Training in progress, step 56000, checkpoint

checkpoint-56000/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:720f533079fc6130ce14b30b8eafdb5e4076469e1faa3ba38a4e72947d565b0b
+oid sha256:e1430e45fb9aa1d2dd302daca3a02ada6cc2a15d8b72c15ae1e157f8dab5c87c
 size 51007160

checkpoint-56000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8d601afe35edd64872bb14c04719f9300b470651e7e555f58f816d084fa31c3
+oid sha256:405387cf742a92e6a4a87130d6734b66a495dfa2d16a9efcafee341d7bd3369a
 size 102078202

checkpoint-56000/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6504f403ec0124563e90a8d77beb4d07baefb7092c70b6d168b7c081e93523f
+oid sha256:24576edc0ba16167d3315213d64436fa9419567478a4365800ac736ef77967d8
 size 14308

checkpoint-56000/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f1181b919b10206263db791048c88fc007cb1b4a9fe395db90858e73bef9c62
+oid sha256:65a845dd3870415240fe85e91ead23713a7c358d33828635effb41406f964c7a
 size 1000

checkpoint-56000/trainer_state.json

@@ -1,7 +1,7 @@
 {
-  "best_metric": 4.237761974334717,
-  "best_model_checkpoint": "/home/p318482/babyLM_controlled/models_trained/en_clm/wikipedia_30/checkpoint-56000",
-  "epoch": 29.45043386799895,
+  "best_metric": 3.5649592876434326,
+  "best_model_checkpoint": "/home/p318482/babyLM_controlled/models_trained/fr_clm/wikipedia_30/checkpoint-36000",
+  "epoch": 60.11808910359635,
   "eval_steps": 2000,
   "global_step": 56000,
   "is_hyper_param_search": false,
@@ -9,332 +9,332 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 1.0518012095713911,
-      "eval_loss": 7.518920421600342,
-      "eval_runtime": 2.2399,
-      "eval_samples_per_second": 1422.808,
-      "eval_steps_per_second": 89.288,
+      "epoch": 2.1470746108427265,
+      "eval_loss": 7.138043403625488,
+      "eval_runtime": 0.8415,
+      "eval_samples_per_second": 1273.93,
+      "eval_steps_per_second": 79.621,
       "step": 2000
     },
     {
-      "epoch": 2.1036024191427822,
-      "grad_norm": 1.2947496175765991,
+      "epoch": 4.294149221685453,
+      "grad_norm": 1.455039381980896,
       "learning_rate": 1e-05,
-      "loss": 7.6141,
+      "loss": 7.213,
       "step": 4000
     },
     {
-      "epoch": 2.1036024191427822,
-      "eval_loss": 6.568221092224121,
-      "eval_runtime": 2.1236,
-      "eval_samples_per_second": 1500.783,
-      "eval_steps_per_second": 94.182,
+      "epoch": 4.294149221685453,
+      "eval_loss": 5.8541717529296875,
+      "eval_runtime": 0.7753,
+      "eval_samples_per_second": 1382.676,
+      "eval_steps_per_second": 86.417,
       "step": 4000
     },
     {
-      "epoch": 3.155403628714173,
-      "eval_loss": 6.16925573348999,
-      "eval_runtime": 2.1624,
-      "eval_samples_per_second": 1473.825,
-      "eval_steps_per_second": 92.49,
+      "epoch": 6.4412238325281805,
+      "eval_loss": 5.403579235076904,
+      "eval_runtime": 0.7542,
+      "eval_samples_per_second": 1421.437,
+      "eval_steps_per_second": 88.84,
       "step": 6000
     },
     {
-      "epoch": 4.2072048382855645,
-      "grad_norm": 2.330080270767212,
-      "learning_rate": 2e-05,
-      "loss": 6.2396,
+      "epoch": 8.588298443370906,
+      "grad_norm": 3.117489814758301,
+      "learning_rate": 1.9997500000000003e-05,
+      "loss": 5.4304,
       "step": 8000
     },
     {
-      "epoch": 4.2072048382855645,
-      "eval_loss": 5.901575088500977,
-      "eval_runtime": 2.1326,
-      "eval_samples_per_second": 1494.388,
-      "eval_steps_per_second": 93.78,
+      "epoch": 8.588298443370906,
+      "eval_loss": 5.049880504608154,
+      "eval_runtime": 0.7597,
