Retrained with larger dataset

all_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 10.0,
-    "total_flos": 1.7358769019486208e+16,
-    "train_loss": 1.7481338489563856,
-    "train_runtime": 1121.6717,
-    "train_samples": 10693,
-    "train_samples_per_second": 95.331,
-    "train_steps_per_second": 11.92
+    "total_flos": 3.2973681787109376e+16,
+    "train_loss": 1.6927287036132812,
+    "train_runtime": 2178.3738,
+    "train_samples": 20000,
+    "train_samples_per_second": 91.812,
+    "train_steps_per_second": 11.476
 }

config.json

@@ -56,7 +56,7 @@
   },
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
-  "transformers_version": "4.44.2",
+  "transformers_version": "4.45.1",
   "use_cache": true,
   "vocab_size": 32128
 }

generation_config.json

@@ -3,5 +3,5 @@
   "decoder_start_token_id": 0,
   "eos_token_id": 1,
   "pad_token_id": 0,
-  "transformers_version": "4.44.2"
+  "transformers_version": "4.45.1"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0e210fd97954bbbde1bc52e61324bb89dc084dc87d3f900928df9f351dacd6a9
+oid sha256:92c440e1e5c5a811a5e853a1f7889a4c1dbb779680e9164c82fb83d1fa7ae909
 size 307867048

tokenizer_config.json

@@ -927,7 +927,7 @@
     "<extra_id_98>",
     "<extra_id_99>"
   ],
-  "clean_up_tokenization_spaces": true,
+  "clean_up_tokenization_spaces": false,
   "eos_token": "</s>",
   "extra_ids": 100,
   "model_max_length": 512,

train_results.json

@@ -1,9 +1,9 @@
 {
     "epoch": 10.0,
-    "total_flos": 1.7358769019486208e+16,
-    "train_loss": 1.7481338489563856,
-    "train_runtime": 1121.6717,
-    "train_samples": 10693,
-    "train_samples_per_second": 95.331,
-    "train_steps_per_second": 11.92
+    "total_flos": 3.2973681787109376e+16,
+    "train_loss": 1.6927287036132812,
+    "train_runtime": 2178.3738,
+    "train_samples": 20000,
+    "train_samples_per_second": 91.812,
+    "train_steps_per_second": 11.476
 }

