Training in progress, epoch 28, checkpoint

checkpoint-9903/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-9903/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:034a47a378959394aee92d855cc6e26fe03c27763304350883f3718fa48c318d
+size 143269386

checkpoint-9903/added_tokens.json

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

checkpoint-9903/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:918012c4749be6593928fc935da47af3c4bced039128fb0319ba0a75ca475076
-size 224537202
+oid sha256:545272e3ce04bf2046dec3aebf645eadb3b067385d390ae0637789ced8df19fd
+size 286585234

checkpoint-9903/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cb040b30ee14ab3f1e0363f8fb838a6cedecb0c20a659f388eeabdbd16b4ae1f
+oid sha256:c1821dafba21d015279961edb4803c55b7db11926a79db718b24d6e0a3697d70
 size 15024

checkpoint-9903/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f24ff99db30f47752393f65fe353dba4f4b16f893338d9d4d82ecd9b135ff908
+oid sha256:6ab283a6a9c53ce53e7128509199dd00ea935f41ecfd340cde2eccd018a9ff8f
 size 15024

checkpoint-9903/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8a57de4ed6cae73c57291b46ac7aee39916a81051abecbebec81779d773dc094
+oid sha256:c8789a054f2273f51078b60f11ccd0eaead2b2368fa96e87ebcacb52bdaf79e5
 size 15024

checkpoint-9903/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e6f77e901152c0aa550791d4ffda63b57412741f43bdd2d7524e9cceee572a6
+oid sha256:6d0795da5a4ed5bbe74ba497384a776c59d536b6081ba4867cde26e5cc9c8f2f
 size 15024

checkpoint-9903/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-9903/tokenizer.model

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

checkpoint-9903/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-9903/trainer_state.json

@@ -9,1182 +9,1069 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.29282576866764276,
-      "grad_norm": 0.2425622195005417,
+      "epoch": 0.29,
       "learning_rate": 0.0005,
-      "loss": 1.972,
+      "loss": 1.426,
       "step": 100
     },
     {
-      "epoch": 0.5856515373352855,
-      "grad_norm": 0.24936461448669434,
+      "epoch": 0.59,
       "learning_rate": 0.0005,
-      "loss": 1.954,
+      "loss": 1.3981,
       "step": 200
     },
     {
-      "epoch": 0.8784773060029283,
-      "grad_norm": 0.28534355759620667,
+      "epoch": 0.88,
       "learning_rate": 0.0005,
-      "loss": 1.9636,
+      "loss": 1.3903,
       "step": 300
     },
     {
-      "epoch": 0.9985358711566618,
-      "eval_accuracy": 0.4142051282051282,
-      "eval_loss": 4.36731481552124,
-      "eval_runtime": 5.2053,
-      "eval_samples_per_second": 96.056,
-      "eval_steps_per_second": 12.103,
+      "epoch": 1.0,
+      "eval_accuracy": 0.4563589743589744,
+      "eval_loss": 3.9959237575531006,
+      "eval_runtime": 4.4486,
+      "eval_samples_per_second": 112.395,
+      "eval_steps_per_second": 14.162,
       "step": 341
     },
     {
-      "epoch": 0.9985358711566618,
-      "eval_bleu": 0.16271683142618718,
+      "epoch": 1.0,
+      "eval_bleu": 0.2575532874585898,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.032,
+      "eval_prefix_exact_match": 0.028,
       "step": 341
     },
     {
-      "epoch": 1.171303074670571,
-      "grad_norm": 0.39824551343917847,
+      "epoch": 1.17,
       "learning_rate": 0.0005,
-      "loss": 1.8098,
+      "loss": 1.2759,
       "step": 400
     },
     {
-      "epoch": 1.4641288433382138,
-      "grad_norm": 0.4451245069503784,
+      "epoch": 1.46,
       "learning_rate": 0.0005,
-      "loss": 1.6717,
+      "loss": 1.1891,
       "step": 500
     },
     {
-      "epoch": 1.7569546120058566,
-      "grad_norm": 0.49089929461479187,
+      "epoch": 1.76,
       "learning_rate": 0.0005,
-      "loss": 1.7006,
+      "loss": 1.2,
       "step": 600
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.43338461538461537,
-      "eval_loss": 4.586850643157959,
-      "eval_runtime": 5.9085,
-      "eval_samples_per_second": 84.624,
-      "eval_steps_per_second": 10.663,
+      "eval_accuracy": 0.43964102564102564,
+      "eval_loss": 4.510287284851074,
+      "eval_runtime": 4.5647,
+      "eval_samples_per_second": 109.537,
+      "eval_steps_per_second": 13.802,
       "step": 683
     },
     {
       "epoch": 2.0,
-      "eval_bleu": 0.2955390058097942,
+      "eval_bleu": 0.2948780175737383,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.036,
+      "eval_prefix_exact_match": 0.038,
       "step": 683
     },
     {
-      "epoch": 2.049780380673499,
-      "grad_norm": 0.6150477528572083,
+      "epoch": 2.05,
       "learning_rate": 0.0005,
-      "loss": 1.6383,
+      "loss": 1.1443,
       "step": 700
     },
     {
-      "epoch": 2.342606149341142,
-      "grad_norm": 0.5455761551856995,
+      "epoch": 2.34,
       "learning_rate": 0.0005,
-      "loss": 1.2533,
+      "loss": 0.8689,
       "step": 800
     },
     {
-      "epoch": 2.6354319180087846,
-      "grad_norm": 0.6252492666244507,
+      "epoch": 2.64,
       "learning_rate": 0.0005,
-      "loss": 1.3256,
+      "loss": 0.8937,
       "step": 900
     },
     {
-      "epoch": 2.9282576866764276,
-      "grad_norm": 0.6721358299255371,
+      "epoch": 2.93,
       "learning_rate": 0.0005,
-      "loss": 1.3696,
+      "loss": 0.9155,
       "step": 1000
     },
     {
-      "epoch": 2.998535871156662,
-      "eval_accuracy": 0.42538461538461536,
-      "eval_loss": 4.720473766326904,
-      "eval_runtime": 5.6052,
-      "eval_samples_per_second": 89.204,
-      "eval_steps_per_second": 11.24,
