Training in progress, step 1000, checkpoint

last-checkpoint/config.json

@@ -8,184 +8,184 @@
   "hidden_dropout_prob": 0.1,
   "hidden_size": 768,
   "id2label": {
-    "0": "Global Financing Facility for Women, Children and Adolescents",
-    "1": "Competitiveness",
-    "2": "Sustainable Communities",
-    "3": "Education",
-    "4": "Natural Capital",
-    "5": "Regional Integration",
-    "6": "Jobs and Development",
-    "7": "Innovation and Entrepreneurship",
-    "8": "Inclusive Cities",
-    "9": "Jobs & Development",
-    "10": "Transport",
-    "11": "Sustainable Infrastructure Finance",
-    "12": "Pandemic Preparedness and COVID-19",
-    "13": "Taxes and Government Revenue",
-    "14": "Labor Markets",
-    "15": "Measuring Poverty",
-    "16": "Food Security Update",
-    "17": "Disability Inclusion",
-    "18": "Agriculture and Food",
-    "19": "Trade Facilitation and Logistics",
-    "20": "Social Protection",
-    "21": "One Health",
-    "22": "Biodiversity",
-    "23": "Social Inclusion",
-    "24": "Credit Infrastructure",
-    "25": "Water Supply",
-    "26": "Early Childhood Development",
-    "27": "Food System Jobs",
-    "28": "Migration",
-    "29": "Indigenous Peoples",
-    "30": "Universal Health Coverage",
-    "31": "Financial Sector",
-    "32": "Procurement for Development",
-    "33": "Inequality and Shared Prosperity",
-    "34": "COVID-19 Hub",
-    "35": "Poverty",
-    "36": "Financial Stability",
-    "37": "Digital Development",
-    "38": "Long-Term Finance",
-    "39": "Gas Flaring Reduction",
-    "40": "Mining Investment and Governance Review",
-    "41": "Small and Medium Enterprises Finance",
-    "42": "Infrastructure",
-    "43": "Health",
-    "44": "Sexual Orientation and Gender Identity",
-    "45": "Nutrition",
-    "46": "Financial Inclusion",
-    "47": "Fragility, Conflict, and Violence",
-    "48": "Debt Relief",
-    "49": "Disaster Risk Management",
-    "50": "Water in Agriculture",
-    "51": "Livestock and Sustainability",
-    "52": "Global Value Chains",
-    "53": "Competition Policy",
-    "54": "Pollution",
-    "55": "Urban Development",
-    "56": "Gender",
-    "57": "Safety Nets and Cash Transfers",
-    "58": "Forests and Landscapes",
-    "59": "Water Resources Management",
-    "60": "Extractive Industries",
-    "61": "Social Sustainability and Inclusion",
-    "62": "Energy",
-    "63": "Girls' Education",
-    "64": "Environment",
-    "65": "Marine Plastic Pollution",
-    "66": "Education and Technology",
-    "67": "Financial Integrity",
-    "68": "Oceans, Fisheries, and Coastal Economies",
-    "69": "Sanitation",
-    "70": "Land",
-    "71": "Higher Education",
-    "72": "Teachers",
-    "73": "Investment Climate",
-    "74": "Debt",
-    "75": "Climate Change",
-    "76": "Trade",
-    "77": "Skills Development",
-    "78": "Agribusiness and Value Chains",
-    "79": "Climate-Smart Agriculture",
-    "80": "Pensions",
-    "81": "Infectious diseases and Vaccines",
-    "82": "Payment Systems",
-    "83": "Community-Driven Development",
-    "84": "Water",
-    "85": "Governance",
-    "86": "Macroeconomics"
+    "0": "Indigenous Peoples",
+    "1": "Investment Climate",
+    "2": "Marine Plastic Pollution",
+    "3": "Poverty",
+    "4": "Water",
+    "5": "Measuring Poverty",
+    "6": "Energy",
+    "7": "Sustainable Communities",
+    "8": "Disability Inclusion",
+    "9": "Macroeconomics",
+    "10": "Early Childhood Development",
+    "11": "Mining Investment and Governance Review",
+    "12": "Infrastructure",
+    "13": "Community-Driven Development",
+    "14": "Land",
+    "15": "Innovation and Entrepreneurship",