+      "eval_samples_per_second": 1411.127,
+      "eval_steps_per_second": 88.195,
       "step": 8000
     },
     {
-      "epoch": 5.259006047856955,
-      "eval_loss": 5.67369270324707,
-      "eval_runtime": 2.1364,
-      "eval_samples_per_second": 1491.788,
-      "eval_steps_per_second": 93.617,
+      "epoch": 10.735373054213634,
+      "eval_loss": 4.760603427886963,
+      "eval_runtime": 0.7624,
+      "eval_samples_per_second": 1406.053,
+      "eval_steps_per_second": 87.878,
       "step": 10000
     },
     {
-      "epoch": 6.310807257428346,
-      "grad_norm": 2.565220594406128,
-      "learning_rate": 2.99925e-05,
-      "loss": 5.7217,
+      "epoch": 12.882447665056361,
+      "grad_norm": 2.9511141777038574,
+      "learning_rate": 2.9995e-05,
+      "loss": 4.771,
       "step": 12000
     },
     {
-      "epoch": 6.310807257428346,
-      "eval_loss": 5.460005760192871,
-      "eval_runtime": 2.1835,
-      "eval_samples_per_second": 1459.563,
-      "eval_steps_per_second": 91.595,
+      "epoch": 12.882447665056361,
+      "eval_loss": 4.517208576202393,
+      "eval_runtime": 0.764,
+      "eval_samples_per_second": 1403.097,
+      "eval_steps_per_second": 87.694,
       "step": 12000
     },
     {
-      "epoch": 7.362608466999737,
-      "eval_loss": 5.272489547729492,
-      "eval_runtime": 2.1833,
-      "eval_samples_per_second": 1459.691,
-      "eval_steps_per_second": 91.603,
+      "epoch": 15.029522275899087,
+      "eval_loss": 4.3206024169921875,
+      "eval_runtime": 0.7595,
+      "eval_samples_per_second": 1411.435,
+      "eval_steps_per_second": 88.215,
       "step": 14000
     },
     {
-      "epoch": 8.414409676571129,
-      "grad_norm": 2.7334115505218506,
-      "learning_rate": 3.99875e-05,
-      "loss": 5.3064,
+      "epoch": 17.176596886741816,
+      "grad_norm": 2.955515146255493,
+      "learning_rate": 3.999e-05,
+      "loss": 4.2888,
       "step": 16000
     },
     {
-      "epoch": 8.414409676571129,
-      "eval_loss": 5.1074652671813965,
-      "eval_runtime": 2.2402,
-      "eval_samples_per_second": 1422.652,
-      "eval_steps_per_second": 89.278,
+      "epoch": 17.176596886741816,
+      "eval_loss": 4.152973175048828,
+      "eval_runtime": 0.8247,
+      "eval_samples_per_second": 1299.91,
+      "eval_steps_per_second": 81.244,
       "step": 16000
     },
     {
-      "epoch": 9.46621088614252,
-      "eval_loss": 4.9716596603393555,
-      "eval_runtime": 2.1736,
-      "eval_samples_per_second": 1466.258,
-      "eval_steps_per_second": 92.015,
+      "epoch": 19.32367149758454,
+      "eval_loss": 4.015476703643799,
+      "eval_runtime": 0.7501,
+      "eval_samples_per_second": 1429.237,
+      "eval_steps_per_second": 89.327,
       "step": 18000
     },
     {
-      "epoch": 10.51801209571391,
-      "grad_norm": 2.6722376346588135,
-      "learning_rate": 4.9985e-05,
-      "loss": 4.9744,
+      "epoch": 21.470746108427267,
+      "grad_norm": 3.0531320571899414,
+      "learning_rate": 4.99875e-05,
+      "loss": 3.9141,
       "step": 20000
     },
     {
-      "epoch": 10.51801209571391,
-      "eval_loss": 4.864916801452637,
-      "eval_runtime": 2.2031,
-      "eval_samples_per_second": 1446.585,
-      "eval_steps_per_second": 90.78,
+      "epoch": 21.470746108427267,
+      "eval_loss": 3.896575927734375,
+      "eval_runtime": 0.768,
+      "eval_samples_per_second": 1395.843,
+      "eval_steps_per_second": 87.24,
       "step": 20000
     },
     {
-      "epoch": 11.569813305285301,
-      "eval_loss": 4.763542652130127,
-      "eval_runtime": 2.4382,
