Training in progress, epoch 32, checkpoint

checkpoint-6000/adapter_config.json

@@ -1,6 +1,6 @@
 {
   "auto_mapping": null,
-  "base_model_name_or_path": "Qwen/Qwen1.5-4B",
+  "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,

checkpoint-6000/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afa2f84342528274e06f400f728c0809b27837a336bc44a0c3ee1dec3721c82a
+size 143269386

checkpoint-6000/added_tokens.json

@@ -1,5 +1,5 @@
 {
-  "<|endoftext|>": 151643,
-  "<|im_end|>": 151645,
-  "<|im_start|>": 151644
+  "</s>": 2,
+  "<s>": 1,
+  "<unk>": 0
 }

checkpoint-6000/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:50ed8866980625ff21bf87e0bdcf636655beb72b40dec545d0c60c23f56d8fb5
-size 224537202
+oid sha256:996c3b660137dab653bfe4fcb2e7e862901c52fd9cdcda0cd3880d3096e3787b
+size 286585234

checkpoint-6000/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e166c2499594e2924b8425fe6fc8b55e54743ade2c7abb81a6c6e448decba153
+oid sha256:79cee95dc60d3952ea75fabc9a01edd738df8b9a060c8b2c3ef636fb471bfcb0
 size 15024

checkpoint-6000/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fbc511338ae25d510d2fb1024520a22da4114462494439cf4993f60efb55ae3d
+oid sha256:e0d87b8aa500e7a3af9fef0dcb706bf22396a5be456f8336f82507c0aa20bfba
 size 15024

checkpoint-6000/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c4e811e29e0ec5bff116f34195d95e270787cdfffa8411ef4b2a5d2da3b4bf12
+oid sha256:f176d2b9c099ed7c7c13565810faed791ce4cc8a88dd726dc26a991322eeeb0b
 size 15024

checkpoint-6000/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a7058edd72fae73d84376ff95a60d0fc5e898791861c11a4b3fd7d639ef05223
+oid sha256:f95f3af8677378ed28d6df26f8636932a4f4c096cf49abc96b928804cbda3b6c
 size 15024

checkpoint-6000/special_tokens_map.json

@@ -1,14 +1,6 @@
 {
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "eos_token": {
-    "content": "<|endoftext|>",
-    "lstrip": false,
-    "normalized": false,
-    "rstrip": false,
-    "single_word": false
-  },
-  "pad_token": "<|endoftext|>"
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "pad_token": "</s>",
+  "unk_token": "<unk>"
 }

checkpoint-6000/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

checkpoint-6000/tokenizer_config.json

@@ -1,24 +1,23 @@
 {
-  "add_prefix_space": false,
   "added_tokens_decoder": {
-    "151643": {
-      "content": "<|endoftext|>",
+    "0": {
+      "content": "<unk>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151644": {
-      "content": "<|im_start|>",
+    "1": {
+      "content": "<s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
       "single_word": false,
       "special": true
     },
-    "151645": {
-      "content": "<|im_end|>",
+    "2": {
+      "content": "</s>",
       "lstrip": false,
       "normalized": false,
       "rstrip": false,
@@ -26,18 +25,16 @@
       "special": true
     }
   },
-  "additional_special_tokens": [
-    "<|im_start|>",
-    "<|im_end|>"
-  ],
-  "bos_token": null,
-  "chat_template": "{% for message in messages %}{% if loop.first and messages[0]['role'] != 'system' %}{{ '<|im_start|>system\nYou are a helpful assistant<|im_end|>\n' }}{% endif %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
+  "additional_special_tokens": [],
+  "bos_token": "<s>",
   "clean_up_tokenization_spaces": false,
-  "eos_token": "<|endoftext|>",
-  "errors": "replace",
-  "model_max_length": 32768,
-  "pad_token": "<|endoftext|>",
-  "split_special_tokens": false,
-  "tokenizer_class": "Qwen2Tokenizer",
-  "unk_token": null
+  "eos_token": "</s>",
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "left",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": true
 }

