{
  "best_metric": 0.5477160056657224,
  "best_model_checkpoint": "/mimer/NOBACKUP/groups/naiss2023-6-290/stefano/models//PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25/checkpoint-10000",
  "epoch": 4.438745314657724,
  "eval_steps": 2500,
  "global_step": 45000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.049319392385085814,
      "grad_norm": 0.9255660772323608,
      "learning_rate": 3.57653791130186e-05,
      "loss": 2.3662,
      "step": 500
    },
    {
      "epoch": 0.09863878477017163,
      "grad_norm": 0.6119515895843506,
      "learning_rate": 4.987090502485352e-05,
      "loss": 0.3224,
      "step": 1000
    },
    {
      "epoch": 0.14795817715525744,
      "grad_norm": 0.4047534763813019,
      "learning_rate": 4.9090579537757233e-05,
      "loss": 0.1093,
      "step": 1500
    },
    {
      "epoch": 0.19727756954034326,
      "grad_norm": 0.3516473174095154,
      "learning_rate": 4.762477336752997e-05,
      "loss": 0.0539,
      "step": 2000
    },
    {
      "epoch": 0.24659696192542907,
      "grad_norm": 0.2518511414527893,
      "learning_rate": 4.5515195009118114e-05,
      "loss": 0.0324,
      "step": 2500
    },
    {
      "epoch": 0.2959163543105149,
      "grad_norm": 0.18100735545158386,
      "learning_rate": 4.282187104726926e-05,
      "loss": 0.0224,
      "step": 3000
    },
    {
      "epoch": 0.3452357466956007,
      "grad_norm": 0.17784442007541656,
      "learning_rate": 3.962143814181834e-05,
      "loss": 0.0167,
      "step": 3500
    },
    {
      "epoch": 0.3945551390806865,
      "grad_norm": 0.15665775537490845,
      "learning_rate": 3.600496238484107e-05,
      "loss": 0.0127,
      "step": 4000
    },
    {
      "epoch": 0.4438745314657723,
      "grad_norm": 0.190831258893013,
      "learning_rate": 3.207534807835037e-05,
      "loss": 0.0101,
      "step": 4500
    },
    {
      "epoch": 0.49319392385085814,
      "grad_norm": 0.17124462127685547,
      "learning_rate": 2.7944409663963666e-05,
      "loss": 0.0086,
      "step": 5000
    },
    {
      "epoch": 0.49319392385085814,
      "eval_all_ligands_equal": 0.4899079320113314,
      "eval_e3_equal": 0.7831975920679887,
      "eval_e3_graph_edit_distance": Infinity,
      "eval_e3_graph_edit_distance_norm": Infinity,
      "eval_e3_has_attachment_point(s)": 0.9948654390934845,
      "eval_e3_heavy_atoms_difference": 0.35915368271954673,
      "eval_e3_heavy_atoms_difference_norm": 0.004063330129381431,
      "eval_e3_tanimoto_similarity": 0.0,
      "eval_e3_valid": 0.9948654390934845,
      "eval_has_all_attachment_points": 0.9820290368271954,
      "eval_has_three_substructures": 0.9988491501416431,
      "eval_heavy_atoms_difference": 7.435021246458923,
      "eval_heavy_atoms_difference_norm": 0.09742815646256632,
      "eval_linker_equal": 0.7091890934844193,
      "eval_linker_graph_edit_distance": 4.603399433427762e+61,
      "eval_linker_graph_edit_distance_norm": 0.054428620142835865,
      "eval_linker_has_attachment_point(s)": 0.9953966005665722,
      "eval_linker_heavy_atoms_difference": 0.6527089235127479,
      "eval_linker_heavy_atoms_difference_norm": 0.02000991525067224,
      "eval_linker_tanimoto_similarity": 0.0,
      "eval_linker_valid": 0.9953966005665722,
      "eval_loss": 0.293075829744339,
      "eval_num_fragments": 3.0004426345609065,
      "eval_poi_equal": 0.7415899433427762,
      "eval_poi_graph_edit_distance": Infinity,
      "eval_poi_graph_edit_distance_norm": Infinity,
      "eval_poi_has_attachment_point(s)": 0.9247521246458924,
      "eval_poi_heavy_atoms_difference": 2.4198831444759206,
      "eval_poi_heavy_atoms_difference_norm": 0.07853231753644796,
      "eval_poi_tanimoto_similarity": 0.0,
