Training in progress, epoch 36, checkpoint

checkpoint-12635/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-12635/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aea7f1d9e13fe279c252d2d7e8f70d6d3304198e9d173f2e3ef34af3579697ab
+size 143269386

checkpoint-12635/added_tokens.json

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

checkpoint-12635/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:58c5d4cf8be278a49085103cf84d918a8d8ec4719e014265d67430d1f5560617
-size 224537202
+oid sha256:1f9b729a0c8b47914ace3e3a511b0c1cee9d874934a5e35a6db1b256f7625428
+size 286585234

checkpoint-12635/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b1c73361ad3c882cbe605a408f270fc6ace3d51b0fb03ad8444fe07467c5fe49
+oid sha256:b692ee65eb6786cfe0642969b7329fa27470cf69addb261201851e97556e6c99
 size 15024

checkpoint-12635/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d23e208377918072f5def6cf66ede2011fe4d5b5c4dc4070927a7e793dbf8918
+oid sha256:191f911b948499cafbf7896fa5c8e5df60bfefc2481ec96b46da3e5e4e841404
 size 15024

checkpoint-12635/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1339d1631fba09b25f9014f91ee45e634bb81c5eae2cfdf700eb6cb23c4cc752
+oid sha256:792ff678e7ea32c7947a300ede0857f5d503840075119c503f85daed4c19e868
 size 15024

checkpoint-12635/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a0991a74e41552710ecc952d7ab7620de52cd641a4be8c7a5759d48ed63476d1
+oid sha256:2519c90183245c0ecf426cb1717a4aa71c41e5208f490747d81fed7c95bae5ca
 size 15024

checkpoint-12635/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-12635/tokenizer.model

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

checkpoint-12635/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-12635/trainer_state.json

@@ -9,1499 +9,1359 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.29282576866764276,
-      "grad_norm": 0.26115021109580994,
+      "epoch": 0.29,
       "learning_rate": 3e-05,
-      "loss": 2.0719,
+      "loss": 1.574,
       "step": 100
     },
     {
-      "epoch": 0.5856515373352855,
-      "grad_norm": 0.2679530680179596,
+      "epoch": 0.59,
       "learning_rate": 3e-05,
-      "loss": 1.969,
+      "loss": 1.474,
       "step": 200
     },
     {
-      "epoch": 0.8784773060029283,
-      "grad_norm": 0.2916224002838135,
+      "epoch": 0.88,
       "learning_rate": 3e-05,
-      "loss": 1.9686,
+      "loss": 1.4104,
       "step": 300
     },
     {
-      "epoch": 0.9985358711566618,
-      "eval_accuracy": 0.43743589743589745,
-      "eval_loss": 3.6918885707855225,
-      "eval_runtime": 5.2974,
-      "eval_samples_per_second": 94.385,
-      "eval_steps_per_second": 11.893,
+      "epoch": 1.0,
+      "eval_accuracy": 0.4536923076923077,
+      "eval_loss": 3.3574578762054443,
+      "eval_runtime": 4.4299,
+      "eval_samples_per_second": 112.869,
+      "eval_steps_per_second": 14.221,
       "step": 341
     },
     {
-      "epoch": 0.9985358711566618,
-      "eval_bleu": 0.1527908371131025,
-      "eval_exact_match": 0.012,
-      "eval_prefix_exact_match": 0.026,
+      "epoch": 1.0,
+      "eval_bleu": 0.08840557310176213,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.058,
       "step": 341
     },
     {
-      "epoch": 1.171303074670571,
-      "grad_norm": 0.3534347712993622,
+      "epoch": 1.17,
       "learning_rate": 3e-05,
-      "loss": 1.9603,
+      "loss": 1.3889,
       "step": 400
     },
     {
-      "epoch": 1.4641288433382138,
-      "grad_norm": 0.3338424563407898,
+      "epoch": 1.46,
       "learning_rate": 3e-05,
-      "loss": 1.9235,
+      "loss": 1.387,
       "step": 500
     },
     {
-      "epoch": 1.7569546120058566,
-      "grad_norm": 0.4033390283584595,
+      "epoch": 1.76,
       "learning_rate": 3e-05,
-      "loss": 1.9337,
+      "loss": 1.389,
       "step": 600
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.44764102564102565,
-      "eval_loss": 3.7484843730926514,
-      "eval_runtime": 5.0649,
-      "eval_samples_per_second": 98.718,
-      "eval_steps_per_second": 12.438,
+      "eval_accuracy": 0.4543589743589744,
+      "eval_loss": 3.4179794788360596,
+      "eval_runtime": 4.1997,
+      "eval_samples_per_second": 119.055,
+      "eval_steps_per_second": 15.001,
       "step": 683
     },
     {
       "epoch": 2.0,
-      "eval_bleu": 0.21311912714932216,
-      "eval_exact_match": 0.02,
-      "eval_prefix_exact_match": 0.032,
+      "eval_bleu": 0.09820569565497561,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.056,
       "step": 683
     },
     {
-      "epoch": 2.049780380673499,
-      "grad_norm": 0.38882532715797424,
+      "epoch": 2.05,
       "learning_rate": 3e-05,
-      "loss": 1.9267,
+      "loss": 1.3725,
       "step": 700
     },
     {
-      "epoch": 2.342606149341142,
-      "grad_norm": 0.44310709834098816,
+      "epoch": 2.34,
       "learning_rate": 3e-05,
-      "loss": 1.8823,
+      "loss": 1.3588,
       "step": 800
     },
     {
-      "epoch": 2.6354319180087846,
-      "grad_norm": 0.5196680426597595,
+      "epoch": 2.64,
       "learning_rate": 3e-05,
-      "loss": 1.901,
+      "loss": 1.348,
       "step": 900
     },
     {
-      "epoch": 2.9282576866764276,
-      "grad_norm": 0.5293437242507935,
+      "epoch": 2.93,
       "learning_rate": 3e-05,
-      "loss": 1.9033,
+      "loss": 1.3414,
       "step": 1000
     },
     {
-      "epoch": 2.998535871156662,
-      "eval_accuracy": 0.4495897435897436,
-      "eval_loss": 3.882551670074463,
-      "eval_runtime": 5.1314,
-      "eval_samples_per_second": 97.439,
-      "eval_steps_per_second": 12.277,
+      "epoch": 3.0,
+      "eval_accuracy": 0.45476923076923076,
+      "eval_loss": 3.511923313140869,
+      "eval_runtime": 4.1985,
+      "eval_samples_per_second": 119.089,
+      "eval_steps_per_second": 15.005,
       "step": 1024
     },
     {
-      "epoch": 2.998535871156662,
-      "eval_bleu": 0.24307488023362148,
-      "eval_exact_match": 0.016,
-      "eval_prefix_exact_match": 0.032,
+      "epoch": 3.0,
+      "eval_bleu": 0.09889831624589397,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.056,
       "step": 1024
     },
     {
-      "epoch": 3.22108345534407,
-      "grad_norm": 0.552943229675293,
+      "epoch": 3.22,
       "learning_rate": 3e-05,
-      "loss": 1.8529,
+      "loss": 1.3166,
       "step": 1100
     },
     {
-      "epoch": 3.513909224011713,
-      "grad_norm": 0.6421984434127808,
+      "epoch": 3.51,
       "learning_rate": 3e-05,
-      "loss": 1.8342,
+      "loss": 1.3035,
       "step": 1200
     },
     {
-      "epoch": 3.8067349926793557,
-      "grad_norm": 0.7155198454856873,
+      "epoch": 3.81,
       "learning_rate": 3e-05,
-      "loss": 1.857,
+      "loss": 1.3002,
       "step": 1300
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.4481025641025641,
-      "eval_loss": 3.970107316970825,
-      "eval_runtime": 5.1563,
-      "eval_samples_per_second": 96.969,
-      "eval_steps_per_second": 12.218,