checkpoint-6000/trainer_state.json

@@ -9,1053 +9,979 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.5333333333333333,
-      "grad_norm": 0.275871217250824,
+      "epoch": 0.53,
       "learning_rate": 3e-05,
-      "loss": 1.8545,
+      "loss": 1.4597,
       "step": 100
     },
     {
-      "epoch": 0.9973333333333333,
-      "eval_accuracy": 0.6037399103139014,
-      "eval_loss": 1.7123007774353027,
-      "eval_runtime": 6.2976,
-      "eval_samples_per_second": 79.395,
-      "eval_steps_per_second": 10.004,
+      "epoch": 1.0,
+      "eval_accuracy": 0.6618980392156862,
+      "eval_loss": 1.2718806266784668,
+      "eval_runtime": 7.6194,
+      "eval_samples_per_second": 65.622,
+      "eval_steps_per_second": 8.268,
       "step": 187
     },
     {
-      "epoch": 0.9973333333333333,
-      "eval_exact_match": 14.4,
-      "eval_f1": 22.843015873015876,
-      "eval_qa_bleu": 9.471299126615676,
-      "eval_qa_exact_match": 0.112,
-      "eval_recite_bleu": 12.267251092702137,
+      "epoch": 1.0,
+      "eval_exact_match": 22.4,
+      "eval_f1": 30.941064425770318,
+      "eval_qa_bleu": 7.097403515221558,
+      "eval_qa_exact_match": 0.196,
+      "eval_recite_bleu": 13.258373740102225,
       "eval_recite_exact_match": 0.0,
       "step": 187
     },
     {
-      "epoch": 1.0666666666666667,
-      "grad_norm": 0.32765188813209534,
+      "epoch": 1.07,
       "learning_rate": 3e-05,
-      "loss": 1.7408,
+      "loss": 1.2905,
       "step": 200
     },
     {
       "epoch": 1.6,
-      "grad_norm": 0.3532409369945526,
       "learning_rate": 3e-05,
-      "loss": 1.7063,
+      "loss": 1.2157,
       "step": 300
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.6058565022421525,
-      "eval_loss": 1.69453763961792,
-      "eval_runtime": 6.5928,
-      "eval_samples_per_second": 75.841,
-      "eval_steps_per_second": 9.556,
+      "eval_accuracy": 0.6670039215686274,
+      "eval_loss": 1.206207036972046,
+      "eval_runtime": 7.6793,
+      "eval_samples_per_second": 65.11,
+      "eval_steps_per_second": 8.204,
       "step": 375
     },
     {
       "epoch": 2.0,
-      "eval_exact_match": 14.4,
-      "eval_f1": 23.29731601731603,
-      "eval_qa_bleu": 9.4408132575913,
-      "eval_qa_exact_match": 0.11,
-      "eval_recite_bleu": 13.027794116434837,
+      "eval_exact_match": 3.2,
+      "eval_f1": 3.96976236704394,
+      "eval_qa_bleu": 0.20484366294597936,
+      "eval_qa_exact_match": 0.026,
+      "eval_recite_bleu": 2.5778714769111824e-05,
       "eval_recite_exact_match": 0.0,
       "step": 375
     },
     {
-      "epoch": 2.1333333333333333,
-      "grad_norm": 0.3777780532836914,
+      "epoch": 2.13,
       "learning_rate": 3e-05,
-      "loss": 1.7094,
+      "loss": 1.1852,
       "step": 400
     },
     {
-      "epoch": 2.6666666666666665,
-      "grad_norm": 0.3971036374568939,
+      "epoch": 2.67,
       "learning_rate": 3e-05,
-      "loss": 1.6702,
+      "loss": 1.1861,
       "step": 500
     },
     {
-      "epoch": 2.997333333333333,
-      "eval_accuracy": 0.6065919282511211,
-      "eval_loss": 1.6851236820220947,
-      "eval_runtime": 6.3104,
-      "eval_samples_per_second": 79.234,
-      "eval_steps_per_second": 9.984,
+      "epoch": 3.0,
+      "eval_accuracy": 0.667513725490196,
+      "eval_loss": 1.19828462600708,
+      "eval_runtime": 7.1425,
+      "eval_samples_per_second": 70.004,
+      "eval_steps_per_second": 8.82,
       "step": 562
     },
     {
-      "epoch": 2.997333333333333,
-      "eval_exact_match": 14.0,
-      "eval_f1": 23.306825396825417,
-      "eval_qa_bleu": 8.683248465366908,
-      "eval_qa_exact_match": 0.11,
-      "eval_recite_bleu": 12.39336052958371,
-      "eval_recite_exact_match": 0.006,
+      "epoch": 3.0,
+      "eval_exact_match": 4.4,
+      "eval_f1": 6.333660342551587,
+      "eval_qa_bleu": 0.522391008398391,
+      "eval_qa_exact_match": 0.028,
+      "eval_recite_bleu": 0.027569975851374308,
+      "eval_recite_exact_match": 0.0,
       "step": 562
     },
     {
       "epoch": 3.2,
-      "grad_norm": 0.4782978594303131,
       "learning_rate": 3e-05,
-      "loss": 1.6574,
+      "loss": 1.1722,
       "step": 600
     },
     {
-      "epoch": 3.7333333333333334,
-      "grad_norm": 0.52582186460495,
+      "epoch": 3.73,
       "learning_rate": 3e-05,
-      "loss": 1.6356,
+      "loss": 1.1445,
       "step": 700
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.6074977578475337,
-      "eval_loss": 1.682567834854126,
-      "eval_runtime": 6.65,
-      "eval_samples_per_second": 75.188,
-      "eval_steps_per_second": 9.474,
+      "eval_accuracy": 0.6672,
+      "eval_loss": 1.197709560394287,
+      "eval_runtime": 7.7427,
+      "eval_samples_per_second": 64.577,
+      "eval_steps_per_second": 8.137,
       "step": 750
     },
     {
       "epoch": 4.0,
-      "eval_exact_match": 15.0,
-      "eval_f1": 23.366825396825416,
-      "eval_qa_bleu": 9.05994637010322,
-      "eval_qa_exact_match": 0.114,
-      "eval_recite_bleu": 13.060415058384661,
-      "eval_recite_exact_match": 0.006,
+      "eval_exact_match": 7.4,
+      "eval_f1": 10.950016966795278,
+      "eval_qa_bleu": 0.6700083846206139,
+      "eval_qa_exact_match": 0.052,
+      "eval_recite_bleu": 1.259776140465531,
+      "eval_recite_exact_match": 0.0,
       "step": 750
     },
     {
-      "epoch": 4.266666666666667,
-      "grad_norm": 0.6418241262435913,
+      "epoch": 4.27,
       "learning_rate": 3e-05,
-      "loss": 1.6005,
+      "loss": 1.1018,
       "step": 800
     },
     {
       "epoch": 4.8,
-      "grad_norm": 0.6805307269096375,
       "learning_rate": 3e-05,
-      "loss": 1.5775,
+      "loss": 1.1,
       "step": 900
     },
     {
-      "epoch": 4.997333333333334,
-      "eval_accuracy": 0.6071031390134529,
-      "eval_loss": 1.69106125831604,
-      "eval_runtime": 5.6598,
-      "eval_samples_per_second": 88.343,
-      "eval_steps_per_second": 11.131,
+      "epoch": 5.0,
+      "eval_accuracy": 0.6664941176470588,
+      "eval_loss": 1.2078713178634644,
+      "eval_runtime": 7.7721,
+      "eval_samples_per_second": 64.333,
+      "eval_steps_per_second": 8.106,
       "step": 937
     },
     {
-      "epoch": 4.997333333333334,
-      "eval_exact_match": 15.2,
-      "eval_f1": 23.38539682539684,
-      "eval_qa_bleu": 11.854591678140883,