      "eval_poi_valid": 0.9247521246458924,
      "eval_reassembly": 0.4971671388101983,
      "eval_reassembly_nostereo": 0.5238137393767706,
      "eval_runtime": 2417.0464,
      "eval_samples_per_second": 4.673,
      "eval_steps_per_second": 0.073,
      "eval_tanimoto_similarity": 0.0,
      "eval_valid": 0.9182011331444759,
      "step": 5000
    },
    {
      "epoch": 0.542513316235944,
      "grad_norm": 0.12277983129024506,
      "learning_rate": 2.372969012074575e-05,
      "loss": 0.007,
      "step": 5500
    },
    {
      "epoch": 0.5918327086210298,
      "grad_norm": 0.11204977333545685,
      "learning_rate": 1.9551116361503457e-05,
      "loss": 0.006,
      "step": 6000
    },
    {
      "epoch": 0.6411521010061156,
      "grad_norm": 0.1196436733007431,
      "learning_rate": 1.55275867959039e-05,
      "loss": 0.0051,
      "step": 6500
    },
    {
      "epoch": 0.6904714933912014,
      "grad_norm": 0.12161414325237274,
      "learning_rate": 1.1773588158934163e-05,
      "loss": 0.0045,
      "step": 7000
    },
    {
      "epoch": 0.7397908857762873,
      "grad_norm": 0.14829416573047638,
      "learning_rate": 8.395937870495241e-06,
      "loss": 0.004,
      "step": 7500
    },
    {
      "epoch": 0.7397908857762873,
      "eval_all_ligands_equal": 0.5380665722379604,
      "eval_e3_equal": 0.8043555240793201,
      "eval_e3_graph_edit_distance": Infinity,
      "eval_e3_graph_edit_distance_norm": Infinity,
      "eval_e3_has_attachment_point(s)": 0.9923866855524079,
      "eval_e3_heavy_atoms_difference": 0.4004957507082153,
      "eval_e3_heavy_atoms_difference_norm": 0.00806031722408992,
      "eval_e3_tanimoto_similarity": 0.0,
      "eval_e3_valid": 0.9923866855524079,
      "eval_has_all_attachment_points": 0.9817634560906515,
      "eval_has_three_substructures": 0.9977868271954674,
      "eval_heavy_atoms_difference": 6.986809490084986,
      "eval_heavy_atoms_difference_norm": 0.09237955806043094,
      "eval_linker_equal": 0.7604461756373938,
      "eval_linker_graph_edit_distance": 5.93130311614731e+61,
      "eval_linker_graph_edit_distance_norm": Infinity,
      "eval_linker_has_attachment_point(s)": 0.9940686968838527,
      "eval_linker_heavy_atoms_difference": 0.24982294617563738,
      "eval_linker_heavy_atoms_difference_norm": -0.0010343910857730165,
      "eval_linker_tanimoto_similarity": 0.0,
      "eval_linker_valid": 0.9940686968838527,
      "eval_loss": 0.30975914001464844,
      "eval_num_fragments": 3.000619688385269,
      "eval_poi_equal": 0.7602691218130312,
      "eval_poi_graph_edit_distance": Infinity,
      "eval_poi_graph_edit_distance_norm": Infinity,
      "eval_poi_has_attachment_point(s)": 0.9225389518413598,
      "eval_poi_heavy_atoms_difference": 2.202372521246459,
      "eval_poi_heavy_atoms_difference_norm": 0.07091974582407527,
      "eval_poi_tanimoto_similarity": 0.0,
      "eval_poi_valid": 0.9225389518413598,
      "eval_reassembly": 0.5450601983002833,
      "eval_reassembly_nostereo": 0.5729461756373938,
      "eval_runtime": 2364.3909,
      "eval_samples_per_second": 4.778,
      "eval_steps_per_second": 0.075,
      "eval_tanimoto_similarity": 0.0,
      "eval_valid": 0.9147485835694051,
      "step": 7500
    },
    {
      "epoch": 0.789110278161373,
      "grad_norm": 0.11315891146659851,
      "learning_rate": 5.490744620021837e-06,
      "loss": 0.0036,
      "step": 8000
    },
    {
      "epoch": 0.8384296705464589,
      "grad_norm": 0.10538846999406815,
      "learning_rate": 3.1406736605785858e-06,
      "loss": 0.0034,
      "step": 8500
    },
    {
      "epoch": 0.8877490629315447,
      "grad_norm": 0.09490078687667847,
      "learning_rate": 1.412594626586694e-06,
      "loss": 0.0032,