-      "eval_samples_per_second": 1307.131,
-      "eval_steps_per_second": 82.029,
+      "epoch": 23.617820719269993,
+      "eval_loss": 3.8046905994415283,
+      "eval_runtime": 0.7601,
+      "eval_samples_per_second": 1410.272,
+      "eval_steps_per_second": 88.142,
       "step": 22000
     },
     {
-      "epoch": 12.621614514856692,
-      "grad_norm": 2.4539408683776855,
-      "learning_rate": 5.9985e-05,
-      "loss": 4.7273,
+      "epoch": 25.764895330112722,
+      "grad_norm": 3.089989185333252,
+      "learning_rate": 5.998250000000001e-05,
+      "loss": 3.6154,
       "step": 24000
     },
     {
-      "epoch": 12.621614514856692,
-      "eval_loss": 4.683297634124756,
-      "eval_runtime": 2.452,
-      "eval_samples_per_second": 1299.733,
-      "eval_steps_per_second": 81.565,
+      "epoch": 25.764895330112722,
+      "eval_loss": 3.7359092235565186,
+      "eval_runtime": 0.7778,
+      "eval_samples_per_second": 1378.238,
+      "eval_steps_per_second": 86.14,
       "step": 24000
     },
     {
-      "epoch": 13.673415724428082,
-      "eval_loss": 4.621574401855469,
-      "eval_runtime": 2.1825,
-      "eval_samples_per_second": 1460.226,
-      "eval_steps_per_second": 91.636,
+      "epoch": 27.911969940955448,
+      "eval_loss": 3.6784231662750244,
+      "eval_runtime": 0.7541,
+      "eval_samples_per_second": 1421.491,
+      "eval_steps_per_second": 88.843,
       "step": 26000
     },
     {
-      "epoch": 14.725216933999475,
-      "grad_norm": 2.4913346767425537,
-      "learning_rate": 6.998e-05,
-      "loss": 4.5397,
+      "epoch": 30.059044551798173,
+      "grad_norm": 3.3495922088623047,
+      "learning_rate": 6.997500000000001e-05,
+      "loss": 3.3661,
       "step": 28000
     },
     {
-      "epoch": 14.725216933999475,
-      "eval_loss": 4.562623977661133,
-      "eval_runtime": 2.201,
-      "eval_samples_per_second": 1447.965,
-      "eval_steps_per_second": 90.867,
+      "epoch": 30.059044551798173,
+      "eval_loss": 3.6360318660736084,
+      "eval_runtime": 0.7603,
+      "eval_samples_per_second": 1409.913,
+      "eval_steps_per_second": 88.12,
       "step": 28000
     },
     {
-      "epoch": 15.777018143570865,
-      "eval_loss": 4.506168842315674,
-      "eval_runtime": 2.2491,
-      "eval_samples_per_second": 1416.988,
-      "eval_steps_per_second": 88.923,
+      "epoch": 32.2061191626409,
+      "eval_loss": 3.6019041538238525,
+      "eval_runtime": 0.7638,
+      "eval_samples_per_second": 1403.547,
+      "eval_steps_per_second": 87.722,
       "step": 30000
     },
     {
-      "epoch": 16.828819353142258,
-      "grad_norm": 2.2661523818969727,
-      "learning_rate": 7.99775e-05,
-      "loss": 4.3839,
+      "epoch": 34.35319377348363,
+      "grad_norm": 3.1256051063537598,
+      "learning_rate": 7.997250000000001e-05,
+      "loss": 3.1473,
       "step": 32000
     },
     {
-      "epoch": 16.828819353142258,
-      "eval_loss": 4.464449882507324,
-      "eval_runtime": 2.253,
-      "eval_samples_per_second": 1414.55,
-      "eval_steps_per_second": 88.77,
+      "epoch": 34.35319377348363,
+      "eval_loss": 3.5816054344177246,
+      "eval_runtime": 0.7628,
+      "eval_samples_per_second": 1405.391,
+      "eval_steps_per_second": 87.837,
       "step": 32000
     },
     {
-      "epoch": 17.88062056271365,
-      "eval_loss": 4.426318168640137,
-      "eval_runtime": 2.2688,
-      "eval_samples_per_second": 1404.709,
-      "eval_steps_per_second": 88.152,
+      "epoch": 36.500268384326354,
+      "eval_loss": 3.5698702335357666,
+      "eval_runtime": 0.7617,
