Training in progress, epoch 24, checkpoint

checkpoint-4500/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-4500/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4191dd9908988e20077b302bfe705bfd349730890fd190946ac99842c98f1a2d
+size 143269386

checkpoint-4500/added_tokens.json

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

checkpoint-4500/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0d8e33cd85fb7022bb4dd5003eeeab717a0b2be59f19077e7dca148c0d1ee86
-size 224537202
+oid sha256:baa866aba5c4e6aa0b33b4acd861fef0e7cf7960fcd709d61272ebf4780c88a6
+size 286585234

checkpoint-4500/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bc1e9c221b2fe6245e04d8caca3b3c47357882aa34b51f101fd13d21acc3f75d
+oid sha256:87fd405a3e89daad30368a84660af9f3db91c5524c2460c0db95322eff31e04b
 size 15024

checkpoint-4500/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:97c3e45302caea18b01a06b44f236a7ab0166383be7f953e3b7ec8ba4e286a65
+oid sha256:ddc0d8775e3c478bdf4dcaf784a9d54059e7d8dc17cec5cdd735e264d0d1e3ab
 size 15024

checkpoint-4500/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:755161384c6d2a51c3c48252a5a4964b5ceaa321bc4bf0336f515e917590410b
+oid sha256:d4bc9328491e8608a26f3c6ec2f33c7cd57a3e4868f04403a76e7b194cb9d69c
 size 15024

checkpoint-4500/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30ea4c12ebc01b961a5a43509b429f6b6f6f28f53e47a24f7e0f1221c808f2d6
+oid sha256:02e78ae98aa77297e09bb303cca045fc3cea48001913aced831ed212ac619dfd
 size 15024

checkpoint-4500/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-4500/tokenizer.model

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

checkpoint-4500/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-4500/trainer_state.json

@@ -9,796 +9,737 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.5333333333333333,
-      "grad_norm": 0.271308958530426,
+      "epoch": 0.53,
       "learning_rate": 5e-05,
-      "loss": 1.8262,
+      "loss": 1.3988,
       "step": 100
     },
     {
-      "epoch": 0.9973333333333333,
-      "eval_accuracy": 0.6047174887892377,
-      "eval_loss": 1.7027028799057007,
-      "eval_runtime": 6.6338,
-      "eval_samples_per_second": 75.372,
-      "eval_steps_per_second": 9.497,
+      "epoch": 1.0,
+      "eval_accuracy": 0.6660784313725491,
+      "eval_loss": 1.2107256650924683,
+      "eval_runtime": 7.728,
+      "eval_samples_per_second": 64.7,
+      "eval_steps_per_second": 8.152,
       "step": 187
     },
     {
-      "epoch": 0.9973333333333333,
-      "eval_exact_match": 14.6,
-      "eval_f1": 23.665714285714298,
-      "eval_qa_bleu": 10.24028390049623,
-      "eval_qa_exact_match": 0.114,
-      "eval_recite_bleu": 12.525534079250063,
+      "epoch": 1.0,
+      "eval_exact_match": 19.4,
+      "eval_f1": 27.911959489312448,
+      "eval_qa_bleu": 2.842200543476061,
+      "eval_qa_exact_match": 0.152,
+      "eval_recite_bleu": 12.149078895446083,
       "eval_recite_exact_match": 0.0,
       "step": 187
     },
     {
-      "epoch": 1.0666666666666667,
-      "grad_norm": 0.301651269197464,
+      "epoch": 1.07,
       "learning_rate": 5e-05,
-      "loss": 1.7424,
+      "loss": 1.238,
       "step": 200
     },
     {
       "epoch": 1.6,
-      "grad_norm": 0.35949045419692993,
       "learning_rate": 5e-05,
-      "loss": 1.6978,
+      "loss": 1.1977,
       "step": 300
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.6068699551569506,
-      "eval_loss": 1.6845285892486572,
-      "eval_runtime": 5.7524,
-      "eval_samples_per_second": 86.92,
-      "eval_steps_per_second": 10.952,
+      "eval_accuracy": 0.6676078431372549,
+      "eval_loss": 1.1987448930740356,
+      "eval_runtime": 7.4628,
+      "eval_samples_per_second": 66.999,
+      "eval_steps_per_second": 8.442,
       "step": 375
     },
     {
       "epoch": 2.0,
-      "eval_exact_match": 14.6,
-      "eval_f1": 23.74190476190477,
-      "eval_qa_bleu": 9.70432698897927,
-      "eval_qa_exact_match": 0.114,
-      "eval_recite_bleu": 13.014822566317982,
-      "eval_recite_exact_match": 0.006,
+      "eval_exact_match": 21.4,
+      "eval_f1": 27.979969946667268,
+      "eval_qa_bleu": 3.4273838915653356,
+      "eval_qa_exact_match": 0.158,
+      "eval_recite_bleu": 11.474877690710539,
+      "eval_recite_exact_match": 0.0,
       "step": 375
     },
     {
-      "epoch": 2.1333333333333333,
-      "grad_norm": 0.41301965713500977,