+    "16": "Competitiveness",
+    "17": "Biodiversity",
+    "18": "Debt",
+    "19": "Higher Education",
+    "20": "Livestock and Sustainability",
+    "21": "Universal Health Coverage",
+    "22": "Nutrition",
+    "23": "Education and Technology",
+    "24": "Small and Medium Enterprises Finance",
+    "25": "Climate-Smart Agriculture",
+    "26": "Food Security Update",
+    "27": "Inclusive Cities",
+    "28": "Jobs & Development",
+    "29": "Migration",
+    "30": "Infectious diseases and Vaccines",
+    "31": "Social Protection",
+    "32": "Skills Development",
+    "33": "Governance",
+    "34": "Pollution",
+    "35": "Inequality and Shared Prosperity",
+    "36": "Oceans, Fisheries, and Coastal Economies",
+    "37": "Global Value Chains",
+    "38": "Pensions",
+    "39": "Food System Jobs",
+    "40": "Debt Relief",
+    "41": "Sexual Orientation and Gender Identity",
+    "42": "Education",
+    "43": "Regional Integration",
+    "44": "Jobs and Development",
+    "45": "Competition Policy",
+    "46": "Fragility, Conflict, and Violence",
+    "47": "Urban Development",
+    "48": "Pandemic Preparedness and COVID-19",
+    "49": "Safety Nets and Cash Transfers",
+    "50": "Financial Sector",
+    "51": "Climate Change",
+    "52": "Transport",
+    "53": "COVID-19 Hub",
+    "54": "Financial Stability",
+    "55": "Gender",
+    "56": "Natural Capital",
+    "57": "Water in Agriculture",
+    "58": "Taxes and Government Revenue",
+    "59": "Labor Markets",
+    "60": "Trade Facilitation and Logistics",
+    "61": "Long-Term Finance",
+    "62": "Teachers",
+    "63": "Sanitation",
+    "64": "Forests and Landscapes",
+    "65": "Financial Integrity",
+    "66": "Payment Systems",
+    "67": "Social Inclusion",
+    "68": "Health",
+    "69": "Girls' Education",
+    "70": "Credit Infrastructure",
+    "71": "Social Sustainability and Inclusion",
+    "72": "Trade",
+    "73": "Financial Inclusion",
+    "74": "Global Financing Facility for Women, Children and Adolescents",
+    "75": "Sustainable Infrastructure Finance",
+    "76": "Extractive Industries",
+    "77": "Procurement for Development",
+    "78": "Agriculture and Food",
+    "79": "Water Supply",
+    "80": "Disaster Risk Management",
+    "81": "One Health",
+    "82": "Environment",
+    "83": "Digital Development",
+    "84": "Agribusiness and Value Chains",
+    "85": "Gas Flaring Reduction",
+    "86": "Water Resources Management"
   },
   "initializer_range": 0.02,
   "intermediate_size": 3072,
   "label2id": {
-    "Agribusiness and Value Chains": 78,
-    "Agriculture and Food": 18,
-    "Biodiversity": 22,
-    "COVID-19 Hub": 34,
-    "Climate Change": 75,
-    "Climate-Smart Agriculture": 79,
-    "Community-Driven Development": 83,
-    "Competition Policy": 53,
-    "Competitiveness": 1,
-    "Credit Infrastructure": 24,
-    "Debt": 74,
-    "Debt Relief": 48,
-    "Digital Development": 37,
-    "Disability Inclusion": 17,
-    "Disaster Risk Management": 49,
-    "Early Childhood Development": 26,
-    "Education": 3,
-    "Education and Technology": 66,
-    "Energy": 62,
-    "Environment": 64,
-    "Extractive Industries": 60,
-    "Financial Inclusion": 46,
-    "Financial Integrity": 67,
-    "Financial Sector": 31,
-    "Financial Stability": 36,
-    "Food Security Update": 16,
-    "Food System Jobs": 27,
-    "Forests and Landscapes": 58,
-    "Fragility, Conflict, and Violence": 47,
-    "Gas Flaring Reduction": 39,
-    "Gender": 56,
-    "Girls' Education": 63,
-    "Global Financing Facility for Women, Children and Adolescents": 0,
-    "Global Value Chains": 52,
-    "Governance": 85,
-    "Health": 43,