-      "eval_qa_exact_match": 0.112,
-      "eval_recite_bleu": 12.694062454678122,
-      "eval_recite_exact_match": 0.004,
+      "epoch": 5.0,
+      "eval_exact_match": 6.6,
+      "eval_f1": 10.12048028049545,
+      "eval_qa_bleu": 0.8314625442591895,
+      "eval_qa_exact_match": 0.05,
+      "eval_recite_bleu": 0.8030246920772521,
+      "eval_recite_exact_match": 0.0,
       "step": 937
     },
     {
-      "epoch": 5.333333333333333,
-      "grad_norm": 0.8362048864364624,
+      "epoch": 5.33,
       "learning_rate": 3e-05,
-      "loss": 1.5173,
+      "loss": 1.0666,
       "step": 1000
     },
     {
-      "epoch": 5.866666666666667,
-      "grad_norm": 0.8521270155906677,
+      "epoch": 5.87,
       "learning_rate": 3e-05,
-      "loss": 1.529,
+      "loss": 1.0439,
       "step": 1100
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.6069417040358744,
-      "eval_loss": 1.7066757678985596,
-      "eval_runtime": 6.2922,
-      "eval_samples_per_second": 79.464,
-      "eval_steps_per_second": 10.012,
+      "eval_accuracy": 0.6649725490196079,
+      "eval_loss": 1.22383713722229,
+      "eval_runtime": 7.6639,
+      "eval_samples_per_second": 65.241,
+      "eval_steps_per_second": 8.22,
       "step": 1125
     },
     {
       "epoch": 6.0,
-      "eval_exact_match": 15.2,
-      "eval_f1": 24.04682539682541,
-      "eval_qa_bleu": 12.150293484705452,
-      "eval_qa_exact_match": 0.106,
-      "eval_recite_bleu": 12.732083161926754,
-      "eval_recite_exact_match": 0.002,
+      "eval_exact_match": 9.0,
+      "eval_f1": 14.19791549030679,
+      "eval_qa_bleu": 2.7960558519683327,
+      "eval_qa_exact_match": 0.07,
+      "eval_recite_bleu": 2.082318255561628,
+      "eval_recite_exact_match": 0.0,
       "step": 1125
     },
     {
       "epoch": 6.4,
-      "grad_norm": 0.9372308254241943,
       "learning_rate": 3e-05,
-      "loss": 1.4613,
+      "loss": 1.0029,
       "step": 1200
     },
     {
-      "epoch": 6.933333333333334,
-      "grad_norm": 1.0236693620681763,
+      "epoch": 6.93,
       "learning_rate": 3e-05,
-      "loss": 1.457,
+      "loss": 0.9888,
       "step": 1300
     },
     {
-      "epoch": 6.997333333333334,
-      "eval_accuracy": 0.6055515695067265,
-      "eval_loss": 1.72792649269104,
-      "eval_runtime": 6.3278,
-      "eval_samples_per_second": 79.016,
-      "eval_steps_per_second": 9.956,
+      "epoch": 7.0,
+      "eval_accuracy": 0.6638117647058823,
+      "eval_loss": 1.2457345724105835,
+      "eval_runtime": 6.7577,
+      "eval_samples_per_second": 73.99,
+      "eval_steps_per_second": 9.323,
       "step": 1312
     },
     {
-      "epoch": 6.997333333333334,
-      "eval_exact_match": 15.0,
-      "eval_f1": 23.28492063492065,
-      "eval_qa_bleu": 9.849658422779212,
-      "eval_qa_exact_match": 0.114,
-      "eval_recite_bleu": 12.3997433702838,
-      "eval_recite_exact_match": 0.006,
+      "epoch": 7.0,
+      "eval_exact_match": 10.6,
+      "eval_f1": 15.30586844512479,
+      "eval_qa_bleu": 0.94556412718174,
+      "eval_qa_exact_match": 0.086,
+      "eval_recite_bleu": 3.1558565534273777,
+      "eval_recite_exact_match": 0.0,
       "step": 1312
     },
     {
-      "epoch": 7.466666666666667,
-      "grad_norm": 1.0773049592971802,
+      "epoch": 7.47,
       "learning_rate": 3e-05,
-      "loss": 1.3996,
+      "loss": 0.9453,
       "step": 1400
     },
     {
       "epoch": 8.0,
-      "grad_norm": 1.0705862045288086,
       "learning_rate": 3e-05,
-      "loss": 1.3907,
+      "loss": 0.9364,
       "step": 1500
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.6046188340807175,
-      "eval_loss": 1.7511730194091797,
-      "eval_runtime": 5.5895,
-      "eval_samples_per_second": 89.453,
-      "eval_steps_per_second": 11.271,
+      "eval_accuracy": 0.6615921568627451,
+      "eval_loss": 1.2816376686096191,
+      "eval_runtime": 7.5902,
+      "eval_samples_per_second": 65.875,
+      "eval_steps_per_second": 8.3,
       "step": 1500
     },
     {
       "epoch": 8.0,
-      "eval_exact_match": 13.0,
-      "eval_f1": 19.93238095238096,
-      "eval_qa_bleu": 11.93942669425287,
-      "eval_qa_exact_match": 0.098,
-      "eval_recite_bleu": 11.949414001720045,
-      "eval_recite_exact_match": 0.006,
+      "eval_exact_match": 9.2,
+      "eval_f1": 12.989213707726288,
+      "eval_qa_bleu": 1.9797940311992752,
+      "eval_qa_exact_match": 0.074,
+      "eval_recite_bleu": 2.551621610315755,
+      "eval_recite_exact_match": 0.0,
       "step": 1500
     },
     {
-      "epoch": 8.533333333333333,
-      "grad_norm": 1.2424529790878296,
+      "epoch": 8.53,
       "learning_rate": 3e-05,
-      "loss": 1.3309,
+      "loss": 0.8889,
       "step": 1600
     },
     {
-      "epoch": 8.997333333333334,
-      "eval_accuracy": 0.6025022421524664,
-      "eval_loss": 1.7774019241333008,
-      "eval_runtime": 6.4013,
-      "eval_samples_per_second": 78.109,
-      "eval_steps_per_second": 9.842,
+      "epoch": 9.0,
+      "eval_accuracy": 0.6606196078431372,
+      "eval_loss": 1.3036144971847534,
+      "eval_runtime": 7.4845,
+      "eval_samples_per_second": 66.805,
+      "eval_steps_per_second": 8.417,
       "step": 1687
     },
     {
-      "epoch": 8.997333333333334,
-      "eval_exact_match": 13.6,
-      "eval_f1": 21.089523809523815,
-      "eval_qa_bleu": 10.094627644241111,
-      "eval_qa_exact_match": 0.094,
-      "eval_recite_bleu": 12.197681191541987,
-      "eval_recite_exact_match": 0.006,
+      "epoch": 9.0,
+      "eval_exact_match": 14.2,
+      "eval_f1": 19.66897045942666,
+      "eval_qa_bleu": 3.3530752528407004,
+      "eval_qa_exact_match": 0.112,
+      "eval_recite_bleu": 5.3293362575452425,
+      "eval_recite_exact_match": 0.002,
       "step": 1687
     },
     {
-      "epoch": 9.066666666666666,
-      "grad_norm": 1.535229206085205,
+      "epoch": 9.07,
       "learning_rate": 3e-05,
-      "loss": 1.3336,
+      "loss": 0.8853,
       "step": 1700
     },
     {
       "epoch": 9.6,
-      "grad_norm": 1.4936225414276123,
       "learning_rate": 3e-05,
-      "loss": 1.2841,
+      "loss": 0.8373,
       "step": 1800
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.6012645739910314,
-      "eval_loss": 1.804250717163086,
-      "eval_runtime": 6.385,
-      "eval_samples_per_second": 78.309,
-      "eval_steps_per_second": 9.867,