      "step": 9000
    },
    {
      "epoch": 0.9370684553166305,
      "grad_norm": 0.08204534649848938,
      "learning_rate": 3.5567880489165705e-07,
      "loss": 0.0031,
      "step": 9500
    },
    {
      "epoch": 0.9863878477017163,
      "grad_norm": 0.10336631536483765,
      "learning_rate": 0.0,
      "loss": 0.003,
      "step": 10000
    },
    {
      "epoch": 0.9863878477017163,
      "eval_all_ligands_equal": 0.5477160056657224,
      "eval_e3_equal": 0.8034702549575071,
      "eval_e3_graph_edit_distance": Infinity,
      "eval_e3_graph_edit_distance_norm": Infinity,
      "eval_e3_has_attachment_point(s)": 0.9942457507082153,
      "eval_e3_heavy_atoms_difference": 0.3627832861189802,
      "eval_e3_heavy_atoms_difference_norm": 0.004441809193160064,
      "eval_e3_tanimoto_similarity": 0.0,
      "eval_e3_valid": 0.9942457507082153,
      "eval_has_all_attachment_points": 0.9856586402266289,
      "eval_has_three_substructures": 0.9983179886685553,
      "eval_heavy_atoms_difference": 6.492917847025495,
      "eval_heavy_atoms_difference_norm": 0.08538313201004577,
      "eval_linker_equal": 0.7725743626062322,
      "eval_linker_graph_edit_distance": Infinity,
      "eval_linker_graph_edit_distance_norm": Infinity,
      "eval_linker_has_attachment_point(s)": 0.9951310198300283,
      "eval_linker_heavy_atoms_difference": 0.32524787535410765,
      "eval_linker_heavy_atoms_difference_norm": 0.005000241007082242,
      "eval_linker_tanimoto_similarity": 0.0,
      "eval_linker_valid": 0.9951310198300283,
      "eval_loss": 0.312394380569458,
      "eval_num_fragments": 3.0007967422096318,
      "eval_poi_equal": 0.7672627478753541,
      "eval_poi_graph_edit_distance": Infinity,
      "eval_poi_graph_edit_distance_norm": Infinity,
      "eval_poi_has_attachment_point(s)": 0.9293555240793201,
      "eval_poi_heavy_atoms_difference": 2.0848973087818696,
      "eval_poi_heavy_atoms_difference_norm": 0.06681110633224353,
      "eval_poi_tanimoto_similarity": 0.0,
      "eval_poi_valid": 0.9293555240793201,
      "eval_reassembly": 0.5548866855524079,
      "eval_reassembly_nostereo": 0.5845432011331445,
      "eval_runtime": 2338.9582,
      "eval_samples_per_second": 4.83,
      "eval_steps_per_second": 0.076,
      "eval_tanimoto_similarity": 0.0,
      "eval_valid": 0.9231586402266289,
      "step": 10000
    },
    {
      "epoch": 1.0357072400868022,
      "grad_norm": 0.10573304444551468,
      "learning_rate": 4.867483222184158e-05,
      "loss": 0.0134,
      "step": 10500
    },
    {
      "epoch": 1.085026632471888,
      "grad_norm": 0.07383856922388077,
      "learning_rate": 4.8545628694730624e-05,
      "loss": 0.0058,
      "step": 11000
    },
    {
      "epoch": 1.1343460248569737,
      "grad_norm": 0.08897178620100021,
      "learning_rate": 4.841060399978481e-05,
      "loss": 0.0045,
      "step": 11500
    },
    {
      "epoch": 1.1836654172420595,
      "grad_norm": 0.08299541473388672,
      "learning_rate": 4.826979151905655e-05,
      "loss": 0.0042,
      "step": 12000
    },
    {
      "epoch": 1.2329848096271454,
      "grad_norm": 0.08171670883893967,
      "learning_rate": 4.812322606550813e-05,
      "loss": 0.004,
      "step": 12500
    },
    {
      "epoch": 1.2823042020122313,
      "grad_norm": 0.08833315223455429,
      "learning_rate": 4.797094387440491e-05,
      "loss": 0.0037,
      "step": 13000
    },
    {
      "epoch": 1.3316235943973171,
      "grad_norm": 0.04986245930194855,
      "learning_rate": 4.78129825943569e-05,
      "loss": 0.0034,
      "step": 13500
    },
    {
      "epoch": 1.3809429867824028,
      "grad_norm": 0.07238644361495972,
      "learning_rate": 4.7649381278011e-05,