-    "Higher Education": 71,
-    "Inclusive Cities": 8,
-    "Indigenous Peoples": 29,
-    "Inequality and Shared Prosperity": 33,
-    "Infectious diseases and Vaccines": 81,
-    "Infrastructure": 42,
-    "Innovation and Entrepreneurship": 7,
-    "Investment Climate": 73,
-    "Jobs & Development": 9,
-    "Jobs and Development": 6,
-    "Labor Markets": 14,
-    "Land": 70,
-    "Livestock and Sustainability": 51,
-    "Long-Term Finance": 38,
-    "Macroeconomics": 86,
-    "Marine Plastic Pollution": 65,
-    "Measuring Poverty": 15,
-    "Migration": 28,
-    "Mining Investment and Governance Review": 40,
-    "Natural Capital": 4,
-    "Nutrition": 45,
-    "Oceans, Fisheries, and Coastal Economies": 68,
-    "One Health": 21,
-    "Pandemic Preparedness and COVID-19": 12,
-    "Payment Systems": 82,
-    "Pensions": 80,
-    "Pollution": 54,
-    "Poverty": 35,
-    "Procurement for Development": 32,
-    "Regional Integration": 5,
-    "Safety Nets and Cash Transfers": 57,
-    "Sanitation": 69,
-    "Sexual Orientation and Gender Identity": 44,
-    "Skills Development": 77,
-    "Small and Medium Enterprises Finance": 41,
-    "Social Inclusion": 23,
-    "Social Protection": 20,
-    "Social Sustainability and Inclusion": 61,
-    "Sustainable Communities": 2,
-    "Sustainable Infrastructure Finance": 11,
-    "Taxes and Government Revenue": 13,
-    "Teachers": 72,
-    "Trade": 76,
-    "Trade Facilitation and Logistics": 19,
-    "Transport": 10,
-    "Universal Health Coverage": 30,
-    "Urban Development": 55,
-    "Water": 84,
-    "Water Resources Management": 59,
-    "Water Supply": 25,
-    "Water in Agriculture": 50
+    "Agribusiness and Value Chains": 84,
+    "Agriculture and Food": 78,
+    "Biodiversity": 17,
+    "COVID-19 Hub": 53,
+    "Climate Change": 51,
+    "Climate-Smart Agriculture": 25,
+    "Community-Driven Development": 13,
+    "Competition Policy": 45,
+    "Competitiveness": 16,
+    "Credit Infrastructure": 70,
+    "Debt": 18,
+    "Debt Relief": 40,
+    "Digital Development": 83,
+    "Disability Inclusion": 8,
+    "Disaster Risk Management": 80,
+    "Early Childhood Development": 10,
+    "Education": 42,
+    "Education and Technology": 23,
+    "Energy": 6,
+    "Environment": 82,
+    "Extractive Industries": 76,
+    "Financial Inclusion": 73,
+    "Financial Integrity": 65,
+    "Financial Sector": 50,
+    "Financial Stability": 54,
+    "Food Security Update": 26,
+    "Food System Jobs": 39,
+    "Forests and Landscapes": 64,
+    "Fragility, Conflict, and Violence": 46,
+    "Gas Flaring Reduction": 85,
+    "Gender": 55,
+    "Girls' Education": 69,
+    "Global Financing Facility for Women, Children and Adolescents": 74,
+    "Global Value Chains": 37,
+    "Governance": 33,
+    "Health": 68,
+    "Higher Education": 19,
+    "Inclusive Cities": 27,
+    "Indigenous Peoples": 0,
+    "Inequality and Shared Prosperity": 35,
+    "Infectious diseases and Vaccines": 30,
+    "Infrastructure": 12,
+    "Innovation and Entrepreneurship": 15,
+    "Investment Climate": 1,
+    "Jobs & Development": 28,
+    "Jobs and Development": 44,
+    "Labor Markets": 59,
+    "Land": 14,
+    "Livestock and Sustainability": 20,
+    "Long-Term Finance": 61,
+    "Macroeconomics": 9,
+    "Marine Plastic Pollution": 2,
+    "Measuring Poverty": 5,
+    "Migration": 29,
+    "Mining Investment and Governance Review": 11,
+    "Natural Capital": 56,
+    "Nutrition": 22,
+    "Oceans, Fisheries, and Coastal Economies": 36,
+    "One Health": 81,
+    "Pandemic Preparedness and COVID-19": 48,
+    "Payment Systems": 66,
+    "Pensions": 38,
+    "Pollution": 34,