+      "epoch": 2.13,
       "learning_rate": 5e-05,
-      "loss": 1.694,
+      "loss": 1.1628,
       "step": 400
     },
     {
-      "epoch": 2.6666666666666665,
-      "grad_norm": 0.5057485699653625,
+      "epoch": 2.67,
       "learning_rate": 5e-05,
-      "loss": 1.6324,
+      "loss": 1.1458,
       "step": 500
     },
     {
-      "epoch": 2.997333333333333,
-      "eval_accuracy": 0.6079551569506726,
-      "eval_loss": 1.6851012706756592,
-      "eval_runtime": 5.5314,
-      "eval_samples_per_second": 90.394,
-      "eval_steps_per_second": 11.39,
+      "epoch": 3.0,
+      "eval_accuracy": 0.6675764705882353,
+      "eval_loss": 1.1964411735534668,
+      "eval_runtime": 6.6677,
+      "eval_samples_per_second": 74.988,
+      "eval_steps_per_second": 9.448,
       "step": 562
     },
     {
-      "epoch": 2.997333333333333,
-      "eval_exact_match": 14.2,
-      "eval_f1": 23.29190476190477,
-      "eval_qa_bleu": 10.724307586340888,
-      "eval_qa_exact_match": 0.104,
-      "eval_recite_bleu": 12.83531066630853,
-      "eval_recite_exact_match": 0.004,
+      "epoch": 3.0,
+      "eval_exact_match": 23.4,
+      "eval_f1": 31.525761083856903,
+      "eval_qa_bleu": 2.0606924929592334,
+      "eval_qa_exact_match": 0.17,
+      "eval_recite_bleu": 12.027903845545522,
+      "eval_recite_exact_match": 0.0,
       "step": 562
     },
     {
       "epoch": 3.2,
-      "grad_norm": 0.6363334655761719,
       "learning_rate": 5e-05,
-      "loss": 1.6063,
+      "loss": 1.1179,
       "step": 600
     },
     {
-      "epoch": 3.7333333333333334,
-      "grad_norm": 0.6466453671455383,
+      "epoch": 3.73,
       "learning_rate": 5e-05,
-      "loss": 1.5562,
+      "loss": 1.0707,
       "step": 700
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.6071121076233184,
-      "eval_loss": 1.6969172954559326,
-      "eval_runtime": 6.2612,
-      "eval_samples_per_second": 79.857,
-      "eval_steps_per_second": 10.062,
+      "eval_accuracy": 0.6665960784313726,
+      "eval_loss": 1.208784580230713,
+      "eval_runtime": 6.6546,
+      "eval_samples_per_second": 75.136,
+      "eval_steps_per_second": 9.467,
       "step": 750
     },
     {
       "epoch": 4.0,
-      "eval_exact_match": 14.0,
-      "eval_f1": 21.34333333333334,
-      "eval_qa_bleu": 10.250340106635626,
-      "eval_qa_exact_match": 0.084,
-      "eval_recite_bleu": 12.741830585824024,
-      "eval_recite_exact_match": 0.004,
+      "eval_exact_match": 23.4,
+      "eval_f1": 33.18893657873498,
+      "eval_qa_bleu": 3.8183981024520506,
+      "eval_qa_exact_match": 0.176,
+      "eval_recite_bleu": 14.071857337313705,
+      "eval_recite_exact_match": 0.0,
       "step": 750
     },
     {
-      "epoch": 4.266666666666667,
-      "grad_norm": 0.8162885904312134,
+      "epoch": 4.27,
       "learning_rate": 5e-05,
-      "loss": 1.5052,
+      "loss": 1.0179,
       "step": 800
     },
     {
       "epoch": 4.8,
-      "grad_norm": 0.8799439072608948,
       "learning_rate": 5e-05,
-      "loss": 1.469,
+      "loss": 1.0066,
       "step": 900
     },
     {
-      "epoch": 4.997333333333334,
-      "eval_accuracy": 0.6059282511210762,
-      "eval_loss": 1.7216957807540894,
-      "eval_runtime": 6.3012,
-      "eval_samples_per_second": 79.35,
-      "eval_steps_per_second": 9.998,
+      "epoch": 5.0,
+      "eval_accuracy": 0.6651764705882353,
+      "eval_loss": 1.2342489957809448,
+      "eval_runtime": 6.699,
+      "eval_samples_per_second": 74.638,
+      "eval_steps_per_second": 9.404,
       "step": 937
     },
     {
-      "epoch": 4.997333333333334,
-      "eval_exact_match": 12.6,
-      "eval_f1": 21.041428571428575,
-      "eval_qa_bleu": 8.239984727774104,
-      "eval_qa_exact_match": 0.094,
-      "eval_recite_bleu": 12.327373697720192,
-      "eval_recite_exact_match": 0.002,
+      "epoch": 5.0,
+      "eval_exact_match": 24.0,
+      "eval_f1": 32.88692081962977,
+      "eval_qa_bleu": 3.6747877275980905,
+      "eval_qa_exact_match": 0.174,
+      "eval_recite_bleu": 14.36158051041567,
+      "eval_recite_exact_match": 0.006,
       "step": 937
     },
     {
-      "epoch": 5.333333333333333,
-      "grad_norm": 0.9834449291229248,
+      "epoch": 5.33,
       "learning_rate": 5e-05,
-      "loss": 1.3845,
+      "loss": 0.9595,
       "step": 1000
     },
     {