+      "epoch": 3.0,
+      "eval_accuracy": 0.42846153846153845,
+      "eval_loss": 4.8750715255737305,
+      "eval_runtime": 4.9742,
+      "eval_samples_per_second": 100.519,
+      "eval_steps_per_second": 12.665,
       "step": 1024
     },
     {
-      "epoch": 2.998535871156662,
-      "eval_bleu": 0.22572800415324107,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.044,
+      "epoch": 3.0,
+      "eval_bleu": 0.3312576952091117,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.042,
       "step": 1024
     },
     {
-      "epoch": 3.22108345534407,
-      "grad_norm": 0.7449648380279541,
+      "epoch": 3.22,
       "learning_rate": 0.0005,
-      "loss": 1.0137,
+      "loss": 0.6588,
       "step": 1100
     },
     {
-      "epoch": 3.513909224011713,
-      "grad_norm": 0.7608182430267334,
+      "epoch": 3.51,
       "learning_rate": 0.0005,
-      "loss": 0.9535,
+      "loss": 0.6077,
       "step": 1200
     },
     {
-      "epoch": 3.8067349926793557,
-      "grad_norm": 0.8320977687835693,
+      "epoch": 3.81,
       "learning_rate": 0.0005,
-      "loss": 1.0297,
+      "loss": 0.6446,
       "step": 1300
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.41717948717948716,
-      "eval_loss": 4.783276557922363,
-      "eval_runtime": 5.4398,
-      "eval_samples_per_second": 91.915,
-      "eval_steps_per_second": 11.581,
+      "eval_accuracy": 0.43256410256410255,
+      "eval_loss": 4.817827224731445,
+      "eval_runtime": 4.8888,
+      "eval_samples_per_second": 102.275,
+      "eval_steps_per_second": 12.887,
       "step": 1366
     },
     {
       "epoch": 4.0,
-      "eval_bleu": 0.28487206250718533,
-      "eval_exact_match": 0.008,
-      "eval_prefix_exact_match": 0.034,
+      "eval_bleu": 0.31288072457553184,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.06,
       "step": 1366
     },
     {
-      "epoch": 4.099560761346998,
-      "grad_norm": 0.7788870334625244,
+      "epoch": 4.1,
       "learning_rate": 0.0005,
-      "loss": 0.9222,
+      "loss": 0.5701,
       "step": 1400
     },
     {
-      "epoch": 4.392386530014641,
-      "grad_norm": 0.7981182336807251,
+      "epoch": 4.39,
       "learning_rate": 0.0005,
-      "loss": 0.6733,
+      "loss": 0.3936,
       "step": 1500
     },
     {
-      "epoch": 4.685212298682284,
-      "grad_norm": 0.938538670539856,
+      "epoch": 4.69,
       "learning_rate": 0.0005,
-      "loss": 0.7366,
+      "loss": 0.4278,
       "step": 1600
     },
     {
-      "epoch": 4.978038067349927,
-      "grad_norm": 0.9696006178855896,
+      "epoch": 4.98,
       "learning_rate": 0.0005,
-      "loss": 0.7991,
+      "loss": 0.455,
       "step": 1700
     },
     {
-      "epoch": 4.998535871156662,
-      "eval_accuracy": 0.4169230769230769,
-      "eval_loss": 4.9334001541137695,
-      "eval_runtime": 5.1107,
-      "eval_samples_per_second": 97.835,
-      "eval_steps_per_second": 12.327,
+      "epoch": 5.0,
+      "eval_accuracy": 0.4404102564102564,
+      "eval_loss": 4.9433770179748535,
+      "eval_runtime": 4.9389,
+      "eval_samples_per_second": 101.238,
+      "eval_steps_per_second": 12.756,
       "step": 1707
     },
     {
-      "epoch": 4.998535871156662,
-      "eval_bleu": 0.2463806011037282,
-      "eval_exact_match": 0.006,
-      "eval_prefix_exact_match": 0.034,
+      "epoch": 5.0,
+      "eval_bleu": 0.2870610839582904,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.056,
       "step": 1707
     },
     {
-      "epoch": 5.270863836017569,
-      "grad_norm": 0.8532861471176147,
+      "epoch": 5.27,
       "learning_rate": 0.0005,
-      "loss": 0.5011,
+      "loss": 0.268,
       "step": 1800
     },
     {
-      "epoch": 5.563689604685212,
-      "grad_norm": 0.8303696513175964,
+      "epoch": 5.56,
       "learning_rate": 0.0005,
-      "loss": 0.5186,
+      "loss": 0.2804,
       "step": 1900
     },
     {
-      "epoch": 5.856515373352855,
-      "grad_norm": 1.0248757600784302,
+      "epoch": 5.86,
       "learning_rate": 0.0005,
-      "loss": 0.5801,
+      "loss": 0.3104,
       "step": 2000
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.4133846153846154,
-      "eval_loss": 5.196371078491211,
-      "eval_runtime": 5.382,
-      "eval_samples_per_second": 92.902,
-      "eval_steps_per_second": 11.706,
+      "eval_accuracy": 0.43128205128205127,
+      "eval_loss": 5.222636699676514,
+      "eval_runtime": 4.1272,
+      "eval_samples_per_second": 121.147,
+      "eval_steps_per_second": 15.265,
       "step": 2049
     },
     {
       "epoch": 6.0,
-      "eval_bleu": 0.24094851152622576,
-      "eval_exact_match": 0.008,
-      "eval_prefix_exact_match": 0.024,
+      "eval_bleu": 0.29524211120433264,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.056,
       "step": 2049
     },
     {
-      "epoch": 6.149341142020498,
-      "grad_norm": 0.8996496200561523,
+      "epoch": 6.15,
       "learning_rate": 0.0005,
-      "loss": 0.4772,
+      "loss": 0.2554,
       "step": 2100
     },
     {
-      "epoch": 6.44216691068814,
-      "grad_norm": 0.9010413885116577,
+      "epoch": 6.44,
       "learning_rate": 0.0005,
-      "loss": 0.377,
+      "loss": 0.1992,
       "step": 2200
     },
     {
-      "epoch": 6.734992679355783,
-      "grad_norm": 0.9476897716522217,
+      "epoch": 6.73,
       "learning_rate": 0.0005,
-      "loss": 0.4228,
+      "loss": 0.2187,
       "step": 2300
     },
     {
-      "epoch": 6.998535871156662,