+    "Poverty": 3,
+    "Procurement for Development": 77,
+    "Regional Integration": 43,
+    "Safety Nets and Cash Transfers": 49,
+    "Sanitation": 63,
+    "Sexual Orientation and Gender Identity": 41,
+    "Skills Development": 32,
+    "Small and Medium Enterprises Finance": 24,
+    "Social Inclusion": 67,
+    "Social Protection": 31,
+    "Social Sustainability and Inclusion": 71,
+    "Sustainable Communities": 7,
+    "Sustainable Infrastructure Finance": 75,
+    "Taxes and Government Revenue": 58,
+    "Teachers": 62,
+    "Trade": 72,
+    "Trade Facilitation and Logistics": 60,
+    "Transport": 52,
+    "Universal Health Coverage": 21,
+    "Urban Development": 47,
+    "Water": 4,
+    "Water Resources Management": 86,
+    "Water Supply": 79,
+    "Water in Agriculture": 57
   },
   "layer_norm_eps": 1e-07,
   "max_position_embeddings": 512,

last-checkpoint/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cd8e2da575487f0f514d3ac641b54188b83b65293d17518c847ecd0b5e591ca1
+oid sha256:0261c797ffae126153cbbb1467b83780ff91dc1912396077fc38e284df689975
 size 567860028

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:86fe37d6452dbda6b773629c24131b2d3eda42a3d4aa1cf384885d4fe03ef892
+oid sha256:a1d631affb6dccd4f3fd4cfc3390be46f868b7efccefa2043b4f74f85561c51c
 size 1135783354

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2ef6b86d20d6ecc4df3a0c586eaefe72e249b74f60cbccfafd813b5495f54c4f
+oid sha256:34649568145f465443beb487f2295ab3dad9e5f49f758646dd823029413e18fe
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3c3f1ac50882defeade258e77931f3038c02d6276451a81f8ef210f495feca4
+oid sha256:8e136870ce715682b288c29fb4d05a0aec61f1a74b2cc393d7fb9e66e4965261
 size 1064

last-checkpoint/trainer_state.json

@@ -1,298 +1,38 @@
 {
-  "best_metric": 0.03708931431174278,
-  "best_model_checkpoint": "doc-topic-model_eval-00_train-03/checkpoint-11000",
-  "epoch": 5.424063116370808,
+  "best_metric": 0.08951831609010696,
+  "best_model_checkpoint": "doc-topic-model_eval-00_train-03/checkpoint-1000",
+  "epoch": 0.4930966469428008,
   "eval_steps": 1000,
-  "global_step": 11000,
+  "global_step": 1000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.2465483234714004,
-      "grad_norm": 0.3815906345844269,
+      "grad_norm": 0.3881610631942749,
       "learning_rate": 1.9950690335305722e-05,
-      "loss": 0.1667,
+      "loss": 0.1669,
       "step": 500
     },
     {
       "epoch": 0.4930966469428008,
-      "grad_norm": 0.36234650015830994,
+      "grad_norm": 0.3564796447753906,
       "learning_rate": 1.9901380670611442e-05,
-      "loss": 0.0923,
+      "loss": 0.0935,
       "step": 1000
     },
     {
       "epoch": 0.4930966469428008,
       "eval_accuracy": 0.9814660487265615,
       "eval_f1": 0.0,
-      "eval_loss": 0.08645907044410706,
+      "eval_loss": 0.08951831609010696,
       "eval_precision": 0.0,
       "eval_recall": 0.0,
-      "eval_runtime": 11.8049,
-      "eval_samples_per_second": 687.004,
-      "eval_steps_per_second": 2.711,
+      "eval_runtime": 12.2339,
+      "eval_samples_per_second": 662.913,
+      "eval_steps_per_second": 2.616,
       "step": 1000
-    },
-    {
-      "epoch": 0.7396449704142012,
-      "grad_norm": 0.39963558316230774,
-      "learning_rate": 1.9852071005917162e-05,
-      "loss": 0.084,
-      "step": 1500
-    },
-    {
-      "epoch": 0.9861932938856016,
-      "grad_norm": 0.37207868695259094,
-      "learning_rate": 1.980276134122288e-05,
-      "loss": 0.0734,
-      "step": 2000
-    },
-    {
-      "epoch": 0.9861932938856016,
-      "eval_accuracy": 0.9814660487265615,
-      "eval_f1": 0.0,
-      "eval_loss": 0.0668005496263504,