      "loss": 0.0038,
      "step": 14000
    },
    {
      "epoch": 1.4302623791674887,
      "grad_norm": 0.05512389540672302,
      "learning_rate": 4.748018037239592e-05,
      "loss": 0.003,
      "step": 14500
    },
    {
      "epoch": 1.4795817715525745,
      "grad_norm": 0.058981988579034805,
      "learning_rate": 4.73054217089226e-05,
      "loss": 0.0027,
      "step": 15000
    },
    {
      "epoch": 1.5289011639376602,
      "grad_norm": 0.07365080714225769,
      "learning_rate": 4.712514849304219e-05,
      "loss": 0.003,
      "step": 15500
    },
    {
      "epoch": 1.578220556322746,
      "grad_norm": 0.04787248373031616,
      "learning_rate": 4.693940529356444e-05,
      "loss": 0.0025,
      "step": 16000
    },
    {
      "epoch": 1.627539948707832,
      "grad_norm": 0.0840035229921341,
      "learning_rate": 4.674823803163899e-05,
      "loss": 0.0023,
      "step": 16500
    },
    {
      "epoch": 1.6768593410929178,
      "grad_norm": 0.05743684619665146,
      "learning_rate": 4.655169396940229e-05,
      "loss": 0.0024,
      "step": 17000
    },
    {
      "epoch": 1.7261787334780037,
      "grad_norm": 0.048048391938209534,
      "learning_rate": 4.6349821698293025e-05,
      "loss": 0.0022,
      "step": 17500
    },
    {
      "epoch": 1.7754981258630895,
      "grad_norm": 0.062112707644701004,
      "learning_rate": 4.6142671127038905e-05,
      "loss": 0.0026,
      "step": 18000
    },
    {
      "epoch": 1.8248175182481752,
      "grad_norm": 0.034723177552223206,
      "learning_rate": 4.593029346931777e-05,
      "loss": 0.0028,
      "step": 18500
    },
    {
      "epoch": 1.874136910633261,
      "grad_norm": 0.0403965599834919,
      "learning_rate": 4.571274123109606e-05,
      "loss": 0.0016,
      "step": 19000
    },
    {
      "epoch": 1.9234563030183467,
      "grad_norm": 0.022397898137569427,
      "learning_rate": 4.549006819764779e-05,
      "loss": 0.0016,
      "step": 19500
    },
    {
      "epoch": 1.9727756954034326,
      "grad_norm": 0.04490479454398155,
      "learning_rate": 4.52623294202573e-05,
      "loss": 0.0017,
      "step": 20000
    },
    {
      "epoch": 2.0220950877885184,
      "grad_norm": 0.047866348177194595,
      "learning_rate": 4.502958120260894e-05,
      "loss": 0.0017,
      "step": 20500
    },
    {
      "epoch": 2.0714144801736043,
      "grad_norm": 0.053165681660175323,
      "learning_rate": 4.479188108686714e-05,
      "loss": 0.0019,
      "step": 21000
    },
    {
      "epoch": 2.12073387255869,
      "grad_norm": 0.04063253104686737,
      "learning_rate": 4.4549287839450324e-05,
      "loss": 0.0017,
      "step": 21500
    },
    {
      "epoch": 2.170053264943776,
      "grad_norm": 0.059268295764923096,
      "learning_rate": 4.4301861436502156e-05,
      "loss": 0.0015,
      "step": 22000
    },
    {
      "epoch": 2.219372657328862,
      "grad_norm": 0.021407226100564003,
      "learning_rate": 4.404966304906363e-05,
      "loss": 0.0014,
      "step": 22500
    },
    {
      "epoch": 2.2686920497139473,
      "grad_norm": 0.027945173904299736,
      "learning_rate": 4.379275502794983e-05,
      "loss": 0.0014,
      "step": 23000
    },
    {
      "epoch": 2.318011442099033,
      "grad_norm": 0.03261112794280052,
      "learning_rate": 4.353120088833501e-05,
      "loss": 0.0014,
      "step": 23500
    },
    {
      "epoch": 2.367330834484119,
      "grad_norm": 0.05259308964014053,
      "learning_rate": 4.326506529404972e-05,
      "loss": 0.0013,
      "step": 24000
    },
    {
      "epoch": 2.416650226869205,
      "grad_norm": 0.0584435798227787,
      "learning_rate": 4.2994959806435226e-05,
      "loss": 0.0015,
      "step": 24500
    },