trainer_state.json

@@ -2,206 +2,374 @@
   "best_metric": null,
   "best_model_checkpoint": null,
   "epoch": 10.0,
-  "eval_steps": 500,
-  "global_step": 13370,
+  "eval_steps": 1000.0,
+  "global_step": 25000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.3739715781600598,
-      "grad_norm": 3.664752721786499,
-      "learning_rate": 4.813014210919971e-05,
-      "loss": 2.1969,
+      "epoch": 0.2,
+      "grad_norm": 3.1640737056732178,
+      "learning_rate": 4.9e-05,
+      "loss": 2.212,
       "step": 500
     },
     {
-      "epoch": 0.7479431563201197,
-      "grad_norm": 2.6418004035949707,
-      "learning_rate": 4.62602842183994e-05,
-      "loss": 2.0162,
+      "epoch": 0.4,
+      "grad_norm": 23.492929458618164,
+      "learning_rate": 4.8e-05,
+      "loss": 2.0008,
       "step": 1000
     },
     {
-      "epoch": 1.1219147344801794,
-      "grad_norm": 2.6499061584472656,
-      "learning_rate": 4.4390426327599105e-05,
-      "loss": 1.928,
+      "epoch": 0.6,
+      "grad_norm": 2.86879563331604,
+      "learning_rate": 4.7e-05,
+      "loss": 1.9454,
       "step": 1500
     },
     {
-      "epoch": 1.4958863126402393,
-      "grad_norm": 2.92419695854187,
-      "learning_rate": 4.252056843679881e-05,
-      "loss": 1.8714,
+      "epoch": 0.8,
+      "grad_norm": 2.5470571517944336,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 1.9098,
       "step": 2000
     },
     {
-      "epoch": 1.8698578908002992,
-      "grad_norm": 2.9675424098968506,
-      "learning_rate": 4.06507105459985e-05,
-      "loss": 1.8688,
+      "epoch": 1.0,
+      "grad_norm": 3.299417018890381,
+      "learning_rate": 4.5e-05,
+      "loss": 1.8775,
       "step": 2500
     },
     {
-      "epoch": 2.243829468960359,
-      "grad_norm": 2.4510722160339355,
-      "learning_rate": 3.878085265519821e-05,
-      "loss": 1.8199,
+      "epoch": 1.2,
+      "grad_norm": 3.0483977794647217,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 1.8398,
       "step": 3000
     },
     {
-      "epoch": 2.6178010471204187,
-      "grad_norm": 3.3887815475463867,
-      "learning_rate": 3.691099476439791e-05,
-      "loss": 1.7878,
+      "epoch": 1.4,
+      "grad_norm": 3.0028903484344482,
+      "learning_rate": 4.3e-05,
+      "loss": 1.8147,
       "step": 3500
     },
     {
-      "epoch": 2.9917726252804786,
-      "grad_norm": 2.7540111541748047,
-      "learning_rate": 3.5041136873597606e-05,
-      "loss": 1.804,
+      "epoch": 1.6,
+      "grad_norm": 2.972844123840332,
+      "learning_rate": 4.2e-05,
+      "loss": 1.8035,
       "step": 4000
     },
     {
-      "epoch": 3.3657442034405385,
-      "grad_norm": 2.7998127937316895,
-      "learning_rate": 3.317127898279731e-05,
-      "loss": 1.764,
+      "epoch": 1.8,
+      "grad_norm": 2.82574200630188,
+      "learning_rate": 4.1e-05,
+      "loss": 1.7974,
       "step": 4500
     },
     {
-      "epoch": 3.7397157816005984,
-      "grad_norm": 3.000854015350342,
-      "learning_rate": 3.130142109199701e-05,
-      "loss": 1.7486,
+      "epoch": 2.0,
+      "grad_norm": 2.6531100273132324,
+      "learning_rate": 4e-05,
+      "loss": 1.7818,
       "step": 5000
     },
     {
-      "epoch": 4.113687359760658,
-      "grad_norm": 2.85657000541687,
-      "learning_rate": 2.9431563201196712e-05,
-      "loss": 1.7418,
+      "epoch": 2.2,
+      "grad_norm": 3.6078109741210938,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 1.7492,
       "step": 5500
     },
     {
-      "epoch": 4.487658937920718,
-      "grad_norm": 2.8628413677215576,
-      "learning_rate": 2.7561705310396414e-05,
-      "loss": 1.7193,
+      "epoch": 2.4,
+      "grad_norm": 2.573765516281128,
+      "learning_rate": 3.8e-05,
+      "loss": 1.7536,
       "step": 6000
     },
     {
-      "epoch": 4.861630516080778,
-      "grad_norm": 3.066136360168457,
-      "learning_rate": 2.569184741959611e-05,
-      "loss": 1.7083,
+      "epoch": 2.6,
+      "grad_norm": 3.07545804977417,
+      "learning_rate": 3.7e-05,
+      "loss": 1.7341,
       "step": 6500
     },
     {
-      "epoch": 5.2356020942408374,
-      "grad_norm": 2.46604585647583,