+      "eval_accuracy": 0.6587450980392157,
+      "eval_loss": 1.3337457180023193,
+      "eval_runtime": 6.6361,
+      "eval_samples_per_second": 75.345,
+      "eval_steps_per_second": 9.493,
       "step": 1875
     },
     {
       "epoch": 10.0,
-      "eval_exact_match": 13.2,
-      "eval_f1": 21.172832722832737,
-      "eval_qa_bleu": 9.841075001407239,
-      "eval_qa_exact_match": 0.096,
-      "eval_recite_bleu": 11.691398764835105,
-      "eval_recite_exact_match": 0.004,
+      "eval_exact_match": 14.4,
+      "eval_f1": 21.20886352105531,
+      "eval_qa_bleu": 2.50339041197013,
+      "eval_qa_exact_match": 0.106,
+      "eval_recite_bleu": 6.648771889251858,
+      "eval_recite_exact_match": 0.0,
       "step": 1875
     },
     {
-      "epoch": 10.133333333333333,
-      "grad_norm": 1.6687729358673096,
+      "epoch": 10.13,
       "learning_rate": 3e-05,
-      "loss": 1.2562,
+      "loss": 0.8353,
       "step": 1900
     },
     {
-      "epoch": 10.666666666666666,
-      "grad_norm": 1.437538504600525,
+      "epoch": 10.67,
       "learning_rate": 3e-05,
-      "loss": 1.2308,
+      "loss": 0.7949,
       "step": 2000
     },
     {
-      "epoch": 10.997333333333334,
-      "eval_accuracy": 0.6000717488789238,
-      "eval_loss": 1.8528188467025757,
-      "eval_runtime": 6.6644,
-      "eval_samples_per_second": 75.026,
-      "eval_steps_per_second": 9.453,
+      "epoch": 11.0,
+      "eval_accuracy": 0.6575058823529412,
+      "eval_loss": 1.367775797843933,
+      "eval_runtime": 7.4976,
+      "eval_samples_per_second": 66.688,
+      "eval_steps_per_second": 8.403,
       "step": 2062
     },
     {
-      "epoch": 10.997333333333334,
-      "eval_exact_match": 12.6,
-      "eval_f1": 20.865213675213692,
-      "eval_qa_bleu": 7.784433449171618,
-      "eval_qa_exact_match": 0.108,
-      "eval_recite_bleu": 12.305413068460787,
-      "eval_recite_exact_match": 0.008,
+      "epoch": 11.0,
+      "eval_exact_match": 15.4,
+      "eval_f1": 21.670913900913916,
+      "eval_qa_bleu": 4.131625289761948,
+      "eval_qa_exact_match": 0.104,
+      "eval_recite_bleu": 8.111027123751208,
+      "eval_recite_exact_match": 0.0,
       "step": 2062
     },
     {
       "epoch": 11.2,
-      "grad_norm": 1.708260416984558,
       "learning_rate": 3e-05,
-      "loss": 1.2011,
+      "loss": 0.7758,
       "step": 2100
     },
     {
-      "epoch": 11.733333333333333,
-      "grad_norm": 1.7026830911636353,
+      "epoch": 11.73,
       "learning_rate": 3e-05,
-      "loss": 1.1722,
+      "loss": 0.7539,
       "step": 2200
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.5987892376681614,
-      "eval_loss": 1.8851206302642822,
-      "eval_runtime": 5.5764,
-      "eval_samples_per_second": 89.663,
-      "eval_steps_per_second": 11.298,
+      "eval_accuracy": 0.6553725490196078,
+      "eval_loss": 1.3984336853027344,
+      "eval_runtime": 7.6892,
+      "eval_samples_per_second": 65.026,
+      "eval_steps_per_second": 8.193,
       "step": 2250
     },
     {
       "epoch": 12.0,
-      "eval_exact_match": 13.0,
-      "eval_f1": 21.17809523809525,
-      "eval_qa_bleu": 11.19937598521782,
-      "eval_qa_exact_match": 0.096,
-      "eval_recite_bleu": 12.500534642121808,
-      "eval_recite_exact_match": 0.01,
+      "eval_exact_match": 16.2,
+      "eval_f1": 23.074922854922868,
+      "eval_qa_bleu": 10.80394366395148,
+      "eval_qa_exact_match": 0.132,
+      "eval_recite_bleu": 9.158207459098465,
+      "eval_recite_exact_match": 0.002,
       "step": 2250
     },
     {
-      "epoch": 12.266666666666667,
-      "grad_norm": 1.6281794309616089,
+      "epoch": 12.27,
       "learning_rate": 3e-05,
-      "loss": 1.151,
+      "loss": 0.7415,
       "step": 2300
     },
     {
       "epoch": 12.8,
-      "grad_norm": 1.686490535736084,
       "learning_rate": 3e-05,
-      "loss": 1.1354,
+      "loss": 0.7133,
       "step": 2400
     },
     {
-      "epoch": 12.997333333333334,
-      "eval_accuracy": 0.5979910313901345,
-      "eval_loss": 1.9114351272583008,
-      "eval_runtime": 5.5718,
-      "eval_samples_per_second": 89.737,
-      "eval_steps_per_second": 11.307,
+      "epoch": 13.0,
+      "eval_accuracy": 0.6538196078431373,
+      "eval_loss": 1.4471032619476318,
+      "eval_runtime": 6.7177,
+      "eval_samples_per_second": 74.431,
+      "eval_steps_per_second": 9.378,
       "step": 2437
     },
     {
-      "epoch": 12.997333333333334,
-      "eval_exact_match": 13.8,
-      "eval_f1": 21.856666666666673,
-      "eval_qa_bleu": 10.206087008658956,
-      "eval_qa_exact_match": 0.116,
-      "eval_recite_bleu": 12.669262328097592,
-      "eval_recite_exact_match": 0.008,
+      "epoch": 13.0,
+      "eval_exact_match": 16.4,
+      "eval_f1": 23.343965796845062,
+      "eval_qa_bleu": 4.473837470175788,
+      "eval_qa_exact_match": 0.136,
+      "eval_recite_bleu": 10.58390244263022,
+      "eval_recite_exact_match": 0.002,
       "step": 2437
     },
     {
-      "epoch": 13.333333333333334,
-      "grad_norm": 1.8666125535964966,
+      "epoch": 13.33,
       "learning_rate": 3e-05,
-      "loss": 1.097,
+      "loss": 0.682,
       "step": 2500
     },
     {
-      "epoch": 13.866666666666667,
-      "grad_norm": 1.9091932773590088,
+      "epoch": 13.87,
       "learning_rate": 3e-05,
-      "loss": 1.0793,
+      "loss": 0.6704,
       "step": 2600
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.5960986547085202,
-      "eval_loss": 1.9584920406341553,
-      "eval_runtime": 6.3056,
-      "eval_samples_per_second": 79.294,
-      "eval_steps_per_second": 9.991,
+      "eval_accuracy": 0.6524862745098039,
+      "eval_loss": 1.4829775094985962,
+      "eval_runtime": 6.6339,
+      "eval_samples_per_second": 75.37,
+      "eval_steps_per_second": 9.497,
       "step": 2625
     },
     {
       "epoch": 14.0,
-      "eval_exact_match": 10.6,
-      "eval_f1": 18.697936507936518,
-      "eval_qa_bleu": 6.4581555766931436,
-      "eval_qa_exact_match": 0.082,
-      "eval_recite_bleu": 12.843526576123885,
-      "eval_recite_exact_match": 0.01,
+      "eval_exact_match": 15.6,
+      "eval_f1": 23.615572621393074,
+      "eval_qa_bleu": 7.563435997390864,
+      "eval_qa_exact_match": 0.13,
+      "eval_recite_bleu": 11.120038652898813,
+      "eval_recite_exact_match": 0.008,
       "step": 2625