+      "eval_accuracy": 0.45543589743589746,
+      "eval_loss": 3.5288071632385254,
+      "eval_runtime": 4.4066,
+      "eval_samples_per_second": 113.466,
+      "eval_steps_per_second": 14.297,
       "step": 1366
     },
     {
       "epoch": 4.0,
-      "eval_bleu": 0.25309406645138827,
-      "eval_exact_match": 0.016,
-      "eval_prefix_exact_match": 0.044,
+      "eval_bleu": 0.11209345982903633,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.054,
       "step": 1366
     },
     {
-      "epoch": 4.099560761346998,
-      "grad_norm": 0.7837833166122437,
+      "epoch": 4.1,
       "learning_rate": 3e-05,
-      "loss": 1.8284,
+      "loss": 1.2929,
       "step": 1400
     },
     {
-      "epoch": 4.392386530014641,
-      "grad_norm": 0.784609854221344,
+      "epoch": 4.39,
       "learning_rate": 3e-05,
-      "loss": 1.7884,
+      "loss": 1.2693,
       "step": 1500
     },
     {
-      "epoch": 4.685212298682284,
-      "grad_norm": 0.834195613861084,
+      "epoch": 4.69,
       "learning_rate": 3e-05,
-      "loss": 1.7841,
+      "loss": 1.2697,
       "step": 1600
     },
     {
-      "epoch": 4.978038067349927,
-      "grad_norm": 0.8902062177658081,
+      "epoch": 4.98,
       "learning_rate": 3e-05,
-      "loss": 1.8042,
+      "loss": 1.2574,
       "step": 1700
     },
     {
-      "epoch": 4.998535871156662,
-      "eval_accuracy": 0.4472820512820513,
-      "eval_loss": 4.117065906524658,
-      "eval_runtime": 5.2376,
-      "eval_samples_per_second": 95.464,
-      "eval_steps_per_second": 12.028,
+      "epoch": 5.0,
+      "eval_accuracy": 0.45394871794871794,
+      "eval_loss": 3.6893365383148193,
+      "eval_runtime": 4.1122,
+      "eval_samples_per_second": 121.591,
+      "eval_steps_per_second": 15.32,
       "step": 1707
     },
     {
-      "epoch": 4.998535871156662,
-      "eval_bleu": 0.2531049955279983,
-      "eval_exact_match": 0.014,
-      "eval_prefix_exact_match": 0.042,
+      "epoch": 5.0,
+      "eval_bleu": 0.12076940663377221,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.056,
       "step": 1707
     },
     {
-      "epoch": 5.270863836017569,
-      "grad_norm": 0.9495383501052856,
+      "epoch": 5.27,
       "learning_rate": 3e-05,
-      "loss": 1.7493,
+      "loss": 1.2097,
       "step": 1800
     },
     {
-      "epoch": 5.563689604685212,
-      "grad_norm": 0.9647482633590698,
+      "epoch": 5.56,
       "learning_rate": 3e-05,
-      "loss": 1.7334,
+      "loss": 1.2188,
       "step": 1900
     },
     {
-      "epoch": 5.856515373352855,
-      "grad_norm": 0.9825767874717712,
+      "epoch": 5.86,
       "learning_rate": 3e-05,
-      "loss": 1.7443,
+      "loss": 1.2258,
       "step": 2000
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.446974358974359,
-      "eval_loss": 4.183700084686279,
-      "eval_runtime": 5.2674,
-      "eval_samples_per_second": 94.923,
-      "eval_steps_per_second": 11.96,
+      "eval_accuracy": 0.45615384615384613,
+      "eval_loss": 3.725893497467041,
+      "eval_runtime": 4.0989,
+      "eval_samples_per_second": 121.984,
+      "eval_steps_per_second": 15.37,
       "step": 2049
     },
     {
       "epoch": 6.0,
-      "eval_bleu": 0.23587359955369594,
-      "eval_exact_match": 0.012,
-      "eval_prefix_exact_match": 0.04,
+      "eval_bleu": 0.11884272898137646,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.058,
       "step": 2049
     },
     {
-      "epoch": 6.149341142020498,
-      "grad_norm": 1.0430296659469604,
+      "epoch": 6.15,
       "learning_rate": 3e-05,
-      "loss": 1.6959,
+      "loss": 1.205,
       "step": 2100
     },
     {
-      "epoch": 6.44216691068814,
-      "grad_norm": 1.0777732133865356,
+      "epoch": 6.44,
       "learning_rate": 3e-05,
-      "loss": 1.6795,
+      "loss": 1.1643,
       "step": 2200
     },
     {
-      "epoch": 6.734992679355783,
-      "grad_norm": 1.0959677696228027,
+      "epoch": 6.73,
       "learning_rate": 3e-05,
-      "loss": 1.7019,
+      "loss": 1.1844,
       "step": 2300
     },
     {
-      "epoch": 6.998535871156662,
-      "eval_accuracy": 0.4461538461538462,
-      "eval_loss": 4.260408401489258,
-      "eval_runtime": 5.178,
-      "eval_samples_per_second": 96.563,
-      "eval_steps_per_second": 12.167,
+      "epoch": 7.0,
+      "eval_accuracy": 0.45594871794871794,
+      "eval_loss": 3.724449396133423,
+      "eval_runtime": 4.8158,
+      "eval_samples_per_second": 103.825,
+      "eval_steps_per_second": 13.082,
       "step": 2390
     },
     {
-      "epoch": 6.998535871156662,
-      "eval_bleu": 0.2008158466316808,
-      "eval_exact_match": 0.006,
-      "eval_prefix_exact_match": 0.042,
+      "epoch": 7.0,
+      "eval_bleu": 0.12407896336995003,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.048,
       "step": 2390
     },
     {
-      "epoch": 7.027818448023426,
-      "grad_norm": 1.0649549961090088,
+      "epoch": 7.03,
       "learning_rate": 3e-05,
-      "loss": 1.6812,
+      "loss": 1.1771,
       "step": 2400
     },
     {
-      "epoch": 7.320644216691068,
-      "grad_norm": 1.294122576713562,
+      "epoch": 7.32,
       "learning_rate": 3e-05,
-      "loss": 1.6226,
+      "loss": 1.1218,
       "step": 2500
     },
     {
-      "epoch": 7.613469985358711,
-      "grad_norm": 1.2657020092010498,
+      "epoch": 7.61,
       "learning_rate": 3e-05,
-      "loss": 1.6404,
+      "loss": 1.1491,
       "step": 2600
     },
     {
-      "epoch": 7.906295754026354,
-      "grad_norm": 1.1963133811950684,
+      "epoch": 7.91,
       "learning_rate": 3e-05,
-      "loss": 1.6305,
+      "loss": 1.1363,
       "step": 2700
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.44148717948717947,
-      "eval_loss": 4.406544208526611,
-      "eval_runtime": 5.3355,
-      "eval_samples_per_second": 93.711,
-      "eval_steps_per_second": 11.808,
+      "eval_accuracy": 0.4543589743589744,
+      "eval_loss": 3.8138701915740967,
+      "eval_runtime": 4.1963,
+      "eval_samples_per_second": 119.152,
+      "eval_steps_per_second": 15.013,
       "step": 2732
     },
     {
       "epoch": 8.0,
-      "eval_bleu": 0.24435500637461924,
-      "eval_exact_match": 0.008,
-      "eval_prefix_exact_match": 0.038,
+      "eval_bleu": 0.11947629327387163,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.046,
       "step": 2732
     },
     {
-      "epoch": 8.199121522693996,
-      "grad_norm": 1.3801360130310059,
+      "epoch": 8.2,
       "learning_rate": 3e-05,
-      "loss": 1.5734,
+      "loss": 1.0858,
       "step": 2800
     },
     {
-      "epoch": 8.49194729136164,
-      "grad_norm": 1.3306772708892822,
+      "epoch": 8.49,
       "learning_rate": 3e-05,
-      "loss": 1.5651,
+      "loss": 1.0951,
       "step": 2900
     },
     {
-      "epoch": 8.784773060029282,
-      "grad_norm": 1.5165746212005615,
+      "epoch": 8.78,
       "learning_rate": 3e-05,
-      "loss": 1.6056,
