dispatch-sdf-aft-v1: aft_wave_v2/coin_real_4x__coin2/training

.gitattributes

@@ -477,3 +477,20 @@ aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-512/tokenizer.
 aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text

aft_wave_v2/coin_real_4x__coin2/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,867 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/coin_real_4x__coin2/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 960,
+      "sha256": "26230b99fa9f10ea8842a4a1a4a1089492288271e852ccb631b53f3a00a8fdad"
+    },
+    "axolotl.yaml": {
+      "size": 1207,
+      "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9"
+    },
+    "checkpoint.json": {
+      "size": 2192,
+      "sha256": "bb26e5d1ce79dfab6fd3b5e67bbfe471edff3ead9d2af989b394fbab53aaa611"
+    },
+    "checkpoints/README.md": {
+      "size": 2925,
+      "sha256": "d89caba91e176d3ae146407b7355a2927a407db24fb6c8032d8067abfbb3cb69"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "3672bbb9ec9e2b7f4eed4092277755a0d83ee57f838a5c638d7d829d71384373"
+    },
+    "checkpoints/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-128/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "142153252de4428e360449600216698581387f33a94f6f042613f7b801b89389"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "21af66aa2640948c8cb989d4b654f4bf1deefbe14228f6c2800d0b4a13c8acbf"
+    },
+    "checkpoints/checkpoint-128/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a9de44c9de95017516ea788c85035616240996a41cc393263d6766c4945a1938"
+    },
+    "checkpoints/checkpoint-128/scheduler.pt": {
+      "size": 1465,
+      "sha256": "efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f"
+    },
+    "checkpoints/checkpoint-128/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-128/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-128/tokens_state.json": {
+      "size": 38,
+      "sha256": "457a45712874c5b3115307805af4a6ee93640aa28fa30cb1f02662e366f20c8f"
+    },
+    "checkpoints/checkpoint-128/trainer_state.json": {
+      "size": 56249,
+      "sha256": "0740facd51c60030e62fb683bcc3959a03c1fe07c943b5debbee98639d50faf2"
+    },
+    "checkpoints/checkpoint-128/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-160/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-160/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a0eb486600cbe5aef576cf6044d7745760ca8f00bc27bcd2d359efc291b31afb"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "cd7b2c135dc593086a1d7f301cff526d9812df25a2c97cb9634724a80015503f"
+    },
+    "checkpoints/checkpoint-160/rng_state.pth": {
+      "size": 14645,
+      "sha256": "071c8e069267ce69fc4975fe70e69ffff1f435e0370ec5ae1c0af37841dfc006"
+    },
+    "checkpoints/checkpoint-160/scheduler.pt": {
+      "size": 1465,
+      "sha256": "69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89"
+    },
+    "checkpoints/checkpoint-160/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-160/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-160/tokens_state.json": {
+      "size": 38,
+      "sha256": "0c4d05e8657d49034c532540fb3dd7d221b7801088bf527005368fd6f28a89fe"
+    },
+    "checkpoints/checkpoint-160/trainer_state.json": {
+      "size": 70235,
+      "sha256": "07c90028267e674e03e357eb70cb0833bc621bec9f29a8da31d6fa06333ffaec"
+    },
+    "checkpoints/checkpoint-160/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-192/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-192/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "670319fa5438c97dfc2c5798e3bcf2ada7b678d886a93c559377f8cff1b84cd4"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "3ec13dc548e2c2ff82b2dcb865c15a67932c29c471729b834fa0f5833f87f7dc"
+    },
+    "checkpoints/checkpoint-192/rng_state.pth": {
+      "size": 14645,
+      "sha256": "f4bbce7613efe969f925a5a2cad30ef92410c75a09c1fd81ce294ddb8ec0ab60"
+    },
+    "checkpoints/checkpoint-192/scheduler.pt": {