     },
     {
       "epoch": 14.4,
-      "grad_norm": 2.042712688446045,
       "learning_rate": 3e-05,
-      "loss": 1.0376,
+      "loss": 0.6412,
       "step": 2700
     },
     {
-      "epoch": 14.933333333333334,
-      "grad_norm": 1.8875784873962402,
+      "epoch": 14.93,
       "learning_rate": 3e-05,
-      "loss": 1.037,
+      "loss": 0.6436,
       "step": 2800
     },
     {
-      "epoch": 14.997333333333334,
-      "eval_accuracy": 0.5948161434977578,
-      "eval_loss": 1.9966908693313599,
-      "eval_runtime": 6.0322,
-      "eval_samples_per_second": 82.888,
-      "eval_steps_per_second": 10.444,
+      "epoch": 15.0,
+      "eval_accuracy": 0.6507529411764705,
+      "eval_loss": 1.524308681488037,
+      "eval_runtime": 6.6481,
+      "eval_samples_per_second": 75.209,
+      "eval_steps_per_second": 9.476,
       "step": 2812
     },
     {
-      "epoch": 14.997333333333334,
-      "eval_exact_match": 12.8,
-      "eval_f1": 20.438917748917753,
-      "eval_qa_bleu": 10.19364322169534,
-      "eval_qa_exact_match": 0.104,
-      "eval_recite_bleu": 12.69742056629334,
-      "eval_recite_exact_match": 0.012,
+      "epoch": 15.0,
+      "eval_exact_match": 17.8,
+      "eval_f1": 27.00021164021165,
+      "eval_qa_bleu": 7.90349748901623,
+      "eval_qa_exact_match": 0.15,
+      "eval_recite_bleu": 12.277483381668418,
+      "eval_recite_exact_match": 0.008,
       "step": 2812
     },
     {
-      "epoch": 15.466666666666667,
-      "grad_norm": 2.027426242828369,
+      "epoch": 15.47,
       "learning_rate": 3e-05,
-      "loss": 0.9844,
+      "loss": 0.5969,
       "step": 2900
     },
     {
       "epoch": 16.0,
-      "grad_norm": 2.0708699226379395,
       "learning_rate": 3e-05,
-      "loss": 0.9901,
+      "loss": 0.6073,
       "step": 3000
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.5934080717488789,
-      "eval_loss": 2.0336477756500244,
-      "eval_runtime": 5.9811,
-      "eval_samples_per_second": 83.597,
-      "eval_steps_per_second": 10.533,
+      "eval_accuracy": 0.6499764705882353,
+      "eval_loss": 1.5640995502471924,
+      "eval_runtime": 7.7119,
+      "eval_samples_per_second": 64.835,
+      "eval_steps_per_second": 8.169,
       "step": 3000
     },
     {
       "epoch": 16.0,
-      "eval_exact_match": 12.4,
-      "eval_f1": 20.427142857142865,
-      "eval_qa_bleu": 9.13217752730269,
-      "eval_qa_exact_match": 0.096,
-      "eval_recite_bleu": 12.688250071816965,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 19.2,
+      "eval_f1": 27.747719298245627,
+      "eval_qa_bleu": 9.793788404463436,
+      "eval_qa_exact_match": 0.14,
+      "eval_recite_bleu": 12.835117221544726,
+      "eval_recite_exact_match": 0.01,
       "step": 3000
     },
     {
-      "epoch": 16.533333333333335,
-      "grad_norm": 1.995875358581543,
+      "epoch": 16.53,
       "learning_rate": 3e-05,
-      "loss": 0.9316,
+      "loss": 0.5567,
       "step": 3100
     },
     {
-      "epoch": 16.997333333333334,
-      "eval_accuracy": 0.591372197309417,
-      "eval_loss": 2.088000535964966,
-      "eval_runtime": 5.9631,
-      "eval_samples_per_second": 83.849,
-      "eval_steps_per_second": 10.565,
+      "epoch": 17.0,
+      "eval_accuracy": 0.6486588235294117,
+      "eval_loss": 1.6058622598648071,
+      "eval_runtime": 7.7727,
+      "eval_samples_per_second": 64.328,
+      "eval_steps_per_second": 8.105,
       "step": 3187
     },
     {
-      "epoch": 16.997333333333334,
-      "eval_exact_match": 11.4,
-      "eval_f1": 18.528571428571443,
-      "eval_qa_bleu": 7.115393148941145,
-      "eval_qa_exact_match": 0.084,
-      "eval_recite_bleu": 12.219544702526637,
-      "eval_recite_exact_match": 0.014,
+      "epoch": 17.0,
+      "eval_exact_match": 19.6,
+      "eval_f1": 26.446776949066113,
+      "eval_qa_bleu": 10.294098070146385,
+      "eval_qa_exact_match": 0.156,
+      "eval_recite_bleu": 12.34815708077889,
+      "eval_recite_exact_match": 0.004,
       "step": 3187
     },
     {
-      "epoch": 17.066666666666666,
-      "grad_norm": 2.623262643814087,
+      "epoch": 17.07,
       "learning_rate": 3e-05,
-      "loss": 0.9354,
+      "loss": 0.5654,
       "step": 3200
     },
     {
       "epoch": 17.6,
-      "grad_norm": 2.4125397205352783,
       "learning_rate": 3e-05,
-      "loss": 0.8802,
+      "loss": 0.5204,
       "step": 3300
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.5900627802690583,
-      "eval_loss": 2.1439507007598877,
-      "eval_runtime": 5.8546,
-      "eval_samples_per_second": 85.404,
-      "eval_steps_per_second": 10.761,
+      "eval_accuracy": 0.6472705882352942,
+      "eval_loss": 1.6425700187683105,
+      "eval_runtime": 7.1347,
+      "eval_samples_per_second": 70.08,
+      "eval_steps_per_second": 8.83,
       "step": 3375
     },
     {
       "epoch": 18.0,
-      "eval_exact_match": 10.8,
-      "eval_f1": 17.508095238095244,
-      "eval_qa_bleu": 7.769457934505562,
-      "eval_qa_exact_match": 0.09,
-      "eval_recite_bleu": 12.155079189562132,
-      "eval_recite_exact_match": 0.016,
+      "eval_exact_match": 18.4,
+      "eval_f1": 26.083665223665236,
+      "eval_qa_bleu": 12.978225559731756,
+      "eval_qa_exact_match": 0.142,
+      "eval_recite_bleu": 12.629003802964625,
+      "eval_recite_exact_match": 0.006,
       "step": 3375
     },
     {
-      "epoch": 18.133333333333333,
-      "grad_norm": 2.3584065437316895,
+      "epoch": 18.13,
       "learning_rate": 3e-05,
-      "loss": 0.8913,
+      "loss": 0.5348,
       "step": 3400
     },
     {
-      "epoch": 18.666666666666668,
-      "grad_norm": 2.130357265472412,
+      "epoch": 18.67,
       "learning_rate": 3e-05,
-      "loss": 0.8382,
+      "loss": 0.5019,
       "step": 3500
     },
     {
-      "epoch": 18.997333333333334,
-      "eval_accuracy": 0.5892645739910314,
-      "eval_loss": 2.171532154083252,
-      "eval_runtime": 6.6257,
-      "eval_samples_per_second": 75.464,
-      "eval_steps_per_second": 9.508,
+      "epoch": 19.0,
+      "eval_accuracy": 0.6459843137254901,
+      "eval_loss": 1.6983567476272583,
+      "eval_runtime": 7.0372,
+      "eval_samples_per_second": 71.051,
+      "eval_steps_per_second": 8.952,
       "step": 3562
     },
     {
-      "epoch": 18.997333333333334,
-      "eval_exact_match": 13.0,