-      "epoch": 5.866666666666667,
-      "grad_norm": 0.9377098679542542,
+      "epoch": 5.87,
       "learning_rate": 5e-05,
-      "loss": 1.3927,
+      "loss": 0.9353,
       "step": 1100
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.6047354260089686,
-      "eval_loss": 1.7564114332199097,
-      "eval_runtime": 6.3447,
-      "eval_samples_per_second": 78.806,
-      "eval_steps_per_second": 9.93,
+      "eval_accuracy": 0.6628392156862745,
+      "eval_loss": 1.2645816802978516,
+      "eval_runtime": 7.4539,
+      "eval_samples_per_second": 67.079,
+      "eval_steps_per_second": 8.452,
       "step": 1125
     },
     {
       "epoch": 6.0,
-      "eval_exact_match": 13.4,
-      "eval_f1": 21.70317460317461,
-      "eval_qa_bleu": 10.522554916579404,
-      "eval_qa_exact_match": 0.104,
-      "eval_recite_bleu": 12.383604385812577,
-      "eval_recite_exact_match": 0.006,
+      "eval_exact_match": 23.4,
+      "eval_f1": 33.50380952380953,
+      "eval_qa_bleu": 13.968612067397348,
+      "eval_qa_exact_match": 0.168,
+      "eval_recite_bleu": 15.026821124315997,
+      "eval_recite_exact_match": 0.012,
       "step": 1125
     },
     {
       "epoch": 6.4,
-      "grad_norm": 1.1172072887420654,
       "learning_rate": 5e-05,
-      "loss": 1.2924,
+      "loss": 0.8774,
       "step": 1200
     },
     {
-      "epoch": 6.933333333333334,
-      "grad_norm": 1.0980473756790161,
+      "epoch": 6.93,
       "learning_rate": 5e-05,
-      "loss": 1.286,
+      "loss": 0.8629,
       "step": 1300
     },
     {
-      "epoch": 6.997333333333334,
-      "eval_accuracy": 0.6026995515695067,
-      "eval_loss": 1.7977122068405151,
-      "eval_runtime": 5.9021,
-      "eval_samples_per_second": 84.716,
-      "eval_steps_per_second": 10.674,
+      "epoch": 7.0,
+      "eval_accuracy": 0.6617333333333333,
+      "eval_loss": 1.2990769147872925,
+      "eval_runtime": 6.6963,
+      "eval_samples_per_second": 74.668,
+      "eval_steps_per_second": 9.408,
       "step": 1312
     },
     {
-      "epoch": 6.997333333333334,
-      "eval_exact_match": 12.0,
-      "eval_f1": 20.07142857142858,
-      "eval_qa_bleu": 9.89685444999411,
-      "eval_qa_exact_match": 0.102,
-      "eval_recite_bleu": 12.555118604776576,
-      "eval_recite_exact_match": 0.008,
+      "epoch": 7.0,
+      "eval_exact_match": 22.0,
+      "eval_f1": 31.8904761904762,
+      "eval_qa_bleu": 15.632357578213835,
+      "eval_qa_exact_match": 0.166,
+      "eval_recite_bleu": 14.318212267740977,
+      "eval_recite_exact_match": 0.016,
       "step": 1312
     },
     {
-      "epoch": 7.466666666666667,
-      "grad_norm": 1.1534159183502197,
+      "epoch": 7.47,
       "learning_rate": 5e-05,
-      "loss": 1.1956,
+      "loss": 0.8,
       "step": 1400
     },
     {
       "epoch": 8.0,
-      "grad_norm": 1.1209707260131836,
       "learning_rate": 5e-05,
-      "loss": 1.1891,
+      "loss": 0.7933,
       "step": 1500
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.6010403587443947,
-      "eval_loss": 1.840914249420166,
-      "eval_runtime": 6.3016,
-      "eval_samples_per_second": 79.345,
-      "eval_steps_per_second": 9.997,
+      "eval_accuracy": 0.659607843137255,
+      "eval_loss": 1.3466852903366089,
+      "eval_runtime": 6.6563,
+      "eval_samples_per_second": 75.117,
+      "eval_steps_per_second": 9.465,
       "step": 1500
     },
     {
       "epoch": 8.0,
-      "eval_exact_match": 12.8,
-      "eval_f1": 21.598730158730167,
-      "eval_qa_bleu": 9.863863368837224,
-      "eval_qa_exact_match": 0.104,
-      "eval_recite_bleu": 12.09730367063489,
-      "eval_recite_exact_match": 0.01,
+      "eval_exact_match": 22.6,
+      "eval_f1": 31.16493506493508,
+      "eval_qa_bleu": 21.45561748665761,
+      "eval_qa_exact_match": 0.17,
+      "eval_recite_bleu": 14.23612921374107,
+      "eval_recite_exact_match": 0.012,
       "step": 1500
     },
     {
-      "epoch": 8.533333333333333,
-      "grad_norm": 1.3646217584609985,
+      "epoch": 8.53,
       "learning_rate": 5e-05,
-      "loss": 1.0861,
+      "loss": 0.7185,
       "step": 1600
     },
     {
-      "epoch": 8.997333333333334,
-      "eval_accuracy": 0.5981255605381166,
-      "eval_loss": 1.9223005771636963,
-      "eval_runtime": 5.5092,
-      "eval_samples_per_second": 90.757,
-      "eval_steps_per_second": 11.435,
+      "epoch": 9.0,
