch": 0.789110278161373,
-      "grad_norm": 0.6613984704017639,
-      "learning_rate": 9.226013039121686e-07,
-      "loss": 0.2124,
-      "step": 8000
-    },
-    {
-      "epoch": 0.8384296705464589,
-      "grad_norm": 0.6681215167045593,
-      "learning_rate": 1.2292954959110304e-08,
-      "loss": 0.2046,
-      "step": 8500
-    },
-    {
-      "epoch": 0.8877490629315447,
-      "grad_norm": 0.6344013810157776,
-      "learning_rate": 1.3326650929596504e-06,
-      "loss": 0.2034,
-      "step": 9000
-    },
-    {
-      "epoch": 0.9370684553166305,
-      "grad_norm": 0.8783854246139526,
-      "learning_rate": 3.640076296365813e-06,
-      "loss": 0.1964,
-      "step": 9500
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "grad_norm": 0.749924898147583,
-      "learning_rate": 4.7612058274004114e-06,
-      "loss": 0.1799,
-      "step": 10000
-    },
-    {
-      "epoch": 0.9863878477017163,
-      "eval_all_ligands_equal": 0.0009721608484312859,
-      "eval_e3_equal": 0.29155987627043745,
-      "eval_e3_graph_edit_distance": 9.999999999999999e+63,
-      "eval_e3_graph_edit_distance_norm": 1.0,
-      "eval_e3_has_attachment_point(s)": 0.9963764913831198,
-      "eval_e3_heavy_atoms_difference": -1.0244807777286788,
-      "eval_e3_heavy_atoms_difference_norm": -0.05932507175480622,
-      "eval_e3_tanimoto_similarity": 0.0,
-      "eval_e3_valid": 0.9963764913831198,
-      "eval_has_all_attachment_points": 0.9400795404330534,
-      "eval_has_three_substructures": 0.9980556783031375,
-      "eval_heavy_atoms_difference": 22.542465753424658,
-      "eval_heavy_atoms_difference_norm": 0.288556240521232,
-      "eval_linker_equal": 0.008749447635881574,
-      "eval_linker_graph_edit_distance": 9.999999999999999e+63,
-      "eval_linker_graph_edit_distance_norm": 1.0,
-      "eval_linker_has_attachment_point(s)": 0.9902783915156872,
-      "eval_linker_heavy_atoms_difference": 4.60362350861688,
-      "eval_linker_heavy_atoms_difference_norm": 0.1790452881173788,
-      "eval_linker_tanimoto_similarity": 0.0,
-      "eval_linker_valid": 0.9902783915156872,
-      "eval_loss": 0.36306923627853394,
-      "eval_num_fragments": 2.99946973044631,
-      "eval_poi_equal": 0.27370746796288115,
-      "eval_poi_graph_edit_distance": 9.999999999999999e+63,
-      "eval_poi_graph_edit_distance_norm": 1.0,
-      "eval_poi_has_attachment_point(s)": 0.7658859920459566,
-      "eval_poi_heavy_atoms_difference": 8.255148033583739,
-      "eval_poi_heavy_atoms_difference_norm": 0.24317273637314274,
-      "eval_poi_tanimoto_similarity": 0.0,
-      "eval_poi_valid": 0.7658859920459566,
-      "eval_reassembly": 0.0010605391073795847,
-      "eval_reassembly_nostereo": 0.0019443216968625717,
-      "eval_runtime": 1567.7756,
-      "eval_samples_per_second": 7.217,
-      "eval_steps_per_second": 0.113,
-      "eval_tanimoto_similarity": 0.0,
-      "eval_valid": 0.7593460008837826,
-      "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=20-learning_rate=4.8e-06-warmup_ratio=0.030-num_cycles=3.000",
+  "trial_name": "trial-number=21-learning_rate=1.6e-06-warmup_ratio=0.080-num_cycles=1.500",
   "trial_params": {
-    "learning_rate": 4.7612058274004114e-06,
-    "num_cycles": 3.0,
-    "warmup_ratio": 0.03
+    "learning_rate": 1.6155360445904128e-06,
+    "num_cycles": 1.5,
+    "warmup_ratio": 0.08
   }
 }

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecb78bd82e576fde3f68367b3825e6828ecaa9f0f56f0f0c7edba1aa4d63b16c
+oid sha256:840779dc69cef9253cfbecd1665a91f7f3e4ecfabfa209711398b6be673cceed
 size 7416