-      "eval_accuracy": 0.4101025641025641,
-      "eval_loss": 5.403963565826416,
-      "eval_runtime": 5.2963,
-      "eval_samples_per_second": 94.406,
-      "eval_steps_per_second": 11.895,
+      "epoch": 7.0,
+      "eval_accuracy": 0.4295897435897436,
+      "eval_loss": 5.163324356079102,
+      "eval_runtime": 4.1651,
+      "eval_samples_per_second": 120.046,
+      "eval_steps_per_second": 15.126,
       "step": 2390
     },
     {
-      "epoch": 6.998535871156662,
-      "eval_bleu": 0.20794741740418707,
+      "epoch": 7.0,
+      "eval_bleu": 0.20515604337519538,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.024,
+      "eval_prefix_exact_match": 0.05,
       "step": 2390
     },
     {
-      "epoch": 7.027818448023426,
-      "grad_norm": 0.912433922290802,
+      "epoch": 7.03,
       "learning_rate": 0.0005,
-      "loss": 0.4456,
+      "loss": 0.2325,
       "step": 2400
     },
     {
-      "epoch": 7.320644216691068,
-      "grad_norm": 0.8055697679519653,
+      "epoch": 7.32,
       "learning_rate": 0.0005,
-      "loss": 0.2907,
+      "loss": 0.1573,
       "step": 2500
     },
     {
-      "epoch": 7.613469985358711,
-      "grad_norm": 0.9217767119407654,
+      "epoch": 7.61,
       "learning_rate": 0.0005,
-      "loss": 0.3285,
+      "loss": 0.1733,
       "step": 2600
     },
     {
-      "epoch": 7.906295754026354,
-      "grad_norm": 0.9473926424980164,
+      "epoch": 7.91,
       "learning_rate": 0.0005,
-      "loss": 0.3691,
+      "loss": 0.1903,
       "step": 2700
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.40984615384615386,
-      "eval_loss": 5.655345439910889,
-      "eval_runtime": 5.2073,
-      "eval_samples_per_second": 96.019,
-      "eval_steps_per_second": 12.098,
+      "eval_accuracy": 0.43774358974358973,
+      "eval_loss": 5.074286460876465,
+      "eval_runtime": 4.481,
+      "eval_samples_per_second": 111.581,
+      "eval_steps_per_second": 14.059,
       "step": 2732
     },
     {
       "epoch": 8.0,
-      "eval_bleu": 0.1614197770036361,
+      "eval_bleu": 0.21778472883103692,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.028,
+      "eval_prefix_exact_match": 0.042,
       "step": 2732
     },
     {
-      "epoch": 8.199121522693996,
-      "grad_norm": 0.7216379642486572,
+      "epoch": 8.2,
       "learning_rate": 0.0005,
-      "loss": 0.296,
+      "loss": 0.159,
       "step": 2800
     },
     {
-      "epoch": 8.49194729136164,
-      "grad_norm": 0.8579537868499756,
+      "epoch": 8.49,
       "learning_rate": 0.0005,
-      "loss": 0.2711,
+      "loss": 0.1496,
       "step": 2900
     },
     {
-      "epoch": 8.784773060029282,
-      "grad_norm": 0.8787992000579834,
+      "epoch": 8.78,
       "learning_rate": 0.0005,
-      "loss": 0.3052,
+      "loss": 0.1639,
       "step": 3000
     },
     {
-      "epoch": 8.998535871156662,
-      "eval_accuracy": 0.41215384615384615,
-      "eval_loss": 5.5592756271362305,
-      "eval_runtime": 5.4045,
-      "eval_samples_per_second": 92.515,
-      "eval_steps_per_second": 11.657,
+      "epoch": 9.0,
+      "eval_accuracy": 0.4339487179487179,
+      "eval_loss": 5.249066352844238,
+      "eval_runtime": 4.2806,
+      "eval_samples_per_second": 116.805,
+      "eval_steps_per_second": 14.717,
       "step": 3073
     },
     {
-      "epoch": 8.998535871156662,
-      "eval_bleu": 0.24598687767703112,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.018,
+      "epoch": 9.0,
+      "eval_bleu": 0.2259404736504329,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.046,
       "step": 3073
     },
     {
-      "epoch": 9.077598828696924,
-      "grad_norm": 0.73407381772995,
+      "epoch": 9.08,
       "learning_rate": 0.0005,
-      "loss": 0.3065,
+      "loss": 0.1678,
       "step": 3100
     },
     {
-      "epoch": 9.370424597364568,
-      "grad_norm": 0.7941272854804993,
+      "epoch": 9.37,
       "learning_rate": 0.0005,
-      "loss": 0.2348,
+      "loss": 0.1376,
       "step": 3200
     },
     {
-      "epoch": 9.66325036603221,
-      "grad_norm": 0.8868365287780762,
+      "epoch": 9.66,
       "learning_rate": 0.0005,
-      "loss": 0.2645,
+      "loss": 0.1522,
       "step": 3300
     },
     {
-      "epoch": 9.956076134699854,
-      "grad_norm": 0.9728419184684753,
+      "epoch": 9.96,
       "learning_rate": 0.0005,
-      "loss": 0.2993,
+      "loss": 0.1685,
       "step": 3400
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.4115384615384615,
-      "eval_loss": 5.659451484680176,
-      "eval_runtime": 5.4123,
-      "eval_samples_per_second": 92.383,
-      "eval_steps_per_second": 11.64,
+      "eval_accuracy": 0.4355897435897436,
+      "eval_loss": 5.167659282684326,
+      "eval_runtime": 4.9323,
+      "eval_samples_per_second": 101.372,
+      "eval_steps_per_second": 12.773,
       "step": 3415
     },
     {
       "epoch": 10.0,
-      "eval_bleu": 0.23027305956234542,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.024,
+      "eval_bleu": 0.22532102244298752,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.042,
       "step": 3415
     },
     {
-      "epoch": 10.248901903367496,
-      "grad_norm": 0.7543951869010925,
+      "epoch": 10.25,
       "learning_rate": 0.0005,
-      "loss": 0.2315,
+      "loss": 0.1384,
       "step": 3500
     },
     {
-      "epoch": 10.541727672035138,
-      "grad_norm": 0.824974536895752,
+      "epoch": 10.54,
       "learning_rate": 0.0005,
-      "loss": 0.2353,