+      "eval_accuracy": 0.6574980392156863,
+      "eval_loss": 1.391683578491211,
+      "eval_runtime": 7.7056,
+      "eval_samples_per_second": 64.888,
+      "eval_steps_per_second": 8.176,
       "step": 1687
     },
     {
-      "epoch": 8.997333333333334,
-      "eval_exact_match": 14.8,
-      "eval_f1": 22.160000000000004,
-      "eval_qa_bleu": 8.357776258354589,
-      "eval_qa_exact_match": 0.116,
-      "eval_recite_bleu": 12.76267532884902,
-      "eval_recite_exact_match": 0.01,
+      "epoch": 9.0,
+      "eval_exact_match": 21.8,
+      "eval_f1": 31.150317460317464,
+      "eval_qa_bleu": 18.411626756160196,
+      "eval_qa_exact_match": 0.17,
+      "eval_recite_bleu": 15.144735177405446,
+      "eval_recite_exact_match": 0.014,
       "step": 1687
     },
     {
-      "epoch": 9.066666666666666,
-      "grad_norm": 1.6948473453521729,
+      "epoch": 9.07,
       "learning_rate": 5e-05,
-      "loss": 1.0935,
+      "loss": 0.7176,
       "step": 1700
     },
     {
       "epoch": 9.6,
-      "grad_norm": 1.6024364233016968,
       "learning_rate": 5e-05,
-      "loss": 1.0127,
+      "loss": 0.6489,
       "step": 1800
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.5973632286995516,
-      "eval_loss": 1.9414385557174683,
-      "eval_runtime": 5.5824,
-      "eval_samples_per_second": 89.568,
-      "eval_steps_per_second": 11.286,
+      "eval_accuracy": 0.6549647058823529,
+      "eval_loss": 1.4609253406524658,
+      "eval_runtime": 6.6599,
+      "eval_samples_per_second": 75.076,
+      "eval_steps_per_second": 9.46,
       "step": 1875
     },
     {
       "epoch": 10.0,
-      "eval_exact_match": 12.4,
-      "eval_f1": 20.75746031746032,
-      "eval_qa_bleu": 8.760424886550503,
-      "eval_qa_exact_match": 0.088,
-      "eval_recite_bleu": 12.754935192029142,
+      "eval_exact_match": 24.8,
+      "eval_f1": 32.80031746031747,
+      "eval_qa_bleu": 19.203751437503612,
+      "eval_qa_exact_match": 0.204,
+      "eval_recite_bleu": 14.497298759377461,
       "eval_recite_exact_match": 0.016,
       "step": 1875
     },
     {
-      "epoch": 10.133333333333333,
-      "grad_norm": 1.6970032453536987,
+      "epoch": 10.13,
       "learning_rate": 5e-05,
-      "loss": 0.9788,
+      "loss": 0.6426,
       "step": 1900
     },
     {
-      "epoch": 10.666666666666666,
-      "grad_norm": 1.559332251548767,
+      "epoch": 10.67,
       "learning_rate": 5e-05,
-      "loss": 0.926,
+      "loss": 0.586,
       "step": 2000
     },
     {
-      "epoch": 10.997333333333334,
-      "eval_accuracy": 0.5945201793721973,
-      "eval_loss": 2.0455663204193115,
-      "eval_runtime": 5.5259,
-      "eval_samples_per_second": 90.484,
-      "eval_steps_per_second": 11.401,
+      "epoch": 11.0,
+      "eval_accuracy": 0.6531372549019608,
+      "eval_loss": 1.5205447673797607,
+      "eval_runtime": 6.6744,
+      "eval_samples_per_second": 74.913,
+      "eval_steps_per_second": 9.439,
       "step": 2062
     },
     {
-      "epoch": 10.997333333333334,
-      "eval_exact_match": 13.0,
-      "eval_f1": 20.56190476190476,
-      "eval_qa_bleu": 11.437402190353293,
-      "eval_qa_exact_match": 0.106,
-      "eval_recite_bleu": 13.600786696850818,
-      "eval_recite_exact_match": 0.022,
+      "epoch": 11.0,
+      "eval_exact_match": 24.4,
+      "eval_f1": 32.14761904761905,
+      "eval_qa_bleu": 17.66314588893528,
+      "eval_qa_exact_match": 0.198,
+      "eval_recite_bleu": 14.285546168231928,
+      "eval_recite_exact_match": 0.016,
       "step": 2062
     },
     {
       "epoch": 11.2,
-      "grad_norm": 1.667855143547058,
       "learning_rate": 5e-05,
-      "loss": 0.8872,
+      "loss": 0.5625,
       "step": 2100
     },
     {
-      "epoch": 11.733333333333333,
-      "grad_norm": 1.6040037870407104,
+      "epoch": 11.73,
       "learning_rate": 5e-05,
-      "loss": 0.8347,
+      "loss": 0.5267,
       "step": 2200
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.5923946188340807,
-      "eval_loss": 2.116677761077881,
-      "eval_runtime": 6.4038,
-      "eval_samples_per_second": 78.078,
-      "eval_steps_per_second": 9.838,
+      "eval_accuracy": 0.6518117647058823,
+      "eval_loss": 1.5615981817245483,
+      "eval_runtime": 7.0623,
+      "eval_samples_per_second": 70.799,
+      "eval_steps_per_second": 8.921,
       "step": 2250