+      "loss": 1.0903,
       "step": 3000
     },
     {
-      "epoch": 8.998535871156662,
-      "eval_accuracy": 0.4397948717948718,
-      "eval_loss": 4.448728561401367,
-      "eval_runtime": 5.187,
-      "eval_samples_per_second": 96.395,
-      "eval_steps_per_second": 12.146,
+      "epoch": 9.0,
+      "eval_accuracy": 0.4524102564102564,
+      "eval_loss": 3.9115548133850098,
+      "eval_runtime": 4.4011,
+      "eval_samples_per_second": 113.609,
+      "eval_steps_per_second": 14.315,
       "step": 3073
     },
     {
-      "epoch": 8.998535871156662,
-      "eval_bleu": 0.20120283162176952,
-      "eval_exact_match": 0.008,
-      "eval_prefix_exact_match": 0.036,
+      "epoch": 9.0,
+      "eval_bleu": 0.13426062809240774,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.034,
       "step": 3073
     },
     {
-      "epoch": 9.077598828696924,
-      "grad_norm": 1.4077060222625732,
+      "epoch": 9.08,
       "learning_rate": 3e-05,
-      "loss": 1.5612,
+      "loss": 1.0822,
       "step": 3100
     },
     {
-      "epoch": 9.370424597364568,
-      "grad_norm": 1.475984811782837,
+      "epoch": 9.37,
       "learning_rate": 3e-05,
-      "loss": 1.5158,
+      "loss": 1.0414,
       "step": 3200
     },
     {
-      "epoch": 9.66325036603221,
-      "grad_norm": 1.586300015449524,
+      "epoch": 9.66,
       "learning_rate": 3e-05,
-      "loss": 1.5145,
+      "loss": 1.0563,
       "step": 3300
     },
     {
-      "epoch": 9.956076134699854,
-      "grad_norm": 1.510048270225525,
+      "epoch": 9.96,
       "learning_rate": 3e-05,
-      "loss": 1.5521,
+      "loss": 1.0538,
       "step": 3400
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.4388717948717949,
-      "eval_loss": 4.5474138259887695,
-      "eval_runtime": 5.169,
-      "eval_samples_per_second": 96.731,
-      "eval_steps_per_second": 12.188,
+      "eval_accuracy": 0.4515897435897436,
+      "eval_loss": 3.92203426361084,
+      "eval_runtime": 4.4105,
+      "eval_samples_per_second": 113.366,
+      "eval_steps_per_second": 14.284,
       "step": 3415
     },
     {
       "epoch": 10.0,
-      "eval_bleu": 0.2258317707850195,
-      "eval_exact_match": 0.004,
+      "eval_bleu": 0.11780869016319889,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.032,
       "step": 3415
     },
     {
-      "epoch": 10.248901903367496,
-      "grad_norm": 1.6384536027908325,
+      "epoch": 10.25,
       "learning_rate": 3e-05,
-      "loss": 1.4631,
+      "loss": 0.991,
       "step": 3500
     },
     {
-      "epoch": 10.541727672035138,
-      "grad_norm": 1.7626007795333862,
+      "epoch": 10.54,
       "learning_rate": 3e-05,
-      "loss": 1.4623,
+      "loss": 1.0019,
       "step": 3600
     },
     {
-      "epoch": 10.834553440702782,
-      "grad_norm": 1.6483694314956665,
+      "epoch": 10.83,
       "learning_rate": 3e-05,
-      "loss": 1.4934,
+      "loss": 0.9971,
       "step": 3700
     },
     {
-      "epoch": 10.998535871156662,
-      "eval_accuracy": 0.4367179487179487,
-      "eval_loss": 4.589761734008789,
-      "eval_runtime": 5.853,
-      "eval_samples_per_second": 85.426,
-      "eval_steps_per_second": 10.764,
+      "epoch": 11.0,
+      "eval_accuracy": 0.45143589743589746,
+      "eval_loss": 3.967252492904663,
+      "eval_runtime": 4.1003,
+      "eval_samples_per_second": 121.943,
+      "eval_steps_per_second": 15.365,
       "step": 3756
     },
     {
-      "epoch": 10.998535871156662,
-      "eval_bleu": 0.20792247275869374,
-      "eval_exact_match": 0.004,
+      "epoch": 11.0,
+      "eval_bleu": 0.12407049349871409,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.026,
       "step": 3756
     },
     {
-      "epoch": 11.127379209370424,
-      "grad_norm": 1.8367844820022583,
+      "epoch": 11.13,
       "learning_rate": 3e-05,
-      "loss": 1.4521,
+      "loss": 0.9861,
       "step": 3800
     },
     {
-      "epoch": 11.420204978038067,
-      "grad_norm": 1.8123779296875,
+      "epoch": 11.42,
       "learning_rate": 3e-05,
-      "loss": 1.4085,
+      "loss": 0.9436,
       "step": 3900
     },
     {
-      "epoch": 11.71303074670571,
-      "grad_norm": 1.9437752962112427,
+      "epoch": 11.71,
       "learning_rate": 3e-05,
-      "loss": 1.4287,
+      "loss": 0.9699,
       "step": 4000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.43553846153846154,
-      "eval_loss": 4.691086769104004,
-      "eval_runtime": 5.3738,
-      "eval_samples_per_second": 93.043,
-      "eval_steps_per_second": 11.723,
+      "eval_accuracy": 0.45076923076923076,
+      "eval_loss": 4.033609867095947,
+      "eval_runtime": 4.8103,
+      "eval_samples_per_second": 103.943,
+      "eval_steps_per_second": 13.097,
       "step": 4098
     },
     {
       "epoch": 12.0,
-      "eval_bleu": 0.21672573731424186,
-      "eval_exact_match": 0.006,
-      "eval_prefix_exact_match": 0.028,
+      "eval_bleu": 0.12190601380876163,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.026,
       "step": 4098
     },
     {
-      "epoch": 12.005856515373353,
-      "grad_norm": 1.5838289260864258,
+      "epoch": 12.01,
       "learning_rate": 3e-05,
-      "loss": 1.4361,
+      "loss": 0.9783,
       "step": 4100
     },
     {
-      "epoch": 12.298682284040996,
-      "grad_norm": 1.8575290441513062,
+      "epoch": 12.3,
       "learning_rate": 3e-05,
-      "loss": 1.3433,
+      "loss": 0.9064,
       "step": 4200
     },
     {
-      "epoch": 12.591508052708638,
-      "grad_norm": 1.9598041772842407,
+      "epoch": 12.59,
       "learning_rate": 3e-05,
-      "loss": 1.3661,
+      "loss": 0.9178,
       "step": 4300
     },
     {
-      "epoch": 12.88433382137628,
-      "grad_norm": 1.972550868988037,
+      "epoch": 12.88,
       "learning_rate": 3e-05,
-      "loss": 1.3846,
+      "loss": 0.9235,
       "step": 4400
     },
     {
-      "epoch": 12.998535871156662,
-      "eval_accuracy": 0.43553846153846154,
-      "eval_loss": 4.762892246246338,
-      "eval_runtime": 5.106,
-      "eval_samples_per_second": 97.925,
-      "eval_steps_per_second": 12.339,
+      "epoch": 13.0,
+      "eval_accuracy": 0.4492820512820513,
+      "eval_loss": 4.002023696899414,
+      "eval_runtime": 4.8204,
+      "eval_samples_per_second": 103.726,
+      "eval_steps_per_second": 13.069,
       "step": 4439
     },
     {
-      "epoch": 12.998535871156662,
-      "eval_bleu": 0.2416046504676966,
-      "eval_exact_match": 0.006,
-      "eval_prefix_exact_match": 0.028,
+      "epoch": 13.0,
+      "eval_bleu": 0.13115986163477472,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.026,
       "step": 4439
     },
     {
-      "epoch": 13.177159590043924,
-      "grad_norm": 1.9469107389450073,
+      "epoch": 13.18,
       "learning_rate": 3e-05,
-      "loss": 1.3229,
+      "loss": 0.8902,
       "step": 4500
     },
     {
-      "epoch": 13.469985358711567,
-      "grad_norm": 2.1680469512939453,