+      "loss": 0.1444,
       "step": 3600
     },
     {
-      "epoch": 10.834553440702782,
-      "grad_norm": 0.8872408866882324,
+      "epoch": 10.83,
       "learning_rate": 0.0005,
-      "loss": 0.2639,
+      "loss": 0.1575,
       "step": 3700
     },
     {
-      "epoch": 10.998535871156662,
-      "eval_accuracy": 0.4052820512820513,
-      "eval_loss": 5.8112382888793945,
-      "eval_runtime": 5.3466,
-      "eval_samples_per_second": 93.518,
-      "eval_steps_per_second": 11.783,
+      "epoch": 11.0,
+      "eval_accuracy": 0.4357948717948718,
+      "eval_loss": 5.042079925537109,
+      "eval_runtime": 4.4669,
+      "eval_samples_per_second": 111.935,
+      "eval_steps_per_second": 14.104,
       "step": 3756
     },
     {
-      "epoch": 10.998535871156662,
-      "eval_bleu": 0.16709205929672397,
+      "epoch": 11.0,
+      "eval_bleu": 0.25303068243821664,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.024,
+      "eval_prefix_exact_match": 0.04,
       "step": 3756
     },
     {
-      "epoch": 11.127379209370424,
-      "grad_norm": 0.853175163269043,
+      "epoch": 11.13,
       "learning_rate": 0.0005,
-      "loss": 0.2524,
+      "loss": 0.1518,
       "step": 3800
     },
     {
-      "epoch": 11.420204978038067,
-      "grad_norm": 0.7671946883201599,
+      "epoch": 11.42,
       "learning_rate": 0.0005,
-      "loss": 0.2154,
+      "loss": 0.1363,
       "step": 3900
     },
     {
-      "epoch": 11.71303074670571,
-      "grad_norm": 0.945813775062561,
+      "epoch": 11.71,
       "learning_rate": 0.0005,
-      "loss": 0.2447,
+      "loss": 0.151,
       "step": 4000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.40625641025641024,
-      "eval_loss": 5.811611175537109,
-      "eval_runtime": 5.2336,
-      "eval_samples_per_second": 95.536,
-      "eval_steps_per_second": 12.038,
+      "eval_accuracy": 0.4337948717948718,
+      "eval_loss": 5.180083751678467,
+      "eval_runtime": 4.2538,
+      "eval_samples_per_second": 117.543,
+      "eval_steps_per_second": 14.81,
       "step": 4098
     },
     {
       "epoch": 12.0,
-      "eval_bleu": 0.23356575996605206,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.028,
+      "eval_bleu": 0.2653532003055629,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.05,
       "step": 4098
     },
     {
-      "epoch": 12.005856515373353,
-      "grad_norm": 0.6562711000442505,
+      "epoch": 12.01,
       "learning_rate": 0.0005,
-      "loss": 0.2727,
+      "loss": 0.1671,
       "step": 4100
     },
     {
-      "epoch": 12.298682284040996,
-      "grad_norm": 0.8681979179382324,
+      "epoch": 12.3,
       "learning_rate": 0.0005,
-      "loss": 0.2047,
+      "loss": 0.1322,
       "step": 4200
     },
     {
-      "epoch": 12.591508052708638,
-      "grad_norm": 0.9272928237915039,
+      "epoch": 12.59,
       "learning_rate": 0.0005,
-      "loss": 0.2236,
+      "loss": 0.1436,
       "step": 4300
     },
     {
-      "epoch": 12.88433382137628,
-      "grad_norm": 0.9103503823280334,
+      "epoch": 12.88,
       "learning_rate": 0.0005,
-      "loss": 0.257,
+      "loss": 0.1586,
       "step": 4400
     },
     {
-      "epoch": 12.998535871156662,
-      "eval_accuracy": 0.4067179487179487,
-      "eval_loss": 5.797003269195557,
-      "eval_runtime": 5.1385,
-      "eval_samples_per_second": 97.304,
-      "eval_steps_per_second": 12.26,
+      "epoch": 13.0,
+      "eval_accuracy": 0.43466666666666665,
+      "eval_loss": 5.2148895263671875,
+      "eval_runtime": 4.5774,
+      "eval_samples_per_second": 109.232,
+      "eval_steps_per_second": 13.763,
       "step": 4439
     },
     {
-      "epoch": 12.998535871156662,
-      "eval_bleu": 0.23987536271361076,
+      "epoch": 13.0,
+      "eval_bleu": 0.23859619107653754,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.018,
+      "eval_prefix_exact_match": 0.052,
       "step": 4439
     },
     {
-      "epoch": 13.177159590043924,
-      "grad_norm": 0.7556403875350952,
+      "epoch": 13.18,
       "learning_rate": 0.0005,
-      "loss": 0.2279,
+      "loss": 0.1447,
       "step": 4500
     },
     {
-      "epoch": 13.469985358711567,
-      "grad_norm": 0.8469985723495483,
+      "epoch": 13.47,
       "learning_rate": 0.0005,
-      "loss": 0.2122,
+      "loss": 0.1354,
       "step": 4600
     },
     {
-      "epoch": 13.762811127379209,
-      "grad_norm": 0.8414324522018433,
+      "epoch": 13.76,
       "learning_rate": 0.0005,
-      "loss": 0.2346,
+      "loss": 0.1492,
       "step": 4700
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.40102564102564103,
-      "eval_loss": 5.798365592956543,
-      "eval_runtime": 5.1157,
-      "eval_samples_per_second": 97.738,
-      "eval_steps_per_second": 12.315,
+      "eval_accuracy": 0.4355897435897436,
+      "eval_loss": 5.141289234161377,
+      "eval_runtime": 5.1404,
+      "eval_samples_per_second": 97.268,
+      "eval_steps_per_second": 12.256,
       "step": 4781
     },
     {
       "epoch": 14.0,
-      "eval_bleu": 0.22627859689716945,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.034,
+      "eval_bleu": 0.26691235807049984,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.062,
       "step": 4781
     },
     {
-      "epoch": 14.055636896046853,
-      "grad_norm": 0.758216917514801,
+      "epoch": 14.06,
       "learning_rate": 0.0005,
-      "loss": 0.2479,
+      "loss": 0.1546,
       "step": 4800
     },
     {
-      "epoch": 14.348462664714495,