     },
     {
       "epoch": 12.0,
-      "eval_exact_match": 13.6,
-      "eval_f1": 20.470981240981256,
-      "eval_qa_bleu": 8.968260765274712,
-      "eval_qa_exact_match": 0.106,
-      "eval_recite_bleu": 12.793280320213494,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 22.0,
+      "eval_f1": 29.764505494505517,
+      "eval_qa_bleu": 14.595863052742681,
+      "eval_qa_exact_match": 0.174,
+      "eval_recite_bleu": 15.269859538584459,
+      "eval_recite_exact_match": 0.02,
       "step": 2250
     },
     {
-      "epoch": 12.266666666666667,
-      "grad_norm": 1.794092059135437,
+      "epoch": 12.27,
       "learning_rate": 5e-05,
-      "loss": 0.8032,
+      "loss": 0.5074,
       "step": 2300
     },
     {
       "epoch": 12.8,
-      "grad_norm": 1.8774526119232178,
       "learning_rate": 5e-05,
-      "loss": 0.7736,
+      "loss": 0.4702,
       "step": 2400
     },
     {
-      "epoch": 12.997333333333334,
-      "eval_accuracy": 0.5902690582959641,
-      "eval_loss": 2.179492950439453,
-      "eval_runtime": 5.9945,
-      "eval_samples_per_second": 83.409,
-      "eval_steps_per_second": 10.51,
+      "epoch": 13.0,
+      "eval_accuracy": 0.6498588235294117,
+      "eval_loss": 1.6518086194992065,
+      "eval_runtime": 7.4952,
+      "eval_samples_per_second": 66.71,
+      "eval_steps_per_second": 8.405,
       "step": 2437
     },
     {
-      "epoch": 12.997333333333334,
-      "eval_exact_match": 12.8,
-      "eval_f1": 20.780476190476197,
-      "eval_qa_bleu": 11.176823944321667,
-      "eval_qa_exact_match": 0.094,
-      "eval_recite_bleu": 13.292972390103905,
-      "eval_recite_exact_match": 0.02,
+      "epoch": 13.0,
+      "eval_exact_match": 19.8,
+      "eval_f1": 29.199523809523825,
+      "eval_qa_bleu": 15.26272043877601,
+      "eval_qa_exact_match": 0.156,
+      "eval_recite_bleu": 13.712487638236569,
+      "eval_recite_exact_match": 0.012,
       "step": 2437
     },
     {
-      "epoch": 13.333333333333334,
-      "grad_norm": 1.9102168083190918,
+      "epoch": 13.33,
       "learning_rate": 5e-05,
-      "loss": 0.7168,
+      "loss": 0.4335,
       "step": 2500
     },
     {
-      "epoch": 13.866666666666667,
-      "grad_norm": 1.9183365106582642,
+      "epoch": 13.87,
       "learning_rate": 5e-05,
-      "loss": 0.6903,
+      "loss": 0.42,
       "step": 2600
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.5875874439461883,
-      "eval_loss": 2.274404287338257,
-      "eval_runtime": 6.2502,
-      "eval_samples_per_second": 79.998,
-      "eval_steps_per_second": 10.08,
+      "eval_accuracy": 0.6481411764705882,
+      "eval_loss": 1.7193351984024048,
+      "eval_runtime": 7.6621,
+      "eval_samples_per_second": 65.256,
+      "eval_steps_per_second": 8.222,
       "step": 2625
     },
     {
       "epoch": 14.0,
-      "eval_exact_match": 11.2,
-      "eval_f1": 18.26571428571429,
-      "eval_qa_bleu": 10.347091943612007,
-      "eval_qa_exact_match": 0.082,
-      "eval_recite_bleu": 12.720454352082847,
-      "eval_recite_exact_match": 0.014,
+      "eval_exact_match": 18.4,
+      "eval_f1": 27.123376623376636,
+      "eval_qa_bleu": 13.906554273237205,
+      "eval_qa_exact_match": 0.15,
+      "eval_recite_bleu": 14.177162892556794,
+      "eval_recite_exact_match": 0.016,
       "step": 2625
     },
     {
       "epoch": 14.4,
-      "grad_norm": 2.326387405395508,
       "learning_rate": 5e-05,
-      "loss": 0.6312,
+      "loss": 0.3784,
       "step": 2700
     },
     {
-      "epoch": 14.933333333333334,
-      "grad_norm": 2.1555840969085693,
+      "epoch": 14.93,
       "learning_rate": 5e-05,
-      "loss": 0.6267,
+      "loss": 0.3811,
       "step": 2800
     },
     {
-      "epoch": 14.997333333333334,
-      "eval_accuracy": 0.5870582959641255,
-      "eval_loss": 2.338937997817993,
-      "eval_runtime": 6.643,
-      "eval_samples_per_second": 75.267,
-      "eval_steps_per_second": 9.484,
+      "epoch": 15.0,
+      "eval_accuracy": 0.6458509803921568,
+      "eval_loss": 1.8015950918197632,
+      "eval_runtime": 7.7113,
+      "eval_samples_per_second": 64.84,
+      "eval_steps_per_second": 8.17,
       "step": 2812
     },
     {
-      "epoch": 14.997333333333334,
-      "eval_exact_match": 10.6,
-      "eval_f1": 18.311471861471865,
-      "eval_qa_bleu": 9.56012066081651,
-      "eval_qa_exact_match": 0.086,