+      "epoch": 13.47,
       "learning_rate": 3e-05,
-      "loss": 1.3172,
+      "loss": 0.8703,
       "step": 4600
     },
     {
-      "epoch": 13.762811127379209,
-      "grad_norm": 2.1960484981536865,
+      "epoch": 13.76,
       "learning_rate": 3e-05,
-      "loss": 1.3185,
+      "loss": 0.891,
       "step": 4700
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.4327692307692308,
-      "eval_loss": 4.758542537689209,
-      "eval_runtime": 5.778,
-      "eval_samples_per_second": 86.535,
-      "eval_steps_per_second": 10.903,
+      "eval_accuracy": 0.44774358974358974,
+      "eval_loss": 4.071566104888916,
+      "eval_runtime": 4.8316,
+      "eval_samples_per_second": 103.485,
+      "eval_steps_per_second": 13.039,
       "step": 4781
     },
     {
       "epoch": 14.0,
-      "eval_bleu": 0.19085834097154403,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.022,
+      "eval_bleu": 0.12164300660650715,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.024,
       "step": 4781
     },
     {
-      "epoch": 14.055636896046853,
-      "grad_norm": 2.4733147621154785,
+      "epoch": 14.06,
       "learning_rate": 3e-05,
-      "loss": 1.3027,
+      "loss": 0.8698,
       "step": 4800
     },
     {
-      "epoch": 14.348462664714495,
-      "grad_norm": 2.2826030254364014,
+      "epoch": 14.35,
       "learning_rate": 3e-05,
-      "loss": 1.2513,
+      "loss": 0.8338,
       "step": 4900
     },
     {
-      "epoch": 14.641288433382137,
-      "grad_norm": 2.041346788406372,
+      "epoch": 14.64,
       "learning_rate": 3e-05,
-      "loss": 1.2637,
+      "loss": 0.8362,
       "step": 5000
     },
     {
-      "epoch": 14.93411420204978,
-      "grad_norm": 2.3494467735290527,
+      "epoch": 14.93,
       "learning_rate": 3e-05,
-      "loss": 1.2667,
+      "loss": 0.845,
       "step": 5100
     },
     {
-      "epoch": 14.998535871156662,
-      "eval_accuracy": 0.4308717948717949,
-      "eval_loss": 4.938868522644043,
-      "eval_runtime": 5.9024,
-      "eval_samples_per_second": 84.711,
-      "eval_steps_per_second": 10.674,
+      "epoch": 15.0,
+      "eval_accuracy": 0.44774358974358974,
+      "eval_loss": 4.09920597076416,
+      "eval_runtime": 4.4174,
+      "eval_samples_per_second": 113.19,
+      "eval_steps_per_second": 14.262,
       "step": 5122
     },
     {
-      "epoch": 14.998535871156662,
-      "eval_bleu": 0.18549760176612626,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.024,
+      "epoch": 15.0,
+      "eval_bleu": 0.1224977607294519,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.02,
       "step": 5122
     },
     {
-      "epoch": 15.226939970717423,
-      "grad_norm": 2.411022663116455,
+      "epoch": 15.23,
       "learning_rate": 3e-05,
-      "loss": 1.2158,
+      "loss": 0.7981,
       "step": 5200
     },
     {
-      "epoch": 15.519765739385067,
-      "grad_norm": 2.3778491020202637,
+      "epoch": 15.52,
       "learning_rate": 3e-05,
-      "loss": 1.193,
+      "loss": 0.7978,
       "step": 5300
     },
     {
-      "epoch": 15.812591508052709,
-      "grad_norm": 2.2003743648529053,
+      "epoch": 15.81,
       "learning_rate": 3e-05,
-      "loss": 1.2144,
+      "loss": 0.8009,
       "step": 5400
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.43025641025641026,
-      "eval_loss": 4.898688316345215,
-      "eval_runtime": 5.1834,
-      "eval_samples_per_second": 96.463,
-      "eval_steps_per_second": 12.154,
+      "eval_accuracy": 0.4464102564102564,
+      "eval_loss": 4.093270301818848,
+      "eval_runtime": 4.102,
+      "eval_samples_per_second": 121.892,
+      "eval_steps_per_second": 15.358,
       "step": 5464
     },
     {
       "epoch": 16.0,
-      "eval_bleu": 0.20104450766148693,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.026,
+      "eval_bleu": 0.12201237729153987,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.022,
       "step": 5464
     },
     {
-      "epoch": 16.105417276720353,
-      "grad_norm": 2.1213016510009766,
+      "epoch": 16.11,
       "learning_rate": 3e-05,
-      "loss": 1.1842,
+      "loss": 0.7948,
       "step": 5500
     },
     {
-      "epoch": 16.398243045387993,
-      "grad_norm": 2.4457600116729736,
+      "epoch": 16.4,
       "learning_rate": 3e-05,
-      "loss": 1.1527,
+      "loss": 0.746,
       "step": 5600
     },
     {
-      "epoch": 16.691068814055637,
-      "grad_norm": 2.5844104290008545,
+      "epoch": 16.69,
       "learning_rate": 3e-05,
-      "loss": 1.1536,
+      "loss": 0.7627,
       "step": 5700
     },
     {
-      "epoch": 16.98389458272328,
-      "grad_norm": 2.7657299041748047,
+      "epoch": 16.98,
       "learning_rate": 3e-05,
-      "loss": 1.1708,
+      "loss": 0.782,
       "step": 5800
     },
     {
-      "epoch": 16.998535871156662,
-      "eval_accuracy": 0.4297435897435897,
-      "eval_loss": 5.00166654586792,
-      "eval_runtime": 5.4251,
-      "eval_samples_per_second": 92.164,
-      "eval_steps_per_second": 11.613,
+      "epoch": 17.0,
+      "eval_accuracy": 0.44671794871794873,
+      "eval_loss": 4.1283488273620605,
+      "eval_runtime": 4.84,
+      "eval_samples_per_second": 103.305,
+      "eval_steps_per_second": 13.016,
       "step": 5805
     },
     {
-      "epoch": 16.998535871156662,
-      "eval_bleu": 0.2059810081312132,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.028,
+      "epoch": 17.0,
+      "eval_bleu": 0.1263852710208009,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.02,
       "step": 5805
     },
     {
-      "epoch": 17.27672035139092,
-      "grad_norm": 2.4050705432891846,
+      "epoch": 17.28,
       "learning_rate": 3e-05,
-      "loss": 1.083,
+      "loss": 0.7142,
       "step": 5900
     },
     {
-      "epoch": 17.569546120058565,
-      "grad_norm": 2.584331750869751,
+      "epoch": 17.57,
       "learning_rate": 3e-05,
-      "loss": 1.0993,
+      "loss": 0.7253,
       "step": 6000
     },
     {
-      "epoch": 17.86237188872621,
-      "grad_norm": 3.257659912109375,
+      "epoch": 17.86,
       "learning_rate": 3e-05,
-      "loss": 1.1146,
+      "loss": 0.7294,
       "step": 6100
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.43066666666666664,
-      "eval_loss": 4.977816581726074,
-      "eval_runtime": 5.2952,
-      "eval_samples_per_second": 94.425,
-      "eval_steps_per_second": 11.898,
+      "eval_accuracy": 0.44564102564102565,
+      "eval_loss": 4.164257049560547,
+      "eval_runtime": 4.1037,
+      "eval_samples_per_second": 121.84,
+      "eval_steps_per_second": 15.352,
       "step": 6147
     },
     {
       "epoch": 18.0,
-      "eval_bleu": 0.1948234970665932,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.032,
+      "eval_bleu": 0.10715019125675011,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.022,
       "step": 6147
     },
     {
-      "epoch": 18.15519765739385,
-      "grad_norm": 3.068880319595337,