-      "eval_f1": 18.838095238095242,
-      "eval_qa_bleu": 7.682885911703284,
-      "eval_qa_exact_match": 0.106,
-      "eval_recite_bleu": 12.350855746597011,
-      "eval_recite_exact_match": 0.014,
+      "epoch": 19.0,
+      "eval_exact_match": 19.4,
+      "eval_f1": 27.658198757763987,
+      "eval_qa_bleu": 10.765170656210177,
+      "eval_qa_exact_match": 0.146,
+      "eval_recite_bleu": 12.874816956011239,
+      "eval_recite_exact_match": 0.006,
       "step": 3562
     },
     {
       "epoch": 19.2,
-      "grad_norm": 2.6397740840911865,
       "learning_rate": 3e-05,
-      "loss": 0.8248,
+      "loss": 0.4888,
       "step": 3600
     },
     {
-      "epoch": 19.733333333333334,
-      "grad_norm": 3.049431085586548,
+      "epoch": 19.73,
       "learning_rate": 3e-05,
-      "loss": 0.7962,
+      "loss": 0.4676,
       "step": 3700
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.5878834080717489,
-      "eval_loss": 2.2236711978912354,
-      "eval_runtime": 6.764,
-      "eval_samples_per_second": 73.921,
-      "eval_steps_per_second": 9.314,
+      "eval_accuracy": 0.6450431372549019,
+      "eval_loss": 1.74748957157135,
+      "eval_runtime": 6.6386,
+      "eval_samples_per_second": 75.317,
+      "eval_steps_per_second": 9.49,
       "step": 3750
     },
     {
       "epoch": 20.0,
-      "eval_exact_match": 12.2,
-      "eval_f1": 18.373809523809534,
-      "eval_qa_bleu": 8.448974659181921,
-      "eval_qa_exact_match": 0.096,
-      "eval_recite_bleu": 11.298504528713615,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 17.4,
+      "eval_f1": 25.5869051909052,
+      "eval_qa_bleu": 7.975401244418332,
+      "eval_qa_exact_match": 0.144,
+      "eval_recite_bleu": 11.978547399136012,
+      "eval_recite_exact_match": 0.004,
       "step": 3750
     },
     {
-      "epoch": 20.266666666666666,
-      "grad_norm": 2.647123098373413,
+      "epoch": 20.27,
       "learning_rate": 3e-05,
-      "loss": 0.7797,
+      "loss": 0.4528,
       "step": 3800
     },
     {
       "epoch": 20.8,
-      "grad_norm": 2.798877239227295,
       "learning_rate": 3e-05,
-      "loss": 0.7553,
+      "loss": 0.4408,
       "step": 3900
     },
     {
-      "epoch": 20.997333333333334,
-      "eval_accuracy": 0.5860896860986547,
-      "eval_loss": 2.295654535293579,
-      "eval_runtime": 5.5847,
-      "eval_samples_per_second": 89.531,
-      "eval_steps_per_second": 11.281,
+      "epoch": 21.0,
+      "eval_accuracy": 0.6450274509803922,
+      "eval_loss": 1.7759969234466553,
+      "eval_runtime": 7.7092,
+      "eval_samples_per_second": 64.857,
+      "eval_steps_per_second": 8.172,
       "step": 3937
     },
     {
-      "epoch": 20.997333333333334,
-      "eval_exact_match": 10.2,
-      "eval_f1": 16.601428571428578,
-      "eval_qa_bleu": 7.90021052694059,
-      "eval_qa_exact_match": 0.08,
-      "eval_recite_bleu": 11.716166611151118,
-      "eval_recite_exact_match": 0.01,
+      "epoch": 21.0,
+      "eval_exact_match": 19.0,
+      "eval_f1": 27.044218949435447,
+      "eval_qa_bleu": 5.961857744132503,
+      "eval_qa_exact_match": 0.158,
+      "eval_recite_bleu": 12.078600703350125,
+      "eval_recite_exact_match": 0.004,
       "step": 3937
     },
     {
-      "epoch": 21.333333333333332,
-      "grad_norm": 2.7208638191223145,
+      "epoch": 21.33,
       "learning_rate": 3e-05,
-      "loss": 0.7306,
+      "loss": 0.4191,
       "step": 4000
     },
     {
-      "epoch": 21.866666666666667,
-      "grad_norm": 2.9887781143188477,
+      "epoch": 21.87,
       "learning_rate": 3e-05,
-      "loss": 0.7238,
+      "loss": 0.4217,
       "step": 4100
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.5850941704035875,
-      "eval_loss": 2.331202507019043,
-      "eval_runtime": 6.6871,
-      "eval_samples_per_second": 74.771,
-      "eval_steps_per_second": 9.421,
+      "eval_accuracy": 0.644078431372549,
+      "eval_loss": 1.7986315488815308,
+      "eval_runtime": 7.6507,
+      "eval_samples_per_second": 65.353,
+      "eval_steps_per_second": 8.235,
       "step": 4125
     },
     {
       "epoch": 22.0,
-      "eval_exact_match": 10.8,
-      "eval_f1": 17.400692640692647,
-      "eval_qa_bleu": 4.77963863209427,
-      "eval_qa_exact_match": 0.084,
-      "eval_recite_bleu": 11.404518212072734,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 18.2,
+      "eval_f1": 25.124761904761915,
+      "eval_qa_bleu": 15.392825669562168,
+      "eval_qa_exact_match": 0.14,
+      "eval_recite_bleu": 11.50341909728548,
+      "eval_recite_exact_match": 0.006,
       "step": 4125
     },
     {
       "epoch": 22.4,
-      "grad_norm": 2.706368923187256,
       "learning_rate": 3e-05,
-      "loss": 0.6813,
+      "loss": 0.393,
       "step": 4200
     },
     {
-      "epoch": 22.933333333333334,
-      "grad_norm": 2.850799798965454,
+      "epoch": 22.93,
       "learning_rate": 3e-05,
-      "loss": 0.676,
+      "loss": 0.3886,
       "step": 4300
     },
     {
-      "epoch": 22.997333333333334,
-      "eval_accuracy": 0.583237668161435,
-      "eval_loss": 2.4042670726776123,
-      "eval_runtime": 6.5747,
-      "eval_samples_per_second": 76.049,
-      "eval_steps_per_second": 9.582,
+      "epoch": 23.0,
+      "eval_accuracy": 0.6439529411764706,
+      "eval_loss": 1.8378911018371582,
+      "eval_runtime": 7.4779,
+      "eval_samples_per_second": 66.864,
+      "eval_steps_per_second": 8.425,
       "step": 4312
     },
     {
-      "epoch": 22.997333333333334,
-      "eval_exact_match": 8.6,
-      "eval_f1": 16.314285714285724,
-      "eval_qa_bleu": 7.466573197185606,
-      "eval_qa_exact_match": 0.064,
-      "eval_recite_bleu": 11.568247151151132,
-      "eval_recite_exact_match": 0.014,
+      "epoch": 23.0,
+      "eval_exact_match": 14.8,
+      "eval_f1": 21.914444444444456,
+      "eval_qa_bleu": 10.979013053975578,
+      "eval_qa_exact_match": 0.128,
+      "eval_recite_bleu": 11.265801595707984,
+      "eval_recite_exact_match": 0.004,
       "step": 4312
     },
     {
-      "epoch": 23.466666666666665,
-      "grad_norm": 2.95532488822937,
+      "epoch": 23.47,
       "learning_rate": 3e-05,
-      "loss": 0.632,
+      "loss": 0.3605,
       "step": 4400
     },
     {
       "epoch": 24.0,
-      "grad_norm": 3.36663556098938,
       "learning_rate": 3e-05,
-      "loss": 0.644,
+      "loss": 0.3729,