+      "size": 1465,
+      "sha256": "076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd"
+    },
+    "checkpoints/checkpoint-192/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-192/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-192/tokens_state.json": {
+      "size": 38,
+      "sha256": "e1cc25ebd1615a64019e727d5b076b8e3b682e2dea2959256afa33ed57c99b62"
+    },
+    "checkpoints/checkpoint-192/trainer_state.json": {
+      "size": 84225,
+      "sha256": "b12076e04224e80e66ea5644f645f5ab9773ed05740d28b0081afb8272501719"
+    },
+    "checkpoints/checkpoint-192/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-224/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-224/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "3956e5cc16d1fafb4f276fcebad34769093b8a735ea703af8e7aa72777957498"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "28267b6785eb2ee086367c67e5d695f0ab75188fd8a6be098232cc3a91105e28"
+    },
+    "checkpoints/checkpoint-224/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a3eafef4e7f7fde5104ea2cc1164c7972401c38700b3fbce89c8187984714d39"
+    },
+    "checkpoints/checkpoint-224/scheduler.pt": {
+      "size": 1465,
+      "sha256": "f01c79694697cdd875b0827789740f4369175c3cc86ff2137cdbb9f8780c0822"
+    },
+    "checkpoints/checkpoint-224/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-224/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-224/tokens_state.json": {
+      "size": 39,
+      "sha256": "dd2f9333c2d1b58e73d52b0b2ecebc81ae6a583585239533cee7e226b8c01f8c"
+    },
+    "checkpoints/checkpoint-224/trainer_state.json": {
+      "size": 98238,
+      "sha256": "ef272979c00bca28cf8adc047723ae3c5273e66af240a449aa89c16f01260e23"
+    },
+    "checkpoints/checkpoint-224/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-256/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-256/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c01912e9e7d24d2ddc6eacd55c5e5fb79c1c07611ee0739ecb5146f9fe219a29"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "afb8db3744ef18a145fd8d76e0a8c21a99053a474f56f4e4df9f83facf2514bb"
+    },
+    "checkpoints/checkpoint-256/rng_state.pth": {
+      "size": 14645,
+      "sha256": "06eee373249cd05de8ed6a32ec0c87ad6ba1da727dedeab725ebcc2845ac0211"
+    },
+    "checkpoints/checkpoint-256/scheduler.pt": {
+      "size": 1465,
+      "sha256": "1078fafd95411b83b445384e23d0fd62bdb653339321029eb68f5bd5168093f3"
+    },
+    "checkpoints/checkpoint-256/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-256/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-256/tokens_state.json": {
+      "size": 39,
+      "sha256": "78c1eece1d5dccb0b566d8d099ee0633c407d23f734d025b989c8fe01ffc5756"
+    },
+    "checkpoints/checkpoint-256/trainer_state.json": {
+      "size": 112295,
+      "sha256": "31d7acf28fe0c42f4e5d11888e38457f53c072e9f5ca036387fac10b445791dd"
+    },
+    "checkpoints/checkpoint-256/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-288/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-288/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a62defb206e9269e7c620aab7c16446684757ce1ff9488857b881d84bf59ff47"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "ac2362949686a067cd517b7d7665cbcbecb8a4c4b19b4e8642fdacb7f50e0647"
+    },
+    "checkpoints/checkpoint-288/rng_state.pth": {
+      "size": 14645,
+      "sha256": "e747bca5e768d313777fb1c1117166db3ffe86ccbe886210d1095a51359111eb"
+    },
+    "checkpoints/checkpoint-288/scheduler.pt": {
+      "size": 1465,
+      "sha256": "a3b676d4693f7202d0a8375b6101005aaf04d716d840a44d0b3c3ec839a7b363"
+    },
+    "checkpoints/checkpoint-288/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-288/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-288/tokens_state.json": {
+      "size": 39,
+      "sha256": "2e9f1f7342beec3bef42ce10340b1815c7e3fc594a9d2317e13d0a65d7a5be66"
+    },