-      "eval_precision": 0.0,
-      "eval_recall": 0.0,
-      "eval_runtime": 11.8932,
-      "eval_samples_per_second": 681.903,
-      "eval_steps_per_second": 2.691,
-      "step": 2000
-    },
-    {
-      "epoch": 1.232741617357002,
-      "grad_norm": 0.39202699065208435,
-      "learning_rate": 1.9753451676528602e-05,
-      "loss": 0.0653,
-      "step": 2500
-    },
-    {
-      "epoch": 1.4792899408284024,
-      "grad_norm": 0.3334597647190094,
-      "learning_rate": 1.9704142011834322e-05,
-      "loss": 0.0606,
-      "step": 3000
-    },
-    {
-      "epoch": 1.4792899408284024,
-      "eval_accuracy": 0.9823617784202843,
-      "eval_f1": 0.12648276830209867,
-      "eval_loss": 0.05524122714996338,
-      "eval_precision": 0.7700854700854701,
-      "eval_recall": 0.06889959470826643,
-      "eval_runtime": 14.1698,
-      "eval_samples_per_second": 572.346,
-      "eval_steps_per_second": 2.258,
-      "step": 3000
-    },
-    {
-      "epoch": 1.725838264299803,
-      "grad_norm": 0.3517291843891144,
-      "learning_rate": 1.965483234714004e-05,
-      "loss": 0.055,
-      "step": 3500
-    },
-    {
-      "epoch": 1.972386587771203,
-      "grad_norm": 0.43956857919692993,
-      "learning_rate": 1.9605522682445763e-05,
-      "loss": 0.0532,
-      "step": 4000
-    },
-    {
-      "epoch": 1.972386587771203,
-      "eval_accuracy": 0.9840511926527489,
-      "eval_f1": 0.2943500344892456,
-      "eval_loss": 0.0491451695561409,
-      "eval_precision": 0.8177700348432055,
-      "eval_recall": 0.1794754148505009,
-      "eval_runtime": 14.1385,
-      "eval_samples_per_second": 573.612,
-      "eval_steps_per_second": 2.263,
-      "step": 4000
-    },
-    {
-      "epoch": 2.2189349112426036,
-      "grad_norm": 0.33584991097450256,
-      "learning_rate": 1.9556213017751483e-05,
-      "loss": 0.0483,
-      "step": 4500
-    },
-    {
-      "epoch": 2.465483234714004,
-      "grad_norm": 0.35681256651878357,
-      "learning_rate": 1.95069033530572e-05,
-      "loss": 0.0466,
-      "step": 5000
-    },
-    {
-      "epoch": 2.465483234714004,
-      "eval_accuracy": 0.9850886517283898,
-      "eval_f1": 0.43230993363189985,
-      "eval_loss": 0.04674109071493149,
-      "eval_precision": 0.7342375366568915,
-      "eval_recall": 0.3063393744742678,
-      "eval_runtime": 14.36,
-      "eval_samples_per_second": 564.763,
-      "eval_steps_per_second": 2.228,
-      "step": 5000
-    },
-    {
-      "epoch": 2.712031558185404,
-      "grad_norm": 0.36567422747612,
-      "learning_rate": 1.9457593688362923e-05,
-      "loss": 0.0449,
-      "step": 5500
-    },
-    {
-      "epoch": 2.9585798816568047,
-      "grad_norm": 0.4769591689109802,
-      "learning_rate": 1.940828402366864e-05,
-      "loss": 0.0433,
-      "step": 6000
-    },
-    {
-      "epoch": 2.9585798816568047,
-      "eval_accuracy": 0.9859475317828139,
-      "eval_f1": 0.484694142716075,
-      "eval_loss": 0.042755745351314545,
-      "eval_precision": 0.7564892926670993,
-      "eval_recall": 0.3565802554102623,
-      "eval_runtime": 14.1095,
-      "eval_samples_per_second": 574.791,
-      "eval_steps_per_second": 2.268,
-      "step": 6000
-    },
-    {
-      "epoch": 3.2051282051282053,
-      "grad_norm": 0.4941834807395935,
-      "learning_rate": 1.935897435897436e-05,
-      "loss": 0.0405,
-      "step": 6500
-    },
-    {
-      "epoch": 3.4516765285996054,
-      "grad_norm": 0.39021360874176025,
-      "learning_rate": 1.930966469428008e-05,
-      "loss": 0.0391,
-      "step": 7000
-    },
-    {
-      "epoch": 3.4516765285996054,
-      "eval_accuracy": 0.9866122425840101,