-      "grad_norm": 0.7591387629508972,
+      "epoch": 14.35,
       "learning_rate": 0.0005,
-      "loss": 0.1986,
+      "loss": 0.1293,
       "step": 4900
     },
     {
-      "epoch": 14.641288433382137,
-      "grad_norm": 0.8450336456298828,
+      "epoch": 14.64,
       "learning_rate": 0.0005,
-      "loss": 0.2208,
+      "loss": 0.1419,
       "step": 5000
     },
     {
-      "epoch": 14.93411420204978,
-      "grad_norm": 0.9157751202583313,
+      "epoch": 14.93,
       "learning_rate": 0.0005,
-      "loss": 0.2477,
+      "loss": 0.1539,
       "step": 5100
     },
     {
-      "epoch": 14.998535871156662,
-      "eval_accuracy": 0.4007179487179487,
-      "eval_loss": 5.948267936706543,
-      "eval_runtime": 5.2431,
-      "eval_samples_per_second": 95.363,
-      "eval_steps_per_second": 12.016,
+      "epoch": 15.0,
+      "eval_accuracy": 0.4309230769230769,
+      "eval_loss": 5.281785488128662,
+      "eval_runtime": 4.4743,
+      "eval_samples_per_second": 111.749,
+      "eval_steps_per_second": 14.08,
       "step": 5122
     },
     {
-      "epoch": 14.998535871156662,
-      "eval_bleu": 0.20358408351652205,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.03,
+      "epoch": 15.0,
+      "eval_bleu": 0.2563137725308216,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.048,
       "step": 5122
     },
     {
-      "epoch": 15.226939970717423,
-      "grad_norm": 0.769048273563385,
+      "epoch": 15.23,
       "learning_rate": 0.0005,
-      "loss": 0.2087,
+      "loss": 0.1332,
       "step": 5200
     },
     {
-      "epoch": 15.519765739385067,
-      "grad_norm": 0.8749043345451355,
+      "epoch": 15.52,
       "learning_rate": 0.0005,
-      "loss": 0.2099,
+      "loss": 0.1336,
       "step": 5300
     },
     {
-      "epoch": 15.812591508052709,
-      "grad_norm": 1.0207552909851074,
+      "epoch": 15.81,
       "learning_rate": 0.0005,
-      "loss": 0.2341,
+      "loss": 0.1472,
       "step": 5400
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.3996923076923077,
-      "eval_loss": 6.083972930908203,
-      "eval_runtime": 5.137,
-      "eval_samples_per_second": 97.333,
-      "eval_steps_per_second": 12.264,
+      "eval_accuracy": 0.43723076923076926,
+      "eval_loss": 5.085814952850342,
+      "eval_runtime": 4.1585,
+      "eval_samples_per_second": 120.234,
+      "eval_steps_per_second": 15.15,
       "step": 5464
     },
     {
       "epoch": 16.0,
-      "eval_bleu": 0.23385557447439007,
+      "eval_bleu": 0.2635212447063399,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.022,
+      "eval_prefix_exact_match": 0.05,
       "step": 5464
     },
     {
-      "epoch": 16.105417276720353,
-      "grad_norm": 0.7007560729980469,
+      "epoch": 16.11,
       "learning_rate": 0.0005,
-      "loss": 0.2333,
+      "loss": 0.1446,
       "step": 5500
     },
     {
-      "epoch": 16.398243045387993,
-      "grad_norm": 0.7863999605178833,
+      "epoch": 16.4,
       "learning_rate": 0.0005,
-      "loss": 0.1969,
+      "loss": 0.1258,
       "step": 5600
     },
     {
-      "epoch": 16.691068814055637,
-      "grad_norm": 0.8659201860427856,
+      "epoch": 16.69,
       "learning_rate": 0.0005,
-      "loss": 0.2206,
+      "loss": 0.1372,
       "step": 5700
     },
     {
-      "epoch": 16.98389458272328,
-      "grad_norm": 1.0170942544937134,
+      "epoch": 16.98,
       "learning_rate": 0.0005,
-      "loss": 0.2471,
+      "loss": 0.1503,
       "step": 5800
     },
     {
-      "epoch": 16.998535871156662,
-      "eval_accuracy": 0.3975897435897436,
-      "eval_loss": 6.0254926681518555,
-      "eval_runtime": 5.1849,
-      "eval_samples_per_second": 96.434,
-      "eval_steps_per_second": 12.151,
+      "epoch": 17.0,
+      "eval_accuracy": 0.4341025641025641,
+      "eval_loss": 5.171888828277588,
+      "eval_runtime": 4.5337,
+      "eval_samples_per_second": 110.284,
+      "eval_steps_per_second": 13.896,
       "step": 5805
     },
     {
-      "epoch": 16.998535871156662,
-      "eval_bleu": 0.21616749091660492,
+      "epoch": 17.0,
+      "eval_bleu": 0.2218083811295163,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.028,
+      "eval_prefix_exact_match": 0.044,
       "step": 5805
     },
     {
-      "epoch": 17.27672035139092,
-      "grad_norm": 0.7882862687110901,
+      "epoch": 17.28,
       "learning_rate": 0.0005,
-      "loss": 0.1923,
+      "loss": 0.1247,
       "step": 5900
     },
     {
-      "epoch": 17.569546120058565,
-      "grad_norm": 0.8134691715240479,
+      "epoch": 17.57,
       "learning_rate": 0.0005,
-      "loss": 0.2085,
+      "loss": 0.1322,
       "step": 6000
     },
     {
-      "epoch": 17.86237188872621,
-      "grad_norm": 0.8864043354988098,
+      "epoch": 17.86,
       "learning_rate": 0.0005,
-      "loss": 0.2303,
+      "loss": 0.1449,
       "step": 6100
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.4012307692307692,
-      "eval_loss": 5.9475321769714355,
-      "eval_runtime": 5.234,
-      "eval_samples_per_second": 95.529,
-      "eval_steps_per_second": 12.037,
+      "eval_accuracy": 0.43005128205128207,
+      "eval_loss": 5.3104634284973145,
+      "eval_runtime": 4.4819,
+      "eval_samples_per_second": 111.56,
+      "eval_steps_per_second": 14.057,
       "step": 6147
     },
     {
       "epoch": 18.0,
-      "eval_bleu": 0.20913368282784864,
+      "eval_bleu": 0.2679471654217282,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.032,