+      "epoch": 18.16,
       "learning_rate": 3e-05,
-      "loss": 1.0673,
+      "loss": 0.7055,
       "step": 6200
     },
     {
-      "epoch": 18.448023426061493,
-      "grad_norm": 2.737912893295288,
+      "epoch": 18.45,
       "learning_rate": 3e-05,
-      "loss": 1.0381,
+      "loss": 0.672,
       "step": 6300
     },
     {
-      "epoch": 18.740849194729137,
-      "grad_norm": 2.878462076187134,
+      "epoch": 18.74,
       "learning_rate": 3e-05,
-      "loss": 1.0531,
+      "loss": 0.6792,
       "step": 6400
     },
     {
-      "epoch": 18.998535871156662,
-      "eval_accuracy": 0.42866666666666664,
-      "eval_loss": 5.121572494506836,
-      "eval_runtime": 5.3996,
-      "eval_samples_per_second": 92.599,
-      "eval_steps_per_second": 11.667,
+      "epoch": 19.0,
+      "eval_accuracy": 0.44487179487179485,
+      "eval_loss": 4.185911655426025,
+      "eval_runtime": 4.823,
+      "eval_samples_per_second": 103.67,
+      "eval_steps_per_second": 13.062,
       "step": 6488
     },
     {
-      "epoch": 18.998535871156662,
-      "eval_bleu": 0.19788734504730915,
+      "epoch": 19.0,
+      "eval_bleu": 0.11314440439431668,
       "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.02,
       "step": 6488
     },
     {
-      "epoch": 19.03367496339678,
-      "grad_norm": 2.6934640407562256,
+      "epoch": 19.03,
       "learning_rate": 3e-05,
-      "loss": 1.0564,
+      "loss": 0.7053,
       "step": 6500
     },
     {
-      "epoch": 19.32650073206442,
-      "grad_norm": 2.822169542312622,
+      "epoch": 19.33,
       "learning_rate": 3e-05,
-      "loss": 0.99,
+      "loss": 0.6448,
       "step": 6600
     },
     {
-      "epoch": 19.619326500732065,
-      "grad_norm": 2.520814895629883,
+      "epoch": 19.62,
       "learning_rate": 3e-05,
-      "loss": 0.991,
+      "loss": 0.6448,
       "step": 6700
     },
     {
-      "epoch": 19.91215226939971,
-      "grad_norm": 3.1566810607910156,
+      "epoch": 19.91,
       "learning_rate": 3e-05,
-      "loss": 1.0158,
+      "loss": 0.6672,
       "step": 6800
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.42728205128205127,
-      "eval_loss": 5.120981216430664,
-      "eval_runtime": 5.394,
-      "eval_samples_per_second": 92.696,
-      "eval_steps_per_second": 11.68,
+      "eval_accuracy": 0.4436923076923077,
+      "eval_loss": 4.201004981994629,
+      "eval_runtime": 4.1985,
+      "eval_samples_per_second": 119.09,
+      "eval_steps_per_second": 15.005,
       "step": 6830
     },
     {
       "epoch": 20.0,
-      "eval_bleu": 0.15806516253752823,
-      "eval_exact_match": 0.006,
-      "eval_prefix_exact_match": 0.028,
+      "eval_bleu": 0.11913021835274795,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.02,
       "step": 6830
     },
     {
-      "epoch": 20.20497803806735,
-      "grad_norm": 2.937910556793213,
+      "epoch": 20.2,
       "learning_rate": 3e-05,
-      "loss": 0.9615,
+      "loss": 0.6146,
       "step": 6900
     },
     {
-      "epoch": 20.497803806734993,
-      "grad_norm": 3.2016685009002686,
+      "epoch": 20.5,
       "learning_rate": 3e-05,
-      "loss": 0.9404,
+      "loss": 0.6102,
       "step": 7000
     },
     {
-      "epoch": 20.790629575402637,
-      "grad_norm": 3.2996864318847656,
+      "epoch": 20.79,
       "learning_rate": 3e-05,
-      "loss": 0.9555,
+      "loss": 0.6258,
       "step": 7100
     },
     {
-      "epoch": 20.998535871156662,
-      "eval_accuracy": 0.42928205128205127,
-      "eval_loss": 5.198832988739014,
-      "eval_runtime": 5.8905,
-      "eval_samples_per_second": 84.882,
-      "eval_steps_per_second": 10.695,
+      "epoch": 21.0,
+      "eval_accuracy": 0.4428717948717949,
+      "eval_loss": 4.230019569396973,
+      "eval_runtime": 4.4124,
+      "eval_samples_per_second": 113.318,
+      "eval_steps_per_second": 14.278,
       "step": 7171
     },
     {
-      "epoch": 20.998535871156662,
-      "eval_bleu": 0.18303907416923465,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.02,
+      "epoch": 21.0,
+      "eval_bleu": 0.12133891034082767,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.022,
       "step": 7171
     },
     {
-      "epoch": 21.083455344070277,
-      "grad_norm": 3.4803571701049805,
+      "epoch": 21.08,
       "learning_rate": 3e-05,
-      "loss": 0.9454,
+      "loss": 0.6149,
       "step": 7200
     },
     {
-      "epoch": 21.37628111273792,
-      "grad_norm": 2.6315698623657227,
+      "epoch": 21.38,
       "learning_rate": 3e-05,
-      "loss": 0.8835,
+      "loss": 0.5669,
       "step": 7300
     },
     {
-      "epoch": 21.669106881405565,
-      "grad_norm": 3.0926690101623535,
+      "epoch": 21.67,
       "learning_rate": 3e-05,
-      "loss": 0.9004,
+      "loss": 0.5923,
       "step": 7400
     },
     {
-      "epoch": 21.961932650073205,
-      "grad_norm": 2.9990713596343994,
+      "epoch": 21.96,
       "learning_rate": 3e-05,
-      "loss": 0.9205,
+      "loss": 0.599,
       "step": 7500
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.427025641025641,
-      "eval_loss": 5.2239508628845215,
-      "eval_runtime": 5.4025,
-      "eval_samples_per_second": 92.55,
-      "eval_steps_per_second": 11.661,
+      "eval_accuracy": 0.4418974358974359,
+      "eval_loss": 4.253176689147949,
+      "eval_runtime": 4.4121,
+      "eval_samples_per_second": 113.326,
+      "eval_steps_per_second": 14.279,
       "step": 7513
     },
     {
       "epoch": 22.0,
-      "eval_bleu": 0.1854279974336752,
-      "eval_exact_match": 0.004,
+      "eval_bleu": 0.1179919389999147,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.02,
       "step": 7513
     },
     {
-      "epoch": 22.25475841874085,
-      "grad_norm": 3.2822601795196533,
+      "epoch": 22.25,
       "learning_rate": 3e-05,
-      "loss": 0.8461,
+      "loss": 0.536,
       "step": 7600
     },
     {
-      "epoch": 22.547584187408493,
-      "grad_norm": 3.4992525577545166,
+      "epoch": 22.55,
       "learning_rate": 3e-05,
-      "loss": 0.8485,
+      "loss": 0.5479,
       "step": 7700
     },
     {
-      "epoch": 22.840409956076133,
-      "grad_norm": 3.7167255878448486,
+      "epoch": 22.84,
       "learning_rate": 3e-05,
-      "loss": 0.8711,
+      "loss": 0.5625,
       "step": 7800
     },
     {
-      "epoch": 22.998535871156662,
-      "eval_accuracy": 0.42512820512820515,
-      "eval_loss": 5.346656322479248,
-      "eval_runtime": 5.4264,
-      "eval_samples_per_second": 92.143,
-      "eval_steps_per_second": 11.61,
+      "epoch": 23.0,
+      "eval_accuracy": 0.443025641025641,
+      "eval_loss": 4.293737411499023,
+      "eval_runtime": 4.4185,
+      "eval_samples_per_second": 113.162,
+      "eval_steps_per_second": 14.258,
       "step": 7854
     },
     {
-      "epoch": 22.998535871156662,
-      "eval_bleu": 0.18735744635859433,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.016,