-      "eval_recite_bleu": 12.790292247765784,
-      "eval_recite_exact_match": 0.022,
+      "epoch": 15.0,
+      "eval_exact_match": 19.0,
+      "eval_f1": 28.891111111111115,
+      "eval_qa_bleu": 16.83679026235492,
+      "eval_qa_exact_match": 0.154,
+      "eval_recite_bleu": 13.04215589870331,
+      "eval_recite_exact_match": 0.012,
       "step": 2812
     },
     {
-      "epoch": 15.466666666666667,
-      "grad_norm": 1.8364830017089844,
+      "epoch": 15.47,
       "learning_rate": 5e-05,
-      "loss": 0.5582,
+      "loss": 0.3287,
       "step": 2900
     },
     {
       "epoch": 16.0,
-      "grad_norm": 1.9223560094833374,
       "learning_rate": 5e-05,
-      "loss": 0.5673,
+      "loss": 0.3373,
       "step": 3000
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.585067264573991,
-      "eval_loss": 2.4404428005218506,
-      "eval_runtime": 6.2254,
-      "eval_samples_per_second": 80.316,
-      "eval_steps_per_second": 10.12,
+      "eval_accuracy": 0.6464627450980392,
+      "eval_loss": 1.8555551767349243,
+      "eval_runtime": 6.6548,
+      "eval_samples_per_second": 75.133,
+      "eval_steps_per_second": 9.467,
       "step": 3000
     },
     {
       "epoch": 16.0,
-      "eval_exact_match": 11.4,
-      "eval_f1": 18.553809523809527,
-      "eval_qa_bleu": 7.054735448183399,
-      "eval_qa_exact_match": 0.088,
-      "eval_recite_bleu": 12.41180854100743,
-      "eval_recite_exact_match": 0.018,
+      "eval_exact_match": 21.0,
+      "eval_f1": 30.546060606060603,
+      "eval_qa_bleu": 15.51818073374785,
+      "eval_qa_exact_match": 0.178,
+      "eval_recite_bleu": 14.611971996497491,
+      "eval_recite_exact_match": 0.016,
       "step": 3000
     },
     {
-      "epoch": 16.533333333333335,
-      "grad_norm": 2.403259515762329,
+      "epoch": 16.53,
       "learning_rate": 5e-05,
-      "loss": 0.4886,
+      "loss": 0.284,
       "step": 3100
     },
     {
-      "epoch": 16.997333333333334,
-      "eval_accuracy": 0.5844035874439462,
-      "eval_loss": 2.5225181579589844,
-      "eval_runtime": 5.9008,
-      "eval_samples_per_second": 84.734,
-      "eval_steps_per_second": 10.677,
+      "epoch": 17.0,
+      "eval_accuracy": 0.644313725490196,
+      "eval_loss": 1.9406696557998657,
+      "eval_runtime": 6.7553,
+      "eval_samples_per_second": 74.016,
+      "eval_steps_per_second": 9.326,
       "step": 3187
     },
     {
-      "epoch": 16.997333333333334,
-      "eval_exact_match": 10.6,
-      "eval_f1": 17.309567099567108,
-      "eval_qa_bleu": 7.6942420610670945,
-      "eval_qa_exact_match": 0.086,
-      "eval_recite_bleu": 11.974979252669918,
+      "epoch": 17.0,
+      "eval_exact_match": 19.2,
+      "eval_f1": 26.95666666666667,
+      "eval_qa_bleu": 13.067462248071651,
+      "eval_qa_exact_match": 0.148,
+      "eval_recite_bleu": 13.516002177466241,
       "eval_recite_exact_match": 0.016,
       "step": 3187
     },
     {
-      "epoch": 17.066666666666666,
-      "grad_norm": 2.1279492378234863,
+      "epoch": 17.07,
       "learning_rate": 5e-05,
-      "loss": 0.4985,
+      "loss": 0.2933,
       "step": 3200
     },
     {
       "epoch": 17.6,
-      "grad_norm": 2.377486228942871,
       "learning_rate": 5e-05,
-      "loss": 0.4357,
+      "loss": 0.25,
       "step": 3300
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.5838026905829596,
-      "eval_loss": 2.590667724609375,
-      "eval_runtime": 5.984,
-      "eval_samples_per_second": 83.556,
-      "eval_steps_per_second": 10.528,
+      "eval_accuracy": 0.6435764705882353,
+      "eval_loss": 1.9759635925292969,
+      "eval_runtime": 7.1655,
+      "eval_samples_per_second": 69.779,
+      "eval_steps_per_second": 8.792,
       "step": 3375
     },
     {
       "epoch": 18.0,
-      "eval_exact_match": 10.8,
-      "eval_f1": 18.150000000000006,
-      "eval_qa_bleu": 10.081752461945443,
-      "eval_qa_exact_match": 0.078,
-      "eval_recite_bleu": 12.17977433983274,
-      "eval_recite_exact_match": 0.018,
+      "eval_exact_match": 18.6,
+      "eval_f1": 27.20079365079366,
+      "eval_qa_bleu": 12.454425793037345,
+      "eval_qa_exact_match": 0.158,
+      "eval_recite_bleu": 13.86644732825742,
+      "eval_recite_exact_match": 0.014,
       "step": 3375
     },
     {
-      "epoch": 18.133333333333333,
-      "grad_norm": 2.425731897354126,