       "step": 4500
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.5824125560538117,
-      "eval_loss": 2.443950891494751,
-      "eval_runtime": 5.7069,
-      "eval_samples_per_second": 87.614,
-      "eval_steps_per_second": 11.039,
+      "eval_accuracy": 0.6420470588235294,
+      "eval_loss": 1.927443504333496,
+      "eval_runtime": 7.4721,
+      "eval_samples_per_second": 66.915,
+      "eval_steps_per_second": 8.431,
       "step": 4500
     },
     {
       "epoch": 24.0,
-      "eval_exact_match": 8.8,
-      "eval_f1": 15.491948051948057,
-      "eval_qa_bleu": 4.531789460915551,
-      "eval_qa_exact_match": 0.076,
-      "eval_recite_bleu": 10.65729884465588,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 15.6,
+      "eval_f1": 22.679302325581407,
+      "eval_qa_bleu": 7.33716058981409,
+      "eval_qa_exact_match": 0.132,
+      "eval_recite_bleu": 11.564503717897392,
+      "eval_recite_exact_match": 0.002,
       "step": 4500
     },
     {
-      "epoch": 24.533333333333335,
-      "grad_norm": 3.074622631072998,
+      "epoch": 24.53,
       "learning_rate": 3e-05,
-      "loss": 0.5939,
+      "loss": 0.3316,
       "step": 4600
     },
     {
-      "epoch": 24.997333333333334,
-      "eval_accuracy": 0.5818385650224215,
-      "eval_loss": 2.5127182006835938,
-      "eval_runtime": 5.6588,
-      "eval_samples_per_second": 88.357,
-      "eval_steps_per_second": 11.133,
+      "epoch": 25.0,
+      "eval_accuracy": 0.641921568627451,
+      "eval_loss": 1.9378185272216797,
+      "eval_runtime": 7.6806,
+      "eval_samples_per_second": 65.099,
+      "eval_steps_per_second": 8.203,
       "step": 4687
     },
     {
-      "epoch": 24.997333333333334,
-      "eval_exact_match": 9.0,
-      "eval_f1": 16.361587301587303,
-      "eval_qa_bleu": 7.691728089554307,
-      "eval_qa_exact_match": 0.068,
-      "eval_recite_bleu": 11.016738678559781,
-      "eval_recite_exact_match": 0.008,
+      "epoch": 25.0,
+      "eval_exact_match": 16.8,
+      "eval_f1": 23.502857142857145,
+      "eval_qa_bleu": 8.430315553769542,
+      "eval_qa_exact_match": 0.13,
+      "eval_recite_bleu": 11.821782294825612,
+      "eval_recite_exact_match": 0.006,
       "step": 4687
     },
     {
-      "epoch": 25.066666666666666,
-      "grad_norm": 2.702502727508545,
+      "epoch": 25.07,
       "learning_rate": 3e-05,
-      "loss": 0.5904,
+      "loss": 0.3454,
       "step": 4700
     },
     {
       "epoch": 25.6,
-      "grad_norm": 3.233320951461792,
       "learning_rate": 3e-05,
-      "loss": 0.5551,
+      "loss": 0.313,
       "step": 4800
     },
     {
       "epoch": 26.0,
-      "eval_accuracy": 0.5809596412556054,
-      "eval_loss": 2.538985252380371,
-      "eval_runtime": 6.6838,
-      "eval_samples_per_second": 74.808,
-      "eval_steps_per_second": 9.426,
+      "eval_accuracy": 0.6410274509803922,
+      "eval_loss": 2.0155365467071533,
+      "eval_runtime": 6.6459,
+      "eval_samples_per_second": 75.235,
+      "eval_steps_per_second": 9.48,
       "step": 4875
     },
     {
       "epoch": 26.0,
-      "eval_exact_match": 9.2,
-      "eval_f1": 16.819884559884564,
-      "eval_qa_bleu": 4.658892987257927,
-      "eval_qa_exact_match": 0.078,
-      "eval_recite_bleu": 10.853797139069188,
-      "eval_recite_exact_match": 0.012,
+      "eval_exact_match": 15.4,
+      "eval_f1": 23.056190476190483,
+      "eval_qa_bleu": 10.828743073480153,
+      "eval_qa_exact_match": 0.132,
+      "eval_recite_bleu": 11.70611308380729,
+      "eval_recite_exact_match": 0.004,
       "step": 4875
     },
     {
-      "epoch": 26.133333333333333,
-      "grad_norm": 3.280097246170044,
+      "epoch": 26.13,
       "learning_rate": 3e-05,
-      "loss": 0.5503,
+      "loss": 0.316,
       "step": 4900
     },
     {
-      "epoch": 26.666666666666668,
-      "grad_norm": 3.1984827518463135,
+      "epoch": 26.67,
       "learning_rate": 3e-05,
-      "loss": 0.5163,
+      "loss": 0.2891,
       "step": 5000
     },
     {
-      "epoch": 26.997333333333334,
-      "eval_accuracy": 0.579829596412556,
-      "eval_loss": 2.5808944702148438,
-      "eval_runtime": 6.6339,
-      "eval_samples_per_second": 75.37,
-      "eval_steps_per_second": 9.497,
+      "epoch": 27.0,
+      "eval_accuracy": 0.6408313725490196,
+      "eval_loss": 2.0192761421203613,
+      "eval_runtime": 7.4551,
+      "eval_samples_per_second": 67.068,
+      "eval_steps_per_second": 8.451,
       "step": 5062
     },
     {
-      "epoch": 26.997333333333334,
-      "eval_exact_match": 8.0,
-      "eval_f1": 16.641197691197693,
-      "eval_qa_bleu": 5.575351494624609,
-      "eval_qa_exact_match": 0.066,
-      "eval_recite_bleu": 11.099361911324253,
-      "eval_recite_exact_match": 0.01,
+      "epoch": 27.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 22.766833842561553,
+      "eval_qa_bleu": 3.5222468357048804,
+      "eval_qa_exact_match": 0.108,
+      "eval_recite_bleu": 10.474195118057567,
+      "eval_recite_exact_match": 0.002,
       "step": 5062
     },
     {
       "epoch": 27.2,
-      "grad_norm": 3.5599937438964844,
       "learning_rate": 3e-05,
-      "loss": 0.52,
+      "loss": 0.299,
       "step": 5100
     },
     {
-      "epoch": 27.733333333333334,
-      "grad_norm": 2.9382145404815674,
+      "epoch": 27.73,
       "learning_rate": 3e-05,
-      "loss": 0.4892,
+      "loss": 0.281,
       "step": 5200
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.5788609865470852,
-      "eval_loss": 2.6670191287994385,
-      "eval_runtime": 6.0585,
-      "eval_samples_per_second": 82.529,
-      "eval_steps_per_second": 10.399,
+      "eval_accuracy": 0.640243137254902,
+      "eval_loss": 2.074819803237915,
+      "eval_runtime": 7.0598,
+      "eval_samples_per_second": 70.823,
+      "eval_steps_per_second": 8.924,
       "step": 5250
     },
     {
       "epoch": 28.0,
-      "eval_exact_match": 8.6,
-      "eval_f1": 15.144920634920638,
-      "eval_qa_bleu": 5.387088284739065,
-      "eval_qa_exact_match": 0.074,
-      "eval_recite_bleu": 11.457740165981269,
-      "eval_recite_exact_match": 0.012,
+      "eval_exact_match": 14.8,
+      "eval_f1": 22.03527508090615,
+      "eval_qa_bleu": 4.8943988922035935,
+      "eval_qa_exact_match": 0.118,
+      "eval_recite_bleu": 11.435320168344504,
+      "eval_recite_exact_match": 0.008,