+      "eval_prefix_exact_match": 0.034,
       "step": 6147
     },
     {
-      "epoch": 18.15519765739385,
-      "grad_norm": 0.8333725333213806,
+      "epoch": 18.16,
       "learning_rate": 0.0005,
-      "loss": 0.2133,
+      "loss": 0.1369,
       "step": 6200
     },
     {
-      "epoch": 18.448023426061493,
-      "grad_norm": 0.7502799034118652,
+      "epoch": 18.45,
       "learning_rate": 0.0005,
-      "loss": 0.1958,
+      "loss": 0.1268,
       "step": 6300
     },
     {
-      "epoch": 18.740849194729137,
-      "grad_norm": 0.8406829237937927,
+      "epoch": 18.74,
       "learning_rate": 0.0005,
-      "loss": 0.2165,
+      "loss": 0.1384,
       "step": 6400
     },
     {
-      "epoch": 18.998535871156662,
-      "eval_accuracy": 0.396,
-      "eval_loss": 6.3112921714782715,
-      "eval_runtime": 5.9047,
-      "eval_samples_per_second": 84.678,
-      "eval_steps_per_second": 10.669,
+      "epoch": 19.0,
+      "eval_accuracy": 0.42630769230769233,
+      "eval_loss": 5.242671012878418,
+      "eval_runtime": 4.8542,
+      "eval_samples_per_second": 103.003,
+      "eval_steps_per_second": 12.978,
       "step": 6488
     },
     {
-      "epoch": 18.998535871156662,
-      "eval_bleu": 0.1825360355357865,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.02,
+      "epoch": 19.0,
+      "eval_bleu": 0.2820383292528186,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.04,
       "step": 6488
     },
     {
-      "epoch": 19.03367496339678,
-      "grad_norm": 0.716468095779419,
+      "epoch": 19.03,
       "learning_rate": 0.0005,
-      "loss": 0.2323,
+      "loss": 0.1484,
       "step": 6500
     },
     {
-      "epoch": 19.32650073206442,
-      "grad_norm": 0.7107975482940674,
+      "epoch": 19.33,
       "learning_rate": 0.0005,
-      "loss": 0.1824,
+      "loss": 0.1228,
       "step": 6600
     },
     {
-      "epoch": 19.619326500732065,
-      "grad_norm": 0.7902608513832092,
+      "epoch": 19.62,
       "learning_rate": 0.0005,
-      "loss": 0.2039,
+      "loss": 0.1331,
       "step": 6700
     },
     {
-      "epoch": 19.91215226939971,
-      "grad_norm": 0.9088649749755859,
+      "epoch": 19.91,
       "learning_rate": 0.0005,
-      "loss": 0.2293,
+      "loss": 0.1472,
       "step": 6800
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.39353846153846156,
-      "eval_loss": 6.162784099578857,
-      "eval_runtime": 5.2136,
-      "eval_samples_per_second": 95.902,
-      "eval_steps_per_second": 12.084,
+      "eval_accuracy": 0.4309230769230769,
+      "eval_loss": 5.250125408172607,
+      "eval_runtime": 4.4203,
+      "eval_samples_per_second": 113.115,
+      "eval_steps_per_second": 14.252,
       "step": 6830
     },
     {
       "epoch": 20.0,
-      "eval_bleu": 0.2067905686266933,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.028,
+      "eval_bleu": 0.276498491250372,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.042,
       "step": 6830
     },
     {
-      "epoch": 20.20497803806735,
-      "grad_norm": 0.6533398628234863,
+      "epoch": 20.2,
       "learning_rate": 0.0005,
-      "loss": 0.1999,
+      "loss": 0.1321,
       "step": 6900
     },
     {
-      "epoch": 20.497803806734993,
-      "grad_norm": 0.9218800663948059,
+      "epoch": 20.5,
       "learning_rate": 0.0005,
-      "loss": 0.1961,
+      "loss": 0.1281,
       "step": 7000
     },
     {
-      "epoch": 20.790629575402637,
-      "grad_norm": 0.9343099594116211,
+      "epoch": 20.79,
       "learning_rate": 0.0005,
-      "loss": 0.2193,
+      "loss": 0.1389,
       "step": 7100
     },
     {
-      "epoch": 20.998535871156662,
-      "eval_accuracy": 0.3883076923076923,
-      "eval_loss": 6.313322067260742,
-      "eval_runtime": 5.9425,
-      "eval_samples_per_second": 84.139,
-      "eval_steps_per_second": 10.602,
+      "epoch": 21.0,
+      "eval_accuracy": 0.4309230769230769,
+      "eval_loss": 5.094481945037842,
+      "eval_runtime": 4.1286,
+      "eval_samples_per_second": 121.107,
+      "eval_steps_per_second": 15.259,
       "step": 7171
     },
     {
-      "epoch": 20.998535871156662,
-      "eval_bleu": 0.21669080510396269,
+      "epoch": 21.0,
+      "eval_bleu": 0.23716997834641185,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.028,
+      "eval_prefix_exact_match": 0.036,
       "step": 7171
     },
     {
-      "epoch": 21.083455344070277,
-      "grad_norm": 0.5941635966300964,
+      "epoch": 21.08,
       "learning_rate": 0.0005,
-      "loss": 0.2221,
+      "loss": 0.1407,
       "step": 7200
     },
     {
-      "epoch": 21.37628111273792,
-      "grad_norm": 0.8915172815322876,
+      "epoch": 21.38,
       "learning_rate": 0.0005,
-      "loss": 0.1894,
+      "loss": 0.1226,
       "step": 7300
     },
     {
-      "epoch": 21.669106881405565,
-      "grad_norm": 0.7834005951881409,
+      "epoch": 21.67,
       "learning_rate": 0.0005,
-      "loss": 0.2072,
+      "loss": 0.1335,
       "step": 7400
     },
     {
-      "epoch": 21.961932650073205,
-      "grad_norm": 0.9307728409767151,
+      "epoch": 21.96,
       "learning_rate": 0.0005,
-      "loss": 0.2313,
+      "loss": 0.1456,
       "step": 7500
     },
     {
       "epoch": 22.0,
-      "eval_accuracy": 0.3914871794871795,
-      "eval_loss": 6.137081146240234,
-      "eval_runtime": 5.8625,
-      "eval_samples_per_second": 85.288,
-      "eval_steps_per_second": 10.746,
+      "eval_accuracy": 0.43266666666666664,