+      "epoch": 18.13,
       "learning_rate": 5e-05,
-      "loss": 0.4424,
+      "loss": 0.2587,
       "step": 3400
     },
     {
-      "epoch": 18.666666666666668,
-      "grad_norm": 2.4585390090942383,
+      "epoch": 18.67,
       "learning_rate": 5e-05,
-      "loss": 0.3873,
+      "loss": 0.2274,
       "step": 3500
     },
     {
-      "epoch": 18.997333333333334,
-      "eval_accuracy": 0.5823228699551569,
-      "eval_loss": 2.6423120498657227,
-      "eval_runtime": 5.5641,
-      "eval_samples_per_second": 89.861,
-      "eval_steps_per_second": 11.323,
+      "epoch": 19.0,
+      "eval_accuracy": 0.6424313725490196,
+      "eval_loss": 2.100283145904541,
+      "eval_runtime": 7.4607,
+      "eval_samples_per_second": 67.018,
+      "eval_steps_per_second": 8.444,
       "step": 3562
     },
     {
-      "epoch": 18.997333333333334,
-      "eval_exact_match": 11.0,
-      "eval_f1": 18.006349206349213,
-      "eval_qa_bleu": 8.890762850928647,
-      "eval_qa_exact_match": 0.092,
-      "eval_recite_bleu": 12.489755382776915,
-      "eval_recite_exact_match": 0.022,
+      "epoch": 19.0,
+      "eval_exact_match": 18.8,
+      "eval_f1": 26.78857142857144,
+      "eval_qa_bleu": 6.755652921468526,
+      "eval_qa_exact_match": 0.16,
+      "eval_recite_bleu": 13.96083887839662,
+      "eval_recite_exact_match": 0.018,
       "step": 3562
     },
     {
       "epoch": 19.2,
-      "grad_norm": 2.00604510307312,
       "learning_rate": 5e-05,
-      "loss": 0.383,
+      "loss": 0.2209,
       "step": 3600
     },
     {
-      "epoch": 19.733333333333334,
-      "grad_norm": 2.393627405166626,
+      "epoch": 19.73,
       "learning_rate": 5e-05,
-      "loss": 0.3497,
+      "loss": 0.2033,
       "step": 3700
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.5803587443946189,
-      "eval_loss": 2.7260048389434814,
-      "eval_runtime": 6.546,
-      "eval_samples_per_second": 76.382,
-      "eval_steps_per_second": 9.624,
+      "eval_accuracy": 0.6424627450980392,
+      "eval_loss": 2.141883373260498,
+      "eval_runtime": 6.6558,
+      "eval_samples_per_second": 75.122,
+      "eval_steps_per_second": 9.465,
       "step": 3750
     },
     {
       "epoch": 20.0,
-      "eval_exact_match": 9.6,
-      "eval_f1": 15.837460317460323,
-      "eval_qa_bleu": 7.780402070269215,
-      "eval_qa_exact_match": 0.078,
-      "eval_recite_bleu": 12.344573606229106,
-      "eval_recite_exact_match": 0.024,
+      "eval_exact_match": 20.4,
+      "eval_f1": 28.975054945054936,
+      "eval_qa_bleu": 14.791681538972547,
+      "eval_qa_exact_match": 0.168,
+      "eval_recite_bleu": 14.920846894015328,
+      "eval_recite_exact_match": 0.016,
       "step": 3750
     },
     {
-      "epoch": 20.266666666666666,
-      "grad_norm": 2.1227269172668457,
+      "epoch": 20.27,
       "learning_rate": 5e-05,
-      "loss": 0.3321,
+      "loss": 0.1928,
       "step": 3800
     },
     {
       "epoch": 20.8,
-      "grad_norm": 2.2578742504119873,
       "learning_rate": 5e-05,
-      "loss": 0.314,
+      "loss": 0.1832,
       "step": 3900
     },
     {
-      "epoch": 20.997333333333334,
-      "eval_accuracy": 0.5801883408071749,
-      "eval_loss": 2.820314884185791,
-      "eval_runtime": 5.6037,
-      "eval_samples_per_second": 89.227,
-      "eval_steps_per_second": 11.243,
+      "epoch": 21.0,
+      "eval_accuracy": 0.6405333333333333,
+      "eval_loss": 2.2004899978637695,
+      "eval_runtime": 7.0682,
+      "eval_samples_per_second": 70.739,
+      "eval_steps_per_second": 8.913,
       "step": 3937
     },
     {
-      "epoch": 20.997333333333334,
-      "eval_exact_match": 12.4,
-      "eval_f1": 17.80337662337662,
-      "eval_qa_bleu": 9.273311361506353,
-      "eval_qa_exact_match": 0.096,
-      "eval_recite_bleu": 12.39691014320001,
-      "eval_recite_exact_match": 0.018,
+      "epoch": 21.0,
+      "eval_exact_match": 19.8,
+      "eval_f1": 27.03857142857144,
+      "eval_qa_bleu": 14.16698111965369,
+      "eval_qa_exact_match": 0.152,
+      "eval_recite_bleu": 14.34411744472555,
+      "eval_recite_exact_match": 0.016,
       "step": 3937
     },
     {
-      "epoch": 21.333333333333332,
-      "grad_norm": 1.9079784154891968,
+      "epoch": 21.33,
       "learning_rate": 5e-05,
-      "loss": 0.2951,
+      "loss": 0.1674,
       "step": 4000
     },