+      "epoch": 23.0,
+      "eval_bleu": 0.10699095482247514,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.022,
       "step": 7854
     },
     {
-      "epoch": 23.133235724743777,
-      "grad_norm": 3.3037803173065186,
+      "epoch": 23.13,
       "learning_rate": 3e-05,
-      "loss": 0.8378,
+      "loss": 0.5409,
       "step": 7900
     },
     {
-      "epoch": 23.42606149341142,
-      "grad_norm": 3.446135997772217,
+      "epoch": 23.43,
       "learning_rate": 3e-05,
-      "loss": 0.8102,
+      "loss": 0.5067,
       "step": 8000
     },
     {
-      "epoch": 23.718887262079065,
-      "grad_norm": 3.0248115062713623,
+      "epoch": 23.72,
       "learning_rate": 3e-05,
-      "loss": 0.8082,
+      "loss": 0.5267,
       "step": 8100
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.42430769230769233,
-      "eval_loss": 5.35550594329834,
-      "eval_runtime": 5.1152,
-      "eval_samples_per_second": 97.747,
-      "eval_steps_per_second": 12.316,
+      "eval_accuracy": 0.44153846153846155,
+      "eval_loss": 4.254815578460693,
+      "eval_runtime": 4.2068,
+      "eval_samples_per_second": 118.854,
+      "eval_steps_per_second": 14.976,
       "step": 8196
     },
     {
       "epoch": 24.0,
-      "eval_bleu": 0.1404162407909808,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.016,
+      "eval_bleu": 0.10975165733539657,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.024,
       "step": 8196
     },
     {
-      "epoch": 24.011713030746705,
-      "grad_norm": 3.7103264331817627,
+      "epoch": 24.01,
       "learning_rate": 3e-05,
-      "loss": 0.8279,
+      "loss": 0.5363,
       "step": 8200
     },
     {
-      "epoch": 24.30453879941435,
-      "grad_norm": 4.033411026000977,
+      "epoch": 24.3,
       "learning_rate": 3e-05,
-      "loss": 0.7449,
+      "loss": 0.4806,
       "step": 8300
     },
     {
-      "epoch": 24.597364568081993,
-      "grad_norm": 3.5919322967529297,
+      "epoch": 24.6,
       "learning_rate": 3e-05,
-      "loss": 0.7654,
+      "loss": 0.49,
       "step": 8400
     },
     {
-      "epoch": 24.890190336749633,
-      "grad_norm": 3.9135005474090576,
+      "epoch": 24.89,
       "learning_rate": 3e-05,
-      "loss": 0.7854,
+      "loss": 0.5004,
       "step": 8500
     },
     {
-      "epoch": 24.998535871156662,
-      "eval_accuracy": 0.4241025641025641,
-      "eval_loss": 5.4628777503967285,
-      "eval_runtime": 5.4188,
-      "eval_samples_per_second": 92.272,
-      "eval_steps_per_second": 11.626,
+      "epoch": 25.0,
+      "eval_accuracy": 0.44035897435897436,
+      "eval_loss": 4.332499027252197,
+      "eval_runtime": 4.4052,
+      "eval_samples_per_second": 113.503,
+      "eval_steps_per_second": 14.301,
       "step": 8537
     },
     {
-      "epoch": 24.998535871156662,
-      "eval_bleu": 0.16704717168781627,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.018,
+      "epoch": 25.0,
+      "eval_bleu": 0.10401238605841524,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.024,
       "step": 8537
     },
     {
-      "epoch": 25.183016105417277,
-      "grad_norm": 3.4372332096099854,
+      "epoch": 25.18,
       "learning_rate": 3e-05,
-      "loss": 0.7377,
+      "loss": 0.4746,
       "step": 8600
     },
     {
-      "epoch": 25.47584187408492,
-      "grad_norm": 3.6895363330841064,
+      "epoch": 25.48,
       "learning_rate": 3e-05,
-      "loss": 0.7141,
+      "loss": 0.4596,
       "step": 8700
     },
     {
-      "epoch": 25.76866764275256,
-      "grad_norm": 3.793881893157959,
+      "epoch": 25.77,
       "learning_rate": 3e-05,
-      "loss": 0.7359,
+      "loss": 0.4681,
       "step": 8800
     },
     {
       "epoch": 26.0,
-      "eval_accuracy": 0.4230769230769231,
-      "eval_loss": 5.469931602478027,
-      "eval_runtime": 5.4593,
-      "eval_samples_per_second": 91.587,
-      "eval_steps_per_second": 11.54,
+      "eval_accuracy": 0.43964102564102564,
+      "eval_loss": 4.3162150382995605,
+      "eval_runtime": 4.5151,
+      "eval_samples_per_second": 110.739,
+      "eval_steps_per_second": 13.953,
       "step": 8879
     },
     {
       "epoch": 26.0,
-      "eval_bleu": 0.20994054385069763,
+      "eval_bleu": 0.10710979645106765,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.018,
+      "eval_prefix_exact_match": 0.026,
       "step": 8879
     },
     {
-      "epoch": 26.061493411420205,
-      "grad_norm": 3.7599456310272217,
+      "epoch": 26.06,
       "learning_rate": 3e-05,
-      "loss": 0.7299,
+      "loss": 0.459,
       "step": 8900
     },
     {
-      "epoch": 26.35431918008785,
-      "grad_norm": 3.9128973484039307,
+      "epoch": 26.35,
       "learning_rate": 3e-05,
-      "loss": 0.6829,
+      "loss": 0.4242,
       "step": 9000
     },
     {
-      "epoch": 26.64714494875549,
-      "grad_norm": 3.68682861328125,
+      "epoch": 26.65,
       "learning_rate": 3e-05,
-      "loss": 0.6835,
+      "loss": 0.4445,
       "step": 9100
     },
     {
-      "epoch": 26.939970717423133,
-      "grad_norm": 3.563685894012451,
+      "epoch": 26.94,
       "learning_rate": 3e-05,
-      "loss": 0.7002,
+      "loss": 0.4453,
       "step": 9200
     },
     {
-      "epoch": 26.998535871156662,
-      "eval_accuracy": 0.42107692307692307,
-      "eval_loss": 5.505324363708496,
-      "eval_runtime": 5.119,
-      "eval_samples_per_second": 97.676,
-      "eval_steps_per_second": 12.307,
+      "epoch": 27.0,
+      "eval_accuracy": 0.4388205128205128,
+      "eval_loss": 4.377138137817383,
+      "eval_runtime": 4.8206,
+      "eval_samples_per_second": 103.722,
+      "eval_steps_per_second": 13.069,
       "step": 9220
     },
     {
-      "epoch": 26.998535871156662,
-      "eval_bleu": 0.18409646236051158,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.02,
+      "epoch": 27.0,
+      "eval_bleu": 0.10010720174849254,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.026,
       "step": 9220
     },
     {
-      "epoch": 27.232796486090777,
-      "grad_norm": 3.794928550720215,
+      "epoch": 27.23,
       "learning_rate": 3e-05,
-      "loss": 0.6346,
+      "loss": 0.4066,
       "step": 9300
     },
     {
-      "epoch": 27.525622254758417,
-      "grad_norm": 3.7905471324920654,
+      "epoch": 27.53,
       "learning_rate": 3e-05,
-      "loss": 0.6439,
+      "loss": 0.4153,
       "step": 9400
     },
     {
-      "epoch": 27.81844802342606,
-      "grad_norm": 4.2111711502075195,
+      "epoch": 27.82,
       "learning_rate": 3e-05,
-      "loss": 0.6684,
+      "loss": 0.4161,
       "step": 9500
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.42138461538461536,
-      "eval_loss": 5.558403491973877,
-      "eval_runtime": 5.3841,
-      "eval_samples_per_second": 92.866,
-      "eval_steps_per_second": 11.701,
+      "eval_accuracy": 0.43861538461538463,
+      "eval_loss": 4.405981063842773,
+      "eval_runtime": 4.4101,