+      "eval_samples_per_second": 1407.381,
+      "eval_steps_per_second": 87.961,
       "step": 34000
     },
     {
-      "epoch": 18.93242177228504,
-      "grad_norm": 2.226025104522705,
-      "learning_rate": 8.9975e-05,
-      "loss": 4.2529,
+      "epoch": 38.64734299516908,
+      "grad_norm": 2.921318769454956,
+      "learning_rate": 8.997000000000001e-05,
+      "loss": 2.9533,
       "step": 36000
     },
     {
-      "epoch": 18.93242177228504,
-      "eval_loss": 4.3944478034973145,
-      "eval_runtime": 2.294,
-      "eval_samples_per_second": 1389.273,
-      "eval_steps_per_second": 87.184,
+      "epoch": 38.64734299516908,
+      "eval_loss": 3.5649592876434326,
+      "eval_runtime": 0.7607,
+      "eval_samples_per_second": 1409.149,
+      "eval_steps_per_second": 88.072,
       "step": 36000
     },
     {
-      "epoch": 19.98422298185643,
-      "eval_loss": 4.365135669708252,
-      "eval_runtime": 2.2176,
-      "eval_samples_per_second": 1437.165,
-      "eval_steps_per_second": 90.189,
+      "epoch": 40.794417606011805,
+      "eval_loss": 3.5666539669036865,
+      "eval_runtime": 0.7558,
+      "eval_samples_per_second": 1418.297,
+      "eval_steps_per_second": 88.644,
       "step": 38000
     },
     {
-      "epoch": 21.03602419142782,
-      "grad_norm": 2.1669206619262695,
-      "learning_rate": 9.997e-05,
-      "loss": 4.1409,
+      "epoch": 42.941492216854535,
+      "grad_norm": 3.202237129211426,
+      "learning_rate": 9.996500000000001e-05,
+      "loss": 2.777,
       "step": 40000
     },
     {
-      "epoch": 21.03602419142782,
-      "eval_loss": 4.351284503936768,
-      "eval_runtime": 2.1934,
-      "eval_samples_per_second": 1453.028,
-      "eval_steps_per_second": 91.185,
+      "epoch": 42.941492216854535,
+      "eval_loss": 3.5746841430664062,
+      "eval_runtime": 0.7722,
+      "eval_samples_per_second": 1388.315,
+      "eval_steps_per_second": 86.77,
       "step": 40000
     },
     {
-      "epoch": 22.08782540099921,
-      "eval_loss": 4.321634292602539,
-      "eval_runtime": 2.2273,
-      "eval_samples_per_second": 1430.895,
-      "eval_steps_per_second": 89.796,
+      "epoch": 45.088566827697264,
+      "eval_loss": 3.5878281593322754,
+      "eval_runtime": 0.7639,
+      "eval_samples_per_second": 1403.298,
+      "eval_steps_per_second": 87.706,
       "step": 42000
     },
     {
-      "epoch": 23.139626610570602,
-      "grad_norm": 2.0360896587371826,
-      "learning_rate": 9.335333333333333e-05,
-      "loss": 4.0334,
+      "epoch": 47.235641438539986,
+      "grad_norm": 2.9902379512786865,
+      "learning_rate": 9.336e-05,
+      "loss": 2.6015,
       "step": 44000
     },
     {
-      "epoch": 23.139626610570602,
-      "eval_loss": 4.300807476043701,
-      "eval_runtime": 2.3226,
-      "eval_samples_per_second": 1372.142,
-      "eval_steps_per_second": 86.109,
+      "epoch": 47.235641438539986,
+      "eval_loss": 3.6107263565063477,
+      "eval_runtime": 0.7601,
+      "eval_samples_per_second": 1410.262,
+      "eval_steps_per_second": 88.141,
       "step": 44000
     },
     {
-      "epoch": 24.191427820141993,
-      "eval_loss": 4.279116153717041,
-      "eval_runtime": 2.2177,
-      "eval_samples_per_second": 1437.082,
-      "eval_steps_per_second": 90.184,
+      "epoch": 49.382716049382715,
+      "eval_loss": 3.6260793209075928,
+      "eval_runtime": 0.7605,
+      "eval_samples_per_second": 1409.681,
+      "eval_steps_per_second": 88.105,
       "step": 46000
     },
     {
-      "epoch": 25.243229029713383,
-      "grad_norm": 2.137535572052002,