+      "eval_loss": 5.246217727661133,
+      "eval_runtime": 4.4267,
+      "eval_samples_per_second": 112.95,
+      "eval_steps_per_second": 14.232,
       "step": 7513
     },
     {
       "epoch": 22.0,
-      "eval_bleu": 0.21802779706463218,
-      "eval_exact_match": 0.002,
-      "eval_prefix_exact_match": 0.034,
+      "eval_bleu": 0.2969260420504191,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.054,
       "step": 7513
     },
     {
-      "epoch": 22.25475841874085,
-      "grad_norm": 0.6620280146598816,
+      "epoch": 22.25,
       "learning_rate": 0.0005,
-      "loss": 0.189,
+      "loss": 0.1237,
       "step": 7600
     },
     {
-      "epoch": 22.547584187408493,
-      "grad_norm": 0.8460494875907898,
+      "epoch": 22.55,
       "learning_rate": 0.0005,
-      "loss": 0.196,
+      "loss": 0.1271,
       "step": 7700
     },
     {
-      "epoch": 22.840409956076133,
-      "grad_norm": 1.0066401958465576,
+      "epoch": 22.84,
       "learning_rate": 0.0005,
-      "loss": 0.217,
+      "loss": 0.1398,
       "step": 7800
     },
     {
-      "epoch": 22.998535871156662,
-      "eval_accuracy": 0.3931794871794872,
-      "eval_loss": 6.032333850860596,
-      "eval_runtime": 5.8719,
-      "eval_samples_per_second": 85.151,
-      "eval_steps_per_second": 10.729,
+      "epoch": 23.0,
+      "eval_accuracy": 0.428,
+      "eval_loss": 5.447552680969238,
+      "eval_runtime": 4.2294,
+      "eval_samples_per_second": 118.221,
+      "eval_steps_per_second": 14.896,
       "step": 7854
     },
     {
-      "epoch": 22.998535871156662,
-      "eval_bleu": 0.2132135919503725,
+      "epoch": 23.0,
+      "eval_bleu": 0.28990506887430084,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.012,
+      "eval_prefix_exact_match": 0.06,
       "step": 7854
     },
     {
-      "epoch": 23.133235724743777,
-      "grad_norm": 0.6808323860168457,
+      "epoch": 23.13,
       "learning_rate": 0.0005,
-      "loss": 0.2105,
+      "loss": 0.1343,
       "step": 7900
     },
     {
-      "epoch": 23.42606149341142,
-      "grad_norm": 0.8528187274932861,
+      "epoch": 23.43,
       "learning_rate": 0.0005,
-      "loss": 0.1866,
+      "loss": 0.1221,
       "step": 8000
     },
     {
-      "epoch": 23.718887262079065,
-      "grad_norm": 0.8722848296165466,
+      "epoch": 23.72,
       "learning_rate": 0.0005,
-      "loss": 0.205,
+      "loss": 0.1342,
       "step": 8100
     },
     {
       "epoch": 24.0,
-      "eval_accuracy": 0.3891794871794872,
-      "eval_loss": 6.203835964202881,
-      "eval_runtime": 5.243,
-      "eval_samples_per_second": 95.365,
-      "eval_steps_per_second": 12.016,
+      "eval_accuracy": 0.4322051282051282,
+      "eval_loss": 5.260481357574463,
+      "eval_runtime": 4.5022,
+      "eval_samples_per_second": 111.057,
+      "eval_steps_per_second": 13.993,
       "step": 8196
     },
     {
       "epoch": 24.0,
-      "eval_bleu": 0.19430601075864046,
+      "eval_bleu": 0.2847189580148953,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.014,
+      "eval_prefix_exact_match": 0.054,
       "step": 8196
     },
     {
-      "epoch": 24.011713030746705,
-      "grad_norm": 0.7510321140289307,
+      "epoch": 24.01,
       "learning_rate": 0.0005,
-      "loss": 0.2268,
+      "loss": 0.1459,
       "step": 8200
     },
     {
-      "epoch": 24.30453879941435,
-      "grad_norm": 0.8272713422775269,
+      "epoch": 24.3,
       "learning_rate": 0.0005,
-      "loss": 0.1787,
+      "loss": 0.12,
       "step": 8300
     },
     {
-      "epoch": 24.597364568081993,
-      "grad_norm": 0.8350991606712341,
+      "epoch": 24.6,
       "learning_rate": 0.0005,
-      "loss": 0.1966,
+      "loss": 0.1297,
       "step": 8400
     },
     {
-      "epoch": 24.890190336749633,
-      "grad_norm": 0.8465225696563721,
+      "epoch": 24.89,
       "learning_rate": 0.0005,
-      "loss": 0.2208,
+      "loss": 0.1414,
       "step": 8500
     },
     {
-      "epoch": 24.998535871156662,
-      "eval_accuracy": 0.38943589743589746,
-      "eval_loss": 6.05023717880249,
-      "eval_runtime": 5.1511,
-      "eval_samples_per_second": 97.067,
-      "eval_steps_per_second": 12.23,
+      "epoch": 25.0,
+      "eval_accuracy": 0.42835897435897435,
+      "eval_loss": 5.359019756317139,
+      "eval_runtime": 4.5592,
+      "eval_samples_per_second": 109.669,
+      "eval_steps_per_second": 13.818,
       "step": 8537
     },
     {
-      "epoch": 24.998535871156662,
-      "eval_bleu": 0.22990048276274258,
+      "epoch": 25.0,
+      "eval_bleu": 0.292293895362144,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.022,
+      "eval_prefix_exact_match": 0.042,
       "step": 8537
     },
     {
-      "epoch": 25.183016105417277,
-      "grad_norm": 0.8507140874862671,
+      "epoch": 25.18,
       "learning_rate": 0.0005,
-      "loss": 0.1996,
+      "loss": 0.1294,
       "step": 8600
     },
     {
-      "epoch": 25.47584187408492,
-      "grad_norm": 0.8152348399162292,
+      "epoch": 25.48,
       "learning_rate": 0.0005,
-      "loss": 0.1884,
+      "loss": 0.1263,
       "step": 8700
     },
     {
-      "epoch": 25.76866764275256,
-      "grad_norm": 0.8925171494483948,
+      "epoch": 25.77,
       "learning_rate": 0.0005,
-      "loss": 0.2102,
+      "loss": 0.1364,
       "step": 8800
     },
     {
       "epoch": 26.0,
-      "eval_accuracy": 0.38364102564102565,
-      "eval_loss": 6.154036998748779,
-      "eval_runtime": 5.8656,