       "step": 5250
     },
     {
-      "epoch": 28.266666666666666,
-      "grad_norm": 4.097602367401123,
+      "epoch": 28.27,
       "learning_rate": 3e-05,
-      "loss": 0.47,
+      "loss": 0.2637,
       "step": 5300
     },
     {
       "epoch": 28.8,
-      "grad_norm": 3.9295685291290283,
       "learning_rate": 3e-05,
-      "loss": 0.4669,
+      "loss": 0.2639,
       "step": 5400
     },
     {
-      "epoch": 28.997333333333334,
-      "eval_accuracy": 0.578609865470852,
-      "eval_loss": 2.669480085372925,
-      "eval_runtime": 6.6992,
-      "eval_samples_per_second": 74.636,
-      "eval_steps_per_second": 9.404,
+      "epoch": 29.0,
+      "eval_accuracy": 0.6389333333333334,
+      "eval_loss": 2.1383912563323975,
+      "eval_runtime": 7.7452,
+      "eval_samples_per_second": 64.556,
+      "eval_steps_per_second": 8.134,
       "step": 5437
     },
     {
-      "epoch": 28.997333333333334,
-      "eval_exact_match": 9.2,
-      "eval_f1": 16.65718614718615,
-      "eval_qa_bleu": 4.637892076000507,
-      "eval_qa_exact_match": 0.074,
-      "eval_recite_bleu": 11.77754451628722,
-      "eval_recite_exact_match": 0.012,
+      "epoch": 29.0,
+      "eval_exact_match": 16.2,
+      "eval_f1": 24.30754076973256,
+      "eval_qa_bleu": 4.486401752816634,
+      "eval_qa_exact_match": 0.128,
+      "eval_recite_bleu": 11.730902648782326,
+      "eval_recite_exact_match": 0.004,
       "step": 5437
     },
     {
-      "epoch": 29.333333333333332,
-      "grad_norm": 3.4413976669311523,
+      "epoch": 29.33,
       "learning_rate": 3e-05,
-      "loss": 0.4398,
+      "loss": 0.2473,
       "step": 5500
     },
     {
-      "epoch": 29.866666666666667,
-      "grad_norm": 2.88114857673645,
+      "epoch": 29.87,
       "learning_rate": 3e-05,
-      "loss": 0.4353,
+      "loss": 0.249,
       "step": 5600
     },
     {
       "epoch": 30.0,
-      "eval_accuracy": 0.5786995515695067,
-      "eval_loss": 2.7645530700683594,
-      "eval_runtime": 6.3214,
-      "eval_samples_per_second": 79.096,
-      "eval_steps_per_second": 9.966,
+      "eval_accuracy": 0.6398509803921568,
+      "eval_loss": 2.125810384750366,
+      "eval_runtime": 7.6954,
+      "eval_samples_per_second": 64.974,
+      "eval_steps_per_second": 8.187,
       "step": 5625
     },
     {
       "epoch": 30.0,
-      "eval_exact_match": 8.0,
-      "eval_f1": 14.828455988455993,
-      "eval_qa_bleu": 7.042137393349148,
-      "eval_qa_exact_match": 0.068,
-      "eval_recite_bleu": 11.413791110087148,
-      "eval_recite_exact_match": 0.01,
+      "eval_exact_match": 15.8,
+      "eval_f1": 22.123535353535363,
+      "eval_qa_bleu": 3.2440134844965627,
+      "eval_qa_exact_match": 0.126,
+      "eval_recite_bleu": 11.76325197849696,
+      "eval_recite_exact_match": 0.004,
       "step": 5625
     },
     {
       "epoch": 30.4,
-      "grad_norm": 4.096495628356934,
       "learning_rate": 3e-05,
-      "loss": 0.4032,
+      "loss": 0.2263,
       "step": 5700
     },
     {
-      "epoch": 30.933333333333334,
-      "grad_norm": 3.567011833190918,
+      "epoch": 30.93,
       "learning_rate": 3e-05,
-      "loss": 0.4104,
+      "loss": 0.236,
       "step": 5800
     },
     {
-      "epoch": 30.997333333333334,
-      "eval_accuracy": 0.5774887892376681,
-      "eval_loss": 2.829072952270508,
-      "eval_runtime": 6.655,
-      "eval_samples_per_second": 75.131,
-      "eval_steps_per_second": 9.467,
+      "epoch": 31.0,
+      "eval_accuracy": 0.6378666666666667,
+      "eval_loss": 2.2261879444122314,
+      "eval_runtime": 7.6462,
+      "eval_samples_per_second": 65.392,
+      "eval_steps_per_second": 8.239,
       "step": 5812
     },
     {
-      "epoch": 30.997333333333334,
-      "eval_exact_match": 8.8,
-      "eval_f1": 14.71683982683983,
-      "eval_qa_bleu": 7.548806097282496,
-      "eval_qa_exact_match": 0.072,
-      "eval_recite_bleu": 10.549616368516189,
-      "eval_recite_exact_match": 0.01,
+      "epoch": 31.0,
+      "eval_exact_match": 17.6,
+      "eval_f1": 25.120476190476193,
+      "eval_qa_bleu": 5.240120013213656,
+      "eval_qa_exact_match": 0.13,
+      "eval_recite_bleu": 11.888730171325669,
+      "eval_recite_exact_match": 0.008,
       "step": 5812
     },
     {
-      "epoch": 31.466666666666665,
-      "grad_norm": 3.1184980869293213,
+      "epoch": 31.47,
       "learning_rate": 3e-05,
-      "loss": 0.3695,
+      "loss": 0.2062,
       "step": 5900
     },
     {
       "epoch": 32.0,
-      "grad_norm": 3.125758647918701,
       "learning_rate": 3e-05,
-      "loss": 0.3885,
+      "loss": 0.2235,
       "step": 6000
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.5763587443946189,
-      "eval_loss": 2.893293857574463,
-      "eval_runtime": 5.9693,
-      "eval_samples_per_second": 83.762,
-      "eval_steps_per_second": 10.554,
+      "eval_accuracy": 0.6381254901960784,
+      "eval_loss": 2.2489171028137207,
+      "eval_runtime": 7.0486,
+      "eval_samples_per_second": 70.936,
+      "eval_steps_per_second": 8.938,
       "step": 6000
     },
     {
       "epoch": 32.0,
-      "eval_exact_match": 8.6,
-      "eval_f1": 15.2548051948052,
-      "eval_qa_bleu": 5.915179328505316,
-      "eval_qa_exact_match": 0.072,
-      "eval_recite_bleu": 10.647049650020403,
+      "eval_exact_match": 16.6,
+      "eval_f1": 25.16174603174605,
+      "eval_qa_bleu": 12.336439673952102,
+      "eval_qa_exact_match": 0.134,
+      "eval_recite_bleu": 11.520322065980684,
       "eval_recite_exact_match": 0.008,
       "step": 6000
     }
   ],
   "logging_steps": 100,
   "max_steps": 9350,
-  "num_input_tokens_seen": 0,
   "num_train_epochs": 50,
   "save_steps": 500,
-  "stateful_callbacks": {
-    "TrainerControl": {
-      "args": {
-        "should_epoch_stop": false,
-        "should_evaluate": false,
-        "should_log": false,
-        "should_save": true,
-        "should_training_stop": false
-      },
-      "attributes": {}
-    }
-  },
-  "total_flos": 6.620940376159027e+17,
-  "train_batch_size": 2,
+  "total_flos": 1.3906918229000847e+18,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-6000/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e9c21c8367f69c99b21b20aa3dbf722814d1b222850e23dacbe471b235b2dd4
-size 5304
+oid sha256:9347580f7d2a7b3957c29c0f67ee2342d50aa99cfc14c2f59eba6127909f6932
+size 4728