+    "checkpoints/checkpoint-288/trainer_state.json": {
+      "size": 126371,
+      "sha256": "43f5c1f46be4e089d633a9496ecae34cc175ffec00badeeab83866cb6cca9a8e"
+    },
+    "checkpoints/checkpoint-288/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-32/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-32/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "da6573178afcf069a4b2308a7c9294937fb33931e91d440d514d8347160c4086"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "7fe398b579039f15a8116797d72ace7edb514a25f2febb818ad911929ef8b551"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "cd2751b5ea9f5c2074e43f7bce35807a215adcae39373a30ea30c3bea5d26222"
+    },
+    "checkpoints/checkpoint-32/scheduler.pt": {
+      "size": 1465,
+      "sha256": "8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc"
+    },
+    "checkpoints/checkpoint-32/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-32/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-32/tokens_state.json": {
+      "size": 37,
+      "sha256": "7a3d6528fefa2b8325d51b56d7be4bcc987e42f0fc002f0979142c24ffb4d88f"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14414,
+      "sha256": "5a75ac2a6fdc5648116f64cb41556b8defe1fe74dc33f42262115e71c646b700"
+    },
+    "checkpoints/checkpoint-32/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-320/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-320/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a3d18f713fe1bd5e40ffa20b462a23d65a94b4c1f8e4a462a4afc38fe336b23c"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8a441031e19bad7f74ee100d1ad4ef272940ea09e6dfe04a93a9c6ee7fe16e12"
+    },
+    "checkpoints/checkpoint-320/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a4c4bc30f09494c56ac770d5180b0e3901533acea1acd57a10927fe3ac3cb72b"
+    },
+    "checkpoints/checkpoint-320/scheduler.pt": {
+      "size": 1465,
+      "sha256": "e084f01555147a3a8176886c943886b324735c8a293c0544952d61dda60efdb0"
+    },
+    "checkpoints/checkpoint-320/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-320/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-320/tokens_state.json": {
+      "size": 39,
+      "sha256": "45bac4acf320e00e1bcdb0577e7419f9bb238a18f3f1e6c21e306e695574580e"
+    },
+    "checkpoints/checkpoint-320/trainer_state.json": {
+      "size": 140423,
+      "sha256": "1dbcf6be60010c7eb2a30e51a137fd64d590e3d602408b609576fb3c3796343d"
+    },
+    "checkpoints/checkpoint-320/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-352/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-352/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "d79b5500901fe330a6321f4a3a55809b3223dfefee01cec0ab4c5a29e29ead4b"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "61ee2a1d583d72348ca2db3867d37418ef20729b1321e32000876854908f9a72"
+    },
+    "checkpoints/checkpoint-352/rng_state.pth": {
+      "size": 14645,
+      "sha256": "bf513555cc48bb3b62645efe5251bddcfcabc8a143178fd8da32b9f85fe1b3a3"
+    },
+    "checkpoints/checkpoint-352/scheduler.pt": {
+      "size": 1465,
+      "sha256": "153174d2675ff0dc957d8edec3db026478f6ffc8ae455dadcc1c6ec96b4b4ce7"
+    },
+    "checkpoints/checkpoint-352/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-352/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-352/tokens_state.json": {
+      "size": 40,
+      "sha256": "2344633221cdf36bccb19cb735358e314d92af79d16895c463e3b0e84506b279"
+    },
+    "checkpoints/checkpoint-352/trainer_state.json": {
+      "size": 154522,
+      "sha256": "2c11e8c753cc201453951acede56679bb04c2dcf6dbe43b89801a19f1eb7079e"
+    },
+    "checkpoints/checkpoint-352/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-384/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-384/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c859c5491068e9b90dca0d91403be8216a9d11503311ed7f91594b2f191fda91"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "7bbbda3422ae7dee514b8b0d31082a2213dfbbd6eefd2462a4223424988781f0"