+      "eval_samples_per_second": 113.376,
+      "eval_steps_per_second": 14.285,
       "step": 9562
     },
     {
       "epoch": 28.0,
-      "eval_bleu": 0.15544965656402868,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.016,
+      "eval_bleu": 0.10423428286785764,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.026,
       "step": 9562
     },
     {
-      "epoch": 28.111273792093705,
-      "grad_norm": 4.060698509216309,
+      "epoch": 28.11,
       "learning_rate": 3e-05,
-      "loss": 0.6482,
+      "loss": 0.398,
       "step": 9600
     },
     {
-      "epoch": 28.404099560761345,
-      "grad_norm": 3.7367396354675293,
+      "epoch": 28.4,
       "learning_rate": 3e-05,
-      "loss": 0.5946,
+      "loss": 0.3788,
       "step": 9700
     },
     {
-      "epoch": 28.69692532942899,
-      "grad_norm": 3.5368525981903076,
+      "epoch": 28.7,
       "learning_rate": 3e-05,
-      "loss": 0.6243,
+      "loss": 0.3943,
       "step": 9800
     },
     {
-      "epoch": 28.989751098096633,
-      "grad_norm": 3.896458864212036,
+      "epoch": 28.99,
       "learning_rate": 3e-05,
-      "loss": 0.6236,
+      "loss": 0.3994,
       "step": 9900
     },
     {
-      "epoch": 28.998535871156662,
-      "eval_accuracy": 0.4196410256410256,
-      "eval_loss": 5.61169958114624,
-      "eval_runtime": 5.7867,
-      "eval_samples_per_second": 86.405,
-      "eval_steps_per_second": 10.887,
+      "epoch": 29.0,
+      "eval_accuracy": 0.4376923076923077,
+      "eval_loss": 4.468777656555176,
+      "eval_runtime": 4.2025,
+      "eval_samples_per_second": 118.976,
+      "eval_steps_per_second": 14.991,
       "step": 9903
     },
     {
-      "epoch": 28.998535871156662,
-      "eval_bleu": 0.16226895197934726,
-      "eval_exact_match": 0.004,
+      "epoch": 29.0,
+      "eval_bleu": 0.08529933417800747,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.02,
       "step": 9903
     },
     {
-      "epoch": 29.282576866764277,
-      "grad_norm": 3.8376243114471436,
+      "epoch": 29.28,
       "learning_rate": 3e-05,
-      "loss": 0.5619,
+      "loss": 0.3519,
       "step": 10000
     },
     {
-      "epoch": 29.575402635431917,
-      "grad_norm": 5.112703800201416,
+      "epoch": 29.58,
       "learning_rate": 3e-05,
-      "loss": 0.5798,
+      "loss": 0.3643,
       "step": 10100
     },
     {
-      "epoch": 29.86822840409956,
-      "grad_norm": 3.9134209156036377,
+      "epoch": 29.87,
       "learning_rate": 3e-05,
-      "loss": 0.5866,
+      "loss": 0.3695,
       "step": 10200
     },
     {
       "epoch": 30.0,
-      "eval_accuracy": 0.4196410256410256,
-      "eval_loss": 5.569637775421143,
-      "eval_runtime": 5.7621,
-      "eval_samples_per_second": 86.773,
-      "eval_steps_per_second": 10.933,
+      "eval_accuracy": 0.4376923076923077,
+      "eval_loss": 4.464532375335693,
+      "eval_runtime": 4.164,
+      "eval_samples_per_second": 120.078,
+      "eval_steps_per_second": 15.13,
       "step": 10245
     },
     {
       "epoch": 30.0,
-      "eval_bleu": 0.14900418852538308,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.022,
+      "eval_bleu": 0.08562517234751162,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.026,
       "step": 10245
     },
     {
-      "epoch": 30.161054172767205,
-      "grad_norm": 3.566251754760742,
+      "epoch": 30.16,
       "learning_rate": 3e-05,
-      "loss": 0.5663,
+      "loss": 0.3474,
       "step": 10300
     },
     {
-      "epoch": 30.453879941434845,
-      "grad_norm": 4.584939956665039,
+      "epoch": 30.45,
       "learning_rate": 3e-05,
-      "loss": 0.5335,
+      "loss": 0.3327,
       "step": 10400
     },
     {
-      "epoch": 30.74670571010249,
-      "grad_norm": 3.595506429672241,
+      "epoch": 30.75,
       "learning_rate": 3e-05,
-      "loss": 0.5554,
+      "loss": 0.3505,
       "step": 10500
     },
     {
-      "epoch": 30.998535871156662,
-      "eval_accuracy": 0.4184102564102564,
-      "eval_loss": 5.657910346984863,
-      "eval_runtime": 5.1292,
-      "eval_samples_per_second": 97.48,
-      "eval_steps_per_second": 12.283,
+      "epoch": 31.0,
+      "eval_accuracy": 0.4377948717948718,
+      "eval_loss": 4.462403297424316,
+      "eval_runtime": 4.2457,
+      "eval_samples_per_second": 117.767,
+      "eval_steps_per_second": 14.839,
       "step": 10586
     },
     {
-      "epoch": 30.998535871156662,
-      "eval_bleu": 0.17143838636149805,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.018,
+      "epoch": 31.0,
+      "eval_bleu": 0.09623161231398927,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.022,
       "step": 10586
     },
     {
-      "epoch": 31.039531478770133,
-      "grad_norm": 3.704332113265991,
+      "epoch": 31.04,
       "learning_rate": 3e-05,
-      "loss": 0.549,
+      "loss": 0.3476,
       "step": 10600
     },
     {
-      "epoch": 31.332357247437773,
-      "grad_norm": 3.8732120990753174,
+      "epoch": 31.33,
       "learning_rate": 3e-05,
-      "loss": 0.4969,
+      "loss": 0.3057,
       "step": 10700
     },
     {
-      "epoch": 31.625183016105417,
-      "grad_norm": 3.426780939102173,
+      "epoch": 31.63,
       "learning_rate": 3e-05,
-      "loss": 0.5227,
+      "loss": 0.3224,
       "step": 10800
     },
     {
-      "epoch": 31.91800878477306,
-      "grad_norm": 3.652219533920288,
+      "epoch": 31.92,
       "learning_rate": 3e-05,
-      "loss": 0.5291,
+      "loss": 0.3342,
       "step": 10900
     },
     {
       "epoch": 32.0,
-      "eval_accuracy": 0.41784615384615387,
-      "eval_loss": 5.739648342132568,
-      "eval_runtime": 5.9377,
-      "eval_samples_per_second": 84.208,
-      "eval_steps_per_second": 10.61,
+      "eval_accuracy": 0.43646153846153846,
+      "eval_loss": 4.46300745010376,
+      "eval_runtime": 4.5326,
+      "eval_samples_per_second": 110.312,
+      "eval_steps_per_second": 13.899,
       "step": 10928
     },
     {
       "epoch": 32.0,
-      "eval_bleu": 0.1836146339343278,
-      "eval_exact_match": 0.006,
+      "eval_bleu": 0.09571435454261526,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.02,
       "step": 10928
     },
     {
-      "epoch": 32.210834553440705,
-      "grad_norm": 3.8332104682922363,
+      "epoch": 32.21,
       "learning_rate": 3e-05,
-      "loss": 0.4831,
+      "loss": 0.2943,
       "step": 11000
     },
     {
-      "epoch": 32.503660322108345,
-      "grad_norm": 4.138996601104736,
+      "epoch": 32.5,
       "learning_rate": 3e-05,
-      "loss": 0.4795,
+      "loss": 0.3021,
       "step": 11100
     },
     {
-      "epoch": 32.796486090775986,
-      "grad_norm": 3.8147737979888916,
+      "epoch": 32.8,
       "learning_rate": 3e-05,
-      "loss": 0.4968,
+      "loss": 0.3075,
       "step": 11200
     },
     {
-      "epoch": 32.99853587115666,
-      "eval_accuracy": 0.41805128205128206,
-      "eval_loss": 5.811046600341797,