-      "learning_rate": 2.382198952879581e-05,
-      "loss": 1.6959,
+      "epoch": 2.8,
+      "grad_norm": 2.4871368408203125,
+      "learning_rate": 3.6e-05,
+      "loss": 1.7362,
       "step": 7000
     },
     {
-      "epoch": 5.609573672400898,
-      "grad_norm": 2.4070425033569336,
-      "learning_rate": 2.1952131637995513e-05,
-      "loss": 1.6787,
+      "epoch": 3.0,
+      "grad_norm": 2.881721019744873,
+      "learning_rate": 3.5e-05,
+      "loss": 1.7426,
       "step": 7500
     },
     {
-      "epoch": 5.983545250560957,
-      "grad_norm": 2.59104585647583,
-      "learning_rate": 2.0082273747195215e-05,
-      "loss": 1.6979,
+      "epoch": 3.2,
+      "grad_norm": 3.6232352256774902,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 1.6958,
       "step": 8000
     },
     {
-      "epoch": 6.3575168287210175,
-      "grad_norm": 2.270085334777832,
-      "learning_rate": 1.8212415856394914e-05,
-      "loss": 1.6618,
+      "epoch": 3.4,
+      "grad_norm": 2.4172329902648926,
+      "learning_rate": 3.3e-05,
+      "loss": 1.6961,
       "step": 8500
     },
     {
-      "epoch": 6.731488406881077,
-      "grad_norm": 2.6211342811584473,
-      "learning_rate": 1.6342557965594616e-05,
-      "loss": 1.665,
+      "epoch": 3.6,
+      "grad_norm": 2.5114290714263916,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 1.6961,
       "step": 9000
     },
     {
-      "epoch": 7.105459985041137,
-      "grad_norm": 2.876598358154297,
-      "learning_rate": 1.4472700074794315e-05,
-      "loss": 1.6716,
+      "epoch": 3.8,
+      "grad_norm": 2.7974891662597656,
+      "learning_rate": 3.1e-05,
+      "loss": 1.6831,
       "step": 9500
     },
     {
-      "epoch": 7.479431563201197,
-      "grad_norm": 2.5392422676086426,
-      "learning_rate": 1.2602842183994019e-05,
-      "loss": 1.6354,
+      "epoch": 4.0,
+      "grad_norm": 2.511326551437378,
+      "learning_rate": 3e-05,
+      "loss": 1.7117,
       "step": 10000
     },
     {
-      "epoch": 7.853403141361256,
-      "grad_norm": 2.6859071254730225,
-      "learning_rate": 1.0732984293193717e-05,
-      "loss": 1.6581,
+      "epoch": 4.2,
+      "grad_norm": 2.3756415843963623,
+      "learning_rate": 2.9e-05,
+      "loss": 1.6684,
       "step": 10500
     },
     {
-      "epoch": 8.227374719521316,
-      "grad_norm": 2.5266973972320557,
-      "learning_rate": 8.863126402393418e-06,
-      "loss": 1.6501,
+      "epoch": 4.4,
+      "grad_norm": 2.2897322177886963,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 1.6826,
       "step": 11000
     },
     {
-      "epoch": 8.601346297681376,
-      "grad_norm": 1.9921244382858276,
-      "learning_rate": 6.993268511593119e-06,
-      "loss": 1.6351,
+      "epoch": 4.6,
+      "grad_norm": 2.7029635906219482,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 1.6649,
       "step": 11500
     },
     {
-      "epoch": 8.975317875841435,
-      "grad_norm": 2.633855104446411,
-      "learning_rate": 5.12341062079282e-06,
-      "loss": 1.6507,
+      "epoch": 4.8,
+      "grad_norm": 2.453059673309326,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 1.6488,
       "step": 12000
     },
     {
-      "epoch": 9.349289454001497,
-      "grad_norm": 2.9301204681396484,
-      "learning_rate": 3.2535527299925206e-06,
-      "loss": 1.6296,
+      "epoch": 5.0,
+      "grad_norm": 2.173687219619751,
+      "learning_rate": 2.5e-05,
+      "loss": 1.652,
       "step": 12500
     },
     {
-      "epoch": 9.723261032161556,
-      "grad_norm": 2.9712412357330322,
-      "learning_rate": 1.3836948391922214e-06,
-      "loss": 1.6316,
+      "epoch": 5.2,
+      "grad_norm": 2.9122655391693115,
+      "learning_rate": 2.4e-05,
+      "loss": 1.6354,
       "step": 13000
     },
+    {
+      "epoch": 5.4,
+      "grad_norm": 2.8792638778686523,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 1.6423,
+      "step": 13500
+    },
+    {
+      "epoch": 5.6,
+      "grad_norm": 3.0389626026153564,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 1.6591,
+      "step": 14000
+    },
+    {
+      "epoch": 5.8,