-      "eval_f1": 0.5389046177877574,
-      "eval_loss": 0.04075852409005165,
-      "eval_precision": 0.7450398164394655,
-      "eval_recall": 0.4221151640284469,
-      "eval_runtime": 14.2258,
-      "eval_samples_per_second": 570.09,
-      "eval_steps_per_second": 2.249,
-      "step": 7000
-    },
-    {
-      "epoch": 3.698224852071006,
-      "grad_norm": 0.45788881182670593,
-      "learning_rate": 1.92603550295858e-05,
-      "loss": 0.0381,
-      "step": 7500
-    },
-    {
-      "epoch": 3.9447731755424065,
-      "grad_norm": 0.3994843363761902,
-      "learning_rate": 1.921104536489152e-05,
-      "loss": 0.0378,
-      "step": 8000
-    },
-    {
-      "epoch": 3.9447731755424065,
-      "eval_accuracy": 0.9867312952648214,
-      "eval_f1": 0.5526994744386049,
-      "eval_loss": 0.03952678292989731,
-      "eval_precision": 0.7365338087355151,
-      "eval_recall": 0.4423032805689378,
-      "eval_runtime": 14.2307,
-      "eval_samples_per_second": 569.894,
-      "eval_steps_per_second": 2.249,
-      "step": 8000
-    },
-    {
-      "epoch": 4.191321499013807,
-      "grad_norm": 0.40335726737976074,
-      "learning_rate": 1.916173570019724e-05,
-      "loss": 0.0349,
-      "step": 8500
-    },
-    {
-      "epoch": 4.437869822485207,
-      "grad_norm": 0.43202295899391174,
-      "learning_rate": 1.911242603550296e-05,
-      "loss": 0.0338,
-      "step": 9000
-    },
-    {
-      "epoch": 4.437869822485207,
-      "eval_accuracy": 0.986986408152274,
-      "eval_f1": 0.5843744341843201,
-      "eval_loss": 0.03873027116060257,
-      "eval_precision": 0.7160288408208542,
-      "eval_recall": 0.49361474344268563,
-      "eval_runtime": 14.0731,
-      "eval_samples_per_second": 576.277,
-      "eval_steps_per_second": 2.274,
-      "step": 9000
-    },
-    {
-      "epoch": 4.684418145956608,
-      "grad_norm": 0.3353199064731598,
-      "learning_rate": 1.906311637080868e-05,
-      "loss": 0.034,
-      "step": 9500
-    },
-    {
-      "epoch": 4.930966469428008,
-      "grad_norm": 0.3929939270019531,
-      "learning_rate": 1.90138067061144e-05,
-      "loss": 0.0333,
-      "step": 10000
-    },
-    {
-      "epoch": 4.930966469428008,
-      "eval_accuracy": 0.9870771149567017,
-      "eval_f1": 0.5952951620062139,
-      "eval_loss": 0.037975214421749115,
-      "eval_precision": 0.7094044218766529,
-      "eval_recall": 0.5128087481838343,
-      "eval_runtime": 14.136,
-      "eval_samples_per_second": 573.711,
-      "eval_steps_per_second": 2.264,
-      "step": 10000
-    },
-    {
-      "epoch": 5.177514792899408,
-      "grad_norm": 0.2832840085029602,
-      "learning_rate": 1.896459566074951e-05,
-      "loss": 0.0303,
-      "step": 10500
-    },
-    {
-      "epoch": 5.424063116370808,
-      "grad_norm": 0.43404069542884827,
-      "learning_rate": 1.891528599605523e-05,
-      "loss": 0.0301,
-      "step": 11000
-    },
-    {
-      "epoch": 5.424063116370808,
-      "eval_accuracy": 0.9875660813243193,
-      "eval_f1": 0.6041958041958042,
-      "eval_loss": 0.03708931431174278,
-      "eval_precision": 0.7367957746478874,
-      "eval_recall": 0.5120440467997247,
-      "eval_runtime": 14.265,
-      "eval_samples_per_second": 568.525,
-      "eval_steps_per_second": 2.243,
-      "step": 11000
     }
   ],
   "logging_steps": 500,
@@ -321,7 +61,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 248019064848396.0,
+  "total_flos": 22523300934480.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:247cccd0c3dbe5c727fd031150f0944d463e134495ac1c2a8628ab0ed563cdcb
+oid sha256:b0815310043095bb92c789617e2f93d478e30d81c4c444abbe71b4d0c0c4fde6
 size 5240