     {
-      "epoch": 21.866666666666667,
-      "grad_norm": 2.5200765132904053,
+      "epoch": 21.87,
       "learning_rate": 5e-05,
-      "loss": 0.2893,
+      "loss": 0.1678,
       "step": 4100
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.5794260089686099,
-      "eval_loss": 2.905937433242798,
-      "eval_runtime": 5.8733,
-      "eval_samples_per_second": 85.132,
-      "eval_steps_per_second": 10.727,
+      "eval_accuracy": 0.6396627450980392,
+      "eval_loss": 2.2465262413024902,
+      "eval_runtime": 7.6611,
+      "eval_samples_per_second": 65.264,
+      "eval_steps_per_second": 8.223,
       "step": 4125
     },
     {
       "epoch": 22.0,
-      "eval_exact_match": 10.2,
-      "eval_f1": 16.468253968253975,
-      "eval_qa_bleu": 6.31129540696892,
-      "eval_qa_exact_match": 0.084,
-      "eval_recite_bleu": 12.11260232701576,
-      "eval_recite_exact_match": 0.022,
+      "eval_exact_match": 18.8,
+      "eval_f1": 26.868253968253967,
+      "eval_qa_bleu": 16.683806863576216,
+      "eval_qa_exact_match": 0.15,
+      "eval_recite_bleu": 15.201480663492395,
+      "eval_recite_exact_match": 0.018,
       "step": 4125
     },
     {
       "epoch": 22.4,
-      "grad_norm": 1.9616879224777222,
       "learning_rate": 5e-05,
-      "loss": 0.2584,
+      "loss": 0.1516,
       "step": 4200
     },
     {
-      "epoch": 22.933333333333334,
-      "grad_norm": 1.854027271270752,
+      "epoch": 22.93,
       "learning_rate": 5e-05,
-      "loss": 0.2583,
+      "loss": 0.1505,
       "step": 4300
     },
     {
-      "epoch": 22.997333333333334,
-      "eval_accuracy": 0.578914798206278,
-      "eval_loss": 2.9786081314086914,
-      "eval_runtime": 6.2893,
-      "eval_samples_per_second": 79.5,
-      "eval_steps_per_second": 10.017,
+      "epoch": 23.0,
+      "eval_accuracy": 0.6399607843137255,
+      "eval_loss": 2.292829990386963,
+      "eval_runtime": 7.6966,
+      "eval_samples_per_second": 64.963,
+      "eval_steps_per_second": 8.185,
       "step": 4312
     },
     {
-      "epoch": 22.997333333333334,
-      "eval_exact_match": 10.2,
-      "eval_f1": 16.564126984126986,
-      "eval_qa_bleu": 6.2299815854348815,
-      "eval_qa_exact_match": 0.086,
-      "eval_recite_bleu": 11.81704787803335,
-      "eval_recite_exact_match": 0.024,
+      "epoch": 23.0,
+      "eval_exact_match": 20.4,
+      "eval_f1": 28.723809523809532,
+      "eval_qa_bleu": 14.649664836489828,
+      "eval_qa_exact_match": 0.16,
+      "eval_recite_bleu": 14.44141809683148,
+      "eval_recite_exact_match": 0.016,
       "step": 4312
     },
     {
-      "epoch": 23.466666666666665,
-      "grad_norm": 1.8527085781097412,
+      "epoch": 23.47,
       "learning_rate": 5e-05,
-      "loss": 0.2289,
+      "loss": 0.1345,
       "step": 4400
     },
     {
       "epoch": 24.0,
-      "grad_norm": 2.258077621459961,
       "learning_rate": 5e-05,
-      "loss": 0.2382,
+      "loss": 0.1406,
       "step": 4500
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.5789596412556054,
-      "eval_loss": 3.0398380756378174,
-      "eval_runtime": 6.5231,
-      "eval_samples_per_second": 76.651,
-      "eval_steps_per_second": 9.658,
+      "eval_accuracy": 0.6399686274509804,
+      "eval_loss": 2.369138240814209,
+      "eval_runtime": 7.0721,
+      "eval_samples_per_second": 70.701,
+      "eval_steps_per_second": 8.908,
       "step": 4500
     },
     {
       "epoch": 24.0,
-      "eval_exact_match": 10.2,
-      "eval_f1": 16.815281385281384,
-      "eval_qa_bleu": 6.817434246718267,
-      "eval_qa_exact_match": 0.082,
-      "eval_recite_bleu": 12.918682221726373,
-      "eval_recite_exact_match": 0.022,
+      "eval_exact_match": 17.8,
+      "eval_f1": 25.43047619047621,
+      "eval_qa_bleu": 13.339685450002685,
+      "eval_qa_exact_match": 0.148,
+      "eval_recite_bleu": 13.754679861863655,
+      "eval_recite_exact_match": 0.018,
       "step": 4500
     }
   ],
   "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": 4.5327709405551e+17,
-  "train_batch_size": 1,
+  "total_flos": 1.0429601922846556e+18,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-4500/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:01f9f0dc44c57fb44afb03320f4d02bf48066009e3bbcd96d8e4377b0171c5a5
-size 5304
+oid sha256:ae9143de0ad8fd0f027f30f757d262a6e02edc4ceeaea2c21697f4f44b5b1eeb
+size 4728