    {
      "epoch": 2.465969619254291,
      "grad_norm": 0.022548576816916466,
      "learning_rate": 4.2719868638689734e-05,
      "loss": 0.0012,
      "step": 25000
    },
    {
      "epoch": 2.5152890116393767,
      "grad_norm": 0.03915947675704956,
      "learning_rate": 4.244039660132742e-05,
      "loss": 0.0013,
      "step": 25500
    },
    {
      "epoch": 2.5646084040244626,
      "grad_norm": 0.02655004896223545,
      "learning_rate": 4.2156612788007085e-05,
      "loss": 0.0012,
      "step": 26000
    },
    {
      "epoch": 2.6139277964095484,
      "grad_norm": 0.030596962198615074,
      "learning_rate": 4.186858735838457e-05,
      "loss": 0.0012,
      "step": 26500
    },
    {
      "epoch": 2.6632471887946343,
      "grad_norm": 0.03012579120695591,
      "learning_rate": 4.15763915207673e-05,
      "loss": 0.0012,
      "step": 27000
    },
    {
      "epoch": 2.7125665811797197,
      "grad_norm": 0.015622408129274845,
      "learning_rate": 4.1280097514509486e-05,
      "loss": 0.001,
      "step": 27500
    },
    {
      "epoch": 2.7618859735648056,
      "grad_norm": 0.03695574030280113,
      "learning_rate": 4.097977859215244e-05,
      "loss": 0.0011,
      "step": 28000
    },
    {
      "epoch": 2.8112053659498915,
      "grad_norm": 0.03273458406329155,
      "learning_rate": 4.0675509001314444e-05,
      "loss": 0.0011,
      "step": 28500
    },
    {
      "epoch": 2.8605247583349773,
      "grad_norm": 0.023473775014281273,
      "learning_rate": 4.036798407361931e-05,
      "loss": 0.0011,
      "step": 29000
    },
    {
      "epoch": 2.909844150720063,
      "grad_norm": 0.0245220847427845,
      "learning_rate": 4.005604729882533e-05,
      "loss": 0.0009,
      "step": 29500
    },
    {
      "epoch": 2.959163543105149,
      "grad_norm": 0.025120645761489868,
      "learning_rate": 3.974038822893393e-05,
      "loss": 0.001,
      "step": 30000
    },
    {
      "epoch": 3.008482935490235,
      "grad_norm": 0.03636594116687775,
      "learning_rate": 3.942108490409635e-05,
      "loss": 0.0008,
      "step": 30500
    },
    {
      "epoch": 3.0578023278753204,
      "grad_norm": 0.02662217803299427,
      "learning_rate": 3.9098865507980035e-05,
      "loss": 0.0011,
      "step": 31000
    },
    {
      "epoch": 3.1071217202604062,
      "grad_norm": 0.014715551398694515,
      "learning_rate": 3.8772518268796315e-05,
      "loss": 0.0009,
      "step": 31500
    },
    {
      "epoch": 3.156441112645492,
      "grad_norm": 0.024017762392759323,
      "learning_rate": 3.8442766060418925e-05,
      "loss": 0.0008,
      "step": 32000
    },
    {
      "epoch": 3.205760505030578,
      "grad_norm": 0.042992159724235535,
      "learning_rate": 3.8109690407234974e-05,
      "loss": 0.0008,
      "step": 32500
    },
    {
      "epoch": 3.255079897415664,
      "grad_norm": 0.05144192650914192,
      "learning_rate": 3.777337365528434e-05,
      "loss": 0.0009,
      "step": 33000
    },
    {
      "epoch": 3.3043992898007497,
      "grad_norm": 0.02588796801865101,
      "learning_rate": 3.7433898951901246e-05,
      "loss": 0.0008,
      "step": 33500
    },
    {
      "epoch": 3.3537186821858356,
      "grad_norm": 0.03916551172733307,
      "learning_rate": 3.709135022515784e-05,
      "loss": 0.0008,
      "step": 34000
    },
    {
      "epoch": 3.4030380745709214,
      "grad_norm": 0.013717463240027428,
      "learning_rate": 3.674581216311475e-05,
      "loss": 0.0008,
      "step": 34500
    },
    {
      "epoch": 3.4523574669560073,
      "grad_norm": 0.07564926147460938,
      "learning_rate": 3.639806991778405e-05,
      "loss": 0.0009,
      "step": 35000
    },
    {
      "epoch": 3.501676859341093,