+      "grad_norm": 2.6412346363067627,
+      "learning_rate": 2.1e-05,
+      "loss": 1.6365,
+      "step": 14500
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 2.6655163764953613,
+      "learning_rate": 2e-05,
+      "loss": 1.6133,
+      "step": 15000
+    },
+    {
+      "epoch": 6.2,
+      "grad_norm": 2.6776795387268066,
+      "learning_rate": 1.9e-05,
+      "loss": 1.603,
+      "step": 15500
+    },
+    {
+      "epoch": 6.4,
+      "grad_norm": 3.4206550121307373,
+      "learning_rate": 1.8e-05,
+      "loss": 1.6142,
+      "step": 16000
+    },
+    {
+      "epoch": 6.6,
+      "grad_norm": 3.59487247467041,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 1.6193,
+      "step": 16500
+    },
+    {
+      "epoch": 6.8,
+      "grad_norm": 2.707260847091675,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.618,
+      "step": 17000
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 2.4932329654693604,
+      "learning_rate": 1.5e-05,
+      "loss": 1.6179,
+      "step": 17500
+    },
+    {
+      "epoch": 7.2,
+      "grad_norm": 2.7114312648773193,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 1.6064,
+      "step": 18000
+    },
+    {
+      "epoch": 7.4,
+      "grad_norm": 2.9900901317596436,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.6079,
+      "step": 18500
+    },
+    {
+      "epoch": 7.6,
+      "grad_norm": 2.5069007873535156,
+      "learning_rate": 1.2e-05,
+      "loss": 1.5917,
+      "step": 19000
+    },
+    {
+      "epoch": 7.8,
+      "grad_norm": 2.816830635070801,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.6159,
+      "step": 19500
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 2.500258445739746,
+      "learning_rate": 1e-05,
+      "loss": 1.586,
+      "step": 20000
+    },
+    {
+      "epoch": 8.2,
+      "grad_norm": 2.392134189605713,
+      "learning_rate": 9e-06,
+      "loss": 1.5948,
+      "step": 20500
+    },
+    {
+      "epoch": 8.4,
+      "grad_norm": 2.19275164604187,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.5904,
+      "step": 21000
+    },
+    {
+      "epoch": 8.6,
+      "grad_norm": 2.1356213092803955,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 1.5772,
+      "step": 21500
+    },
+    {
+      "epoch": 8.8,
+      "grad_norm": 2.0939230918884277,
+      "learning_rate": 6e-06,
+      "loss": 1.604,
+      "step": 22000
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 2.8503177165985107,
+      "learning_rate": 5e-06,
+      "loss": 1.5878,
+      "step": 22500
+    },
+    {
+      "epoch": 9.2,
+      "grad_norm": 2.6212220191955566,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.5893,
+      "step": 23000
+    },
+    {
+      "epoch": 9.4,
+      "grad_norm": 2.9165279865264893,
+      "learning_rate": 3e-06,
+      "loss": 1.5796,
+      "step": 23500
+    },
+    {
+      "epoch": 9.6,
+      "grad_norm": 2.8491413593292236,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.5832,
+      "step": 24000
+    },
+    {
+      "epoch": 9.8,
+      "grad_norm": 3.690595865249634,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.5872,
+      "step": 24500
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 3.193439245223999,
+      "learning_rate": 0.0,
+      "loss": 1.5781,
+      "step": 25000
+    },
     {
       "epoch": 10.0,
-      "step": 13370,
-      "total_flos": 1.7358769019486208e+16,
-      "train_loss": 1.7481338489563856,
-      "train_runtime": 1121.6717,
-      "train_samples_per_second": 95.331,
-      "train_steps_per_second": 11.92
+      "step": 25000,
+      "total_flos": 3.2973681787109376e+16,
+      "train_loss": 1.6927287036132812,
+      "train_runtime": 2178.3738,
+      "train_samples_per_second": 91.812,
+      "train_steps_per_second": 11.476
     }
   ],
   "logging_steps": 500,
-  "max_steps": 13370,
+  "max_steps": 25000,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 10,
   "save_steps": 500,
@@ -217,7 +385,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 1.7358769019486208e+16,
+  "total_flos": 3.2973681787109376e+16,
   "train_batch_size": 8,
   "trial_name": null,
   "trial_params": null