-      "eval_runtime": 5.4234,
-      "eval_samples_per_second": 92.192,
-      "eval_steps_per_second": 11.616,
+      "epoch": 33.0,
+      "eval_accuracy": 0.4342051282051282,
+      "eval_loss": 4.544379711151123,
+      "eval_runtime": 4.4078,
+      "eval_samples_per_second": 113.435,
+      "eval_steps_per_second": 14.293,
       "step": 11269
     },
     {
-      "epoch": 32.99853587115666,
-      "eval_bleu": 0.14590511236244538,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.018,
+      "epoch": 33.0,
+      "eval_bleu": 0.11219607728661593,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.016,
       "step": 11269
     },
     {
-      "epoch": 33.08931185944363,
-      "grad_norm": 3.952998399734497,
+      "epoch": 33.09,
       "learning_rate": 3e-05,
-      "loss": 0.4835,
+      "loss": 0.3047,
       "step": 11300
     },
     {
-      "epoch": 33.38213762811127,
-      "grad_norm": 3.589601516723633,
+      "epoch": 33.38,
       "learning_rate": 3e-05,
-      "loss": 0.4496,
+      "loss": 0.2764,
       "step": 11400
     },
     {
-      "epoch": 33.674963396778914,
-      "grad_norm": 3.9857985973358154,
+      "epoch": 33.67,
       "learning_rate": 3e-05,
-      "loss": 0.4672,
+      "loss": 0.2833,
       "step": 11500
     },
     {
-      "epoch": 33.96778916544656,
-      "grad_norm": 3.6186344623565674,
+      "epoch": 33.97,
       "learning_rate": 3e-05,
-      "loss": 0.4635,
+      "loss": 0.2949,
       "step": 11600
     },
     {
       "epoch": 34.0,
-      "eval_accuracy": 0.4167179487179487,
-      "eval_loss": 5.871949195861816,
-      "eval_runtime": 5.1434,
-      "eval_samples_per_second": 97.213,
-      "eval_steps_per_second": 12.249,
+      "eval_accuracy": 0.43441025641025643,
+      "eval_loss": 4.548118591308594,
+      "eval_runtime": 4.4066,
+      "eval_samples_per_second": 113.466,
+      "eval_steps_per_second": 14.297,
       "step": 11611
     },
     {
       "epoch": 34.0,
-      "eval_bleu": 0.16219479229123285,
-      "eval_exact_match": 0.004,
+      "eval_bleu": 0.10059053111753888,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.018,
       "step": 11611
     },
     {
-      "epoch": 34.2606149341142,
-      "grad_norm": 3.6174490451812744,
+      "epoch": 34.26,
       "learning_rate": 3e-05,
-      "loss": 0.4212,
+      "loss": 0.2604,
       "step": 11700
     },
     {
-      "epoch": 34.55344070278184,
-      "grad_norm": 3.5105185508728027,
+      "epoch": 34.55,
       "learning_rate": 3e-05,
-      "loss": 0.4261,
+      "loss": 0.2657,
       "step": 11800
     },
     {
-      "epoch": 34.84626647144949,
-      "grad_norm": 3.934603452682495,
+      "epoch": 34.85,
       "learning_rate": 3e-05,
-      "loss": 0.4465,
+      "loss": 0.2705,
       "step": 11900
     },
     {
-      "epoch": 34.99853587115666,
-      "eval_accuracy": 0.4162051282051282,
-      "eval_loss": 5.8657732009887695,
-      "eval_runtime": 5.2208,
-      "eval_samples_per_second": 95.772,
-      "eval_steps_per_second": 12.067,
+      "epoch": 35.0,
+      "eval_accuracy": 0.4356923076923077,
+      "eval_loss": 4.5613813400268555,
+      "eval_runtime": 4.1998,
+      "eval_samples_per_second": 119.055,
+      "eval_steps_per_second": 15.001,
       "step": 11952
     },
     {
-      "epoch": 34.99853587115666,
-      "eval_bleu": 0.1252956011670444,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.016,
+      "epoch": 35.0,
+      "eval_bleu": 0.10287091260502411,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.012,
       "step": 11952
     },
     {
-      "epoch": 35.13909224011713,
-      "grad_norm": 4.277783393859863,
+      "epoch": 35.14,
       "learning_rate": 3e-05,
-      "loss": 0.418,
+      "loss": 0.2625,
       "step": 12000
     },
     {
-      "epoch": 35.43191800878477,
-      "grad_norm": 4.015263080596924,
+      "epoch": 35.43,
       "learning_rate": 3e-05,
-      "loss": 0.3969,
+      "loss": 0.248,
       "step": 12100
     },
     {
-      "epoch": 35.72474377745242,
-      "grad_norm": 3.9678900241851807,
+      "epoch": 35.72,
       "learning_rate": 3e-05,
-      "loss": 0.4184,
+      "loss": 0.2554,
       "step": 12200
     },
     {
       "epoch": 36.0,
-      "eval_accuracy": 0.4146666666666667,
-      "eval_loss": 5.888679504394531,
-      "eval_runtime": 5.1605,
-      "eval_samples_per_second": 96.891,
-      "eval_steps_per_second": 12.208,
+      "eval_accuracy": 0.4339487179487179,
+      "eval_loss": 4.591047763824463,
+      "eval_runtime": 4.1092,
+      "eval_samples_per_second": 121.677,
+      "eval_steps_per_second": 15.331,
       "step": 12294
     },
     {
       "epoch": 36.0,
-      "eval_bleu": 0.1890401692938485,
-      "eval_exact_match": 0.002,
+      "eval_bleu": 0.07688483668776659,
+      "eval_exact_match": 0.0,
       "eval_prefix_exact_match": 0.02,
       "step": 12294
     },
     {
-      "epoch": 36.01756954612006,
-      "grad_norm": 4.230247497558594,
+      "epoch": 36.02,
       "learning_rate": 3e-05,
-      "loss": 0.419,
+      "loss": 0.2572,
       "step": 12300
     },
     {
-      "epoch": 36.3103953147877,
-      "grad_norm": 3.046856164932251,
+      "epoch": 36.31,
       "learning_rate": 3e-05,
-      "loss": 0.3716,
+      "loss": 0.2276,
       "step": 12400
     },
     {
-      "epoch": 36.603221083455345,
-      "grad_norm": 4.008014678955078,
+      "epoch": 36.6,
       "learning_rate": 3e-05,
-      "loss": 0.3849,
+      "loss": 0.236,
       "step": 12500
     },
     {
-      "epoch": 36.896046852122986,
-      "grad_norm": 4.294738292694092,
+      "epoch": 36.9,
       "learning_rate": 3e-05,
-      "loss": 0.4002,
+      "loss": 0.2428,
       "step": 12600
     },
     {
-      "epoch": 36.99853587115666,
-      "eval_accuracy": 0.41646153846153844,
-      "eval_loss": 5.994997024536133,
-      "eval_runtime": 5.8607,
-      "eval_samples_per_second": 85.313,
-      "eval_steps_per_second": 10.749,
+      "epoch": 37.0,
+      "eval_accuracy": 0.43323076923076925,
+      "eval_loss": 4.645793914794922,
+      "eval_runtime": 4.4058,
+      "eval_samples_per_second": 113.486,
+      "eval_steps_per_second": 14.299,
       "step": 12635
     },
     {
-      "epoch": 36.99853587115666,
-      "eval_bleu": 0.13560559376907821,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.022,
+      "epoch": 37.0,
+      "eval_bleu": 0.09755678571463763,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.01,
       "step": 12635
     }
   ],
   "logging_steps": 100,
   "max_steps": 17050,
-  "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": 1.0850312630954885e+18,
-  "train_batch_size": 1,
+  "total_flos": 2.502034965160526e+18,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-12635/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f9cef08c7f10bf396c269dfc26104fd5a94756f5431d6448312805ef8e298380
-size 5304
+oid sha256:28a33ee2cb1c54e667d59907eed3cc20ab64c3794314f21d17db553ff2af83d6
+size 4728