      "grad_norm": 0.023237833753228188,
      "learning_rate": 3.604752099655735e-05,
      "loss": 0.0009,
      "step": 35500
    },
    {
      "epoch": 3.5509962517261786,
      "grad_norm": 0.034029532223939896,
      "learning_rate": 3.569354109150707e-05,
      "loss": 0.0006,
      "step": 36000
    },
    {
      "epoch": 3.6003156441112645,
      "grad_norm": 0.04032694920897484,
      "learning_rate": 3.533691743044375e-05,
      "loss": 0.0007,
      "step": 36500
    },
    {
      "epoch": 3.6496350364963503,
      "grad_norm": 0.03434913605451584,
      "learning_rate": 3.497773818116328e-05,
      "loss": 0.0007,
      "step": 37000
    },
    {
      "epoch": 3.698954428881436,
      "grad_norm": 0.012562757357954979,
      "learning_rate": 3.461609214327767e-05,
      "loss": 0.0007,
      "step": 37500
    },
    {
      "epoch": 3.748273821266522,
      "grad_norm": 0.03795718029141426,
      "learning_rate": 3.4253529436582336e-05,
      "loss": 0.0012,
      "step": 38000
    },
    {
      "epoch": 3.797593213651608,
      "grad_norm": 0.036550700664520264,
      "learning_rate": 3.3887227607163154e-05,
      "loss": 0.0007,
      "step": 38500
    },
    {
      "epoch": 3.8469126060366934,
      "grad_norm": 0.009776749648153782,
      "learning_rate": 3.3518728595236816e-05,
      "loss": 0.0006,
      "step": 39000
    },
    {
      "epoch": 3.8962319984217793,
      "grad_norm": 0.010442009195685387,
      "learning_rate": 3.314812350453317e-05,
      "loss": 0.0006,
      "step": 39500
    },
    {
      "epoch": 3.945551390806865,
      "grad_norm": 0.023100633174180984,
      "learning_rate": 3.27755039594663e-05,
      "loss": 0.0006,
      "step": 40000
    },
    {
      "epoch": 3.994870783191951,
      "grad_norm": 0.028563622385263443,
      "learning_rate": 3.2400962082482306e-05,
      "loss": 0.0007,
      "step": 40500
    },
    {
      "epoch": 4.044190175577037,
      "grad_norm": 0.012048379518091679,
      "learning_rate": 3.2024590471283946e-05,
      "loss": 0.0006,
      "step": 41000
    },
    {
      "epoch": 4.093509567962123,
      "grad_norm": 0.01336103118956089,
      "learning_rate": 3.164648217593777e-05,
      "loss": 0.0006,
      "step": 41500
    },
    {
      "epoch": 4.142828960347209,
      "grad_norm": 0.04073134437203407,
      "learning_rate": 3.1266730675869434e-05,
      "loss": 0.0006,
      "step": 42000
    },
    {
      "epoch": 4.1921483527322945,
      "grad_norm": 0.03634357079863548,
      "learning_rate": 3.088542985675283e-05,
      "loss": 0.0006,
      "step": 42500
    },
    {
      "epoch": 4.24146774511738,
      "grad_norm": 0.03682945668697357,
      "learning_rate": 3.0502673987298837e-05,
      "loss": 0.0007,
      "step": 43000
    },
    {
      "epoch": 4.290787137502466,
      "grad_norm": 0.030003152787685394,
      "learning_rate": 3.0119327223166655e-05,
      "loss": 0.0007,
      "step": 43500
    },
    {
      "epoch": 4.340106529887552,
      "grad_norm": 0.025118093937635422,
      "learning_rate": 2.973471986909034e-05,
      "loss": 0.0008,
      "step": 44000
    },
    {
      "epoch": 4.389425922272638,
      "grad_norm": 0.01078966073691845,
      "learning_rate": 2.9348172431675086e-05,
      "loss": 0.0005,
      "step": 44500
    },
    {
      "epoch": 4.438745314657724,
      "grad_norm": 0.020486511290073395,
      "learning_rate": 2.8960549998903897e-05,
      "loss": 0.0005,
      "step": 45000
    }
  ],
  "logging_steps": 500,
  "max_steps": 100000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 5000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 3.898212104814858e+17,
  "train_batch_size": 128,
  "trial_name": null,
  "trial_params": null
}