-      "learning_rate": 8.668833333333334e-05,
-      "loss": 3.9378,
+      "epoch": 51.529790660225444,
+      "grad_norm": 3.1626901626586914,
+      "learning_rate": 8.669833333333334e-05,
+      "loss": 2.4429,
       "step": 48000
     },
     {
-      "epoch": 25.243229029713383,
-      "eval_loss": 4.2718939781188965,
-      "eval_runtime": 2.2019,
-      "eval_samples_per_second": 1447.39,
-      "eval_steps_per_second": 90.831,
+      "epoch": 51.529790660225444,
+      "eval_loss": 3.6414248943328857,
+      "eval_runtime": 0.7596,
+      "eval_samples_per_second": 1411.254,
+      "eval_steps_per_second": 88.203,
       "step": 48000
     },
     {
-      "epoch": 26.295030239284774,
-      "eval_loss": 4.26237678527832,
-      "eval_runtime": 2.2417,
-      "eval_samples_per_second": 1421.698,
-      "eval_steps_per_second": 89.219,
+      "epoch": 53.676865271068166,
+      "eval_loss": 3.6636786460876465,
+      "eval_runtime": 0.7637,
+      "eval_samples_per_second": 1403.705,
+      "eval_steps_per_second": 87.732,
       "step": 50000
     },
     {
-      "epoch": 27.346831448856165,
-      "grad_norm": 2.1131036281585693,
-      "learning_rate": 8.002333333333334e-05,
-      "loss": 3.8595,
+      "epoch": 55.823939881910896,
+      "grad_norm": 3.162830114364624,
+      "learning_rate": 8.003333333333333e-05,
+      "loss": 2.3125,
       "step": 52000
     },
     {
-      "epoch": 27.346831448856165,
-      "eval_loss": 4.253220558166504,
-      "eval_runtime": 2.2045,
-      "eval_samples_per_second": 1445.692,
-      "eval_steps_per_second": 90.724,
+      "epoch": 55.823939881910896,
+      "eval_loss": 3.6777544021606445,
+      "eval_runtime": 0.7597,
+      "eval_samples_per_second": 1411.137,
+      "eval_steps_per_second": 88.196,
       "step": 52000
     },
     {
-      "epoch": 28.39863265842756,
-      "eval_loss": 4.243188381195068,
-      "eval_runtime": 2.1654,
-      "eval_samples_per_second": 1471.751,
-      "eval_steps_per_second": 92.36,
+      "epoch": 57.971014492753625,
+      "eval_loss": 3.703284502029419,
+      "eval_runtime": 0.757,
+      "eval_samples_per_second": 1416.089,
+      "eval_steps_per_second": 88.506,
       "step": 54000
     },
     {
-      "epoch": 29.45043386799895,
-      "grad_norm": 2.1626858711242676,
-      "learning_rate": 7.335833333333335e-05,
-      "loss": 3.7927,
+      "epoch": 60.11808910359635,
+      "grad_norm": 3.4666566848754883,
+      "learning_rate": 7.337000000000001e-05,
+      "loss": 2.1989,
       "step": 56000
     },
     {
-      "epoch": 29.45043386799895,
-      "eval_loss": 4.237761974334717,
-      "eval_runtime": 2.3384,
-      "eval_samples_per_second": 1362.919,
-      "eval_steps_per_second": 85.53,
+      "epoch": 60.11808910359635,
+      "eval_loss": 3.7409818172454834,
+      "eval_runtime": 0.7476,
+      "eval_samples_per_second": 1433.993,
+      "eval_steps_per_second": 89.625,
       "step": 56000
     }
   ],
   "logging_steps": 4000,
   "max_steps": 100000,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 53,
+  "num_train_epochs": 108,
   "save_steps": 4000,
   "stateful_callbacks": {
     "TrainerControl": {
@@ -348,7 +348,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.4476189099416576e+16,
+  "total_flos": 1.447583199264768e+16,
   "train_batch_size": 16,
   "trial_name": null,
   "trial_params": null

checkpoint-56000/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab68434ec7156d3a63e4783f824871a48d62c7c2fdcc831fdd5471c5f4aec7fe
+oid sha256:26102d0ac750ce531ebda9d7577ccea6c8ecbc93b04bdff23226e376dd5609a2
 size 5368