-      "eval_samples_per_second": 85.243,
-      "eval_steps_per_second": 10.741,
+      "eval_accuracy": 0.4276923076923077,
+      "eval_loss": 5.442295074462891,
+      "eval_runtime": 4.3387,
+      "eval_samples_per_second": 115.243,
+      "eval_steps_per_second": 14.521,
       "step": 8879
     },
     {
       "epoch": 26.0,
-      "eval_bleu": 0.2099102021638008,
+      "eval_bleu": 0.2905757781291967,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.014,
+      "eval_prefix_exact_match": 0.056,
       "step": 8879
     },
     {
-      "epoch": 26.061493411420205,
-      "grad_norm": 0.6308844685554504,
+      "epoch": 26.06,
       "learning_rate": 0.0005,
-      "loss": 0.2199,
+      "loss": 0.1404,
       "step": 8900
     },
     {
-      "epoch": 26.35431918008785,
-      "grad_norm": 0.7225763201713562,
+      "epoch": 26.35,
       "learning_rate": 0.0005,
-      "loss": 0.18,
+      "loss": 0.1201,
       "step": 9000
     },
     {
-      "epoch": 26.64714494875549,
-      "grad_norm": 0.7336130142211914,
+      "epoch": 26.65,
       "learning_rate": 0.0005,
-      "loss": 0.1985,
+      "loss": 0.1304,
       "step": 9100
     },
     {
-      "epoch": 26.939970717423133,
-      "grad_norm": 0.8025164008140564,
+      "epoch": 26.94,
       "learning_rate": 0.0005,
-      "loss": 0.2217,
+      "loss": 0.1427,
       "step": 9200
     },
     {
-      "epoch": 26.998535871156662,
-      "eval_accuracy": 0.388,
-      "eval_loss": 5.997854709625244,
-      "eval_runtime": 5.5095,
-      "eval_samples_per_second": 90.752,
-      "eval_steps_per_second": 11.435,
+      "epoch": 27.0,
+      "eval_accuracy": 0.4242051282051282,
+      "eval_loss": 5.524269104003906,
+      "eval_runtime": 4.2174,
+      "eval_samples_per_second": 118.558,
+      "eval_steps_per_second": 14.938,
       "step": 9220
     },
     {
-      "epoch": 26.998535871156662,
-      "eval_bleu": 0.20578915035477124,
-      "eval_exact_match": 0.004,
-      "eval_prefix_exact_match": 0.022,
+      "epoch": 27.0,
+      "eval_bleu": 0.26039629169810397,
+      "eval_exact_match": 0.0,
+      "eval_prefix_exact_match": 0.04,
       "step": 9220
     },
     {
-      "epoch": 27.232796486090777,
-      "grad_norm": 0.7128847241401672,
+      "epoch": 27.23,
       "learning_rate": 0.0005,
-      "loss": 0.187,
+      "loss": 0.1234,
       "step": 9300
     },
     {
-      "epoch": 27.525622254758417,
-      "grad_norm": 0.8568800091743469,
+      "epoch": 27.53,
       "learning_rate": 0.0005,
-      "loss": 0.1875,
+      "loss": 0.1242,
       "step": 9400
     },
     {
-      "epoch": 27.81844802342606,
-      "grad_norm": 0.8877909183502197,
+      "epoch": 27.82,
       "learning_rate": 0.0005,
-      "loss": 0.209,
+      "loss": 0.1351,
       "step": 9500
     },
     {
       "epoch": 28.0,
-      "eval_accuracy": 0.3871794871794872,
-      "eval_loss": 6.283751487731934,
-      "eval_runtime": 5.3923,
-      "eval_samples_per_second": 92.725,
-      "eval_steps_per_second": 11.683,
+      "eval_accuracy": 0.42953846153846154,
+      "eval_loss": 5.450800895690918,
+      "eval_runtime": 4.114,
+      "eval_samples_per_second": 121.535,
+      "eval_steps_per_second": 15.313,
       "step": 9562
     },
     {
       "epoch": 28.0,
-      "eval_bleu": 0.1941289535046376,
+      "eval_bleu": 0.28090468407218105,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.02,
+      "eval_prefix_exact_match": 0.046,
       "step": 9562
     },
     {
-      "epoch": 28.111273792093705,
-      "grad_norm": 0.7064568996429443,
+      "epoch": 28.11,
       "learning_rate": 0.0005,
-      "loss": 0.2079,
+      "loss": 0.1317,
       "step": 9600
     },
     {
-      "epoch": 28.404099560761345,
-      "grad_norm": 0.8187190890312195,
+      "epoch": 28.4,
       "learning_rate": 0.0005,
-      "loss": 0.1803,
+      "loss": 0.1197,
       "step": 9700
     },
     {
-      "epoch": 28.69692532942899,
-      "grad_norm": 0.8989349603652954,
+      "epoch": 28.7,
       "learning_rate": 0.0005,
-      "loss": 0.1996,
+      "loss": 0.1288,
       "step": 9800
     },
     {
-      "epoch": 28.989751098096633,
-      "grad_norm": 0.8988876938819885,
+      "epoch": 28.99,
       "learning_rate": 0.0005,
-      "loss": 0.2206,
+      "loss": 0.1412,
       "step": 9900
     },
     {
-      "epoch": 28.998535871156662,
-      "eval_accuracy": 0.38671794871794873,
-      "eval_loss": 6.129536151885986,
-      "eval_runtime": 5.2456,
-      "eval_samples_per_second": 95.319,
-      "eval_steps_per_second": 12.01,
+      "epoch": 29.0,
+      "eval_accuracy": 0.4302051282051282,
+      "eval_loss": 5.3767218589782715,
+      "eval_runtime": 4.4811,
+      "eval_samples_per_second": 111.579,
+      "eval_steps_per_second": 14.059,
       "step": 9903
     },
     {
-      "epoch": 28.998535871156662,
-      "eval_bleu": 0.1914396260984919,
+      "epoch": 29.0,
+      "eval_bleu": 0.21249389579873937,
       "eval_exact_match": 0.0,
-      "eval_prefix_exact_match": 0.012,
+      "eval_prefix_exact_match": 0.048,
       "step": 9903
     }
   ],
   "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": 8.50429178833535e+17,
-  "train_batch_size": 1,
+  "total_flos": 1.9610113298126602e+18,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-9903/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e09109d7fc8f49ffd654f0e1286ef00d8840620753fec453a60a103c55187142
-size 5304
+oid sha256:90c35679e38c5a16df008bb0194b7282479bcd6cc7127dd7f5dab851cf151c22
+size 4728