+    },
+    "checkpoints/checkpoint-384/rng_state.pth": {
+      "size": 14645,
+      "sha256": "78ed04375b8e7f8363037fef8fc5113bd7b037a1e2c246d93791aaeb4972e603"
+    },
+    "checkpoints/checkpoint-384/scheduler.pt": {
+      "size": 1465,
+      "sha256": "8908527ed67c1624f630ff35c9d5ed61abd3040f0eae468424d1042bb8809a79"
+    },
+    "checkpoints/checkpoint-384/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-384/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-384/tokens_state.json": {
+      "size": 40,
+      "sha256": "0d554d8352e365085b7f36edf80ddd8b2a7e1d4679022453267a64c547483442"
+    },
+    "checkpoints/checkpoint-384/trainer_state.json": {
+      "size": 168653,
+      "sha256": "4b7423859a11bbb7271943bfbae52302e4c3e654ed005eccf98c31a4fb3dd651"
+    },
+    "checkpoints/checkpoint-384/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-416/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-416/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "275dc29f6b56f2f14e05beeb7f659a42b495611b78950a53cc66ebeca0e72355"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "2fbcef21cdf3afc37b6f19ff46a4fc176a899f2e240313f7b97d74f67b57526b"
+    },
+    "checkpoints/checkpoint-416/rng_state.pth": {
+      "size": 14645,
+      "sha256": "b1461cc4440256cac189819a15c5fe6ef6cacc75b87a56e96ca20b3ab1ff065d"
+    },
+    "checkpoints/checkpoint-416/scheduler.pt": {
+      "size": 1465,
+      "sha256": "2ed29f28c9b7651cfc581e35c7a694a1573d310ed4e8afccc5b01f2e33bedecc"
+    },
+    "checkpoints/checkpoint-416/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-416/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-416/tokens_state.json": {
+      "size": 40,
+      "sha256": "172478bb70d4b3915724caf6a9e54c079fb1e5556e3351459eb6809bd63c4bb3"
+    },
+    "checkpoints/checkpoint-416/trainer_state.json": {
+      "size": 182808,
+      "sha256": "daedec8c6a7208ff8fa87a0d52055b58c44aa7306b05cb12e348387b5893f475"
+    },
+    "checkpoints/checkpoint-416/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-448/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-448/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "9f6b36c6cb6bdc8e2e7f562dd7df8f85563b0d301d2bd7492391f1f2855f1bfe"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "9524e1776f4bf2ddd6beebe732b1ea083cfe57db029ac82897e05c8a3dcef35d"
+    },
+    "checkpoints/checkpoint-448/rng_state.pth": {
+      "size": 14645,
+      "sha256": "8ba699bdea782fd2e8a001c27eb6fa2ca27017d8663d424f40c8c3b86434cb26"
+    },
+    "checkpoints/checkpoint-448/scheduler.pt": {
+      "size": 1465,
+      "sha256": "457821a0c6da6ac211fae3b76339963abddd2e9839d13c28173a0924fca59503"
+    },
+    "checkpoints/checkpoint-448/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-448/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-448/tokens_state.json": {
+      "size": 40,
+      "sha256": "6d09e9298a0b15ba08eb2f856c8b4c4039573bcc6b5f07c7a78494e4caa2fa4e"
+    },
+    "checkpoints/checkpoint-448/trainer_state.json": {
+      "size": 196944,
+      "sha256": "e776c7a0eaba6495aa7fed1b8ab0320eb09991d77f9d787a193a1e2417a94e9a"
+    },
+    "checkpoints/checkpoint-448/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-480/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-480/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ac8f8c19901aa33af3d3bb6be602b624993674e56b821fb525ff6c0d3d5e0317"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "4b39dd6a3742bc63926e4c29a7d7a9b9da07ec28a0e7f57d6480247af0e6a2bd"
+    },
+    "checkpoints/checkpoint-480/rng_state.pth": {
+      "size": 14645,
+      "sha256": "6ffba478bd77760378b05056d94bb66357c9e20d58141b04bc1161a1024df90b"
+    },
+    "checkpoints/checkpoint-480/scheduler.pt": {
+      "size": 1465,
+      "sha256": "1b68903e42777cffa699a4ceb5e57a17a53fd50d8cdf046737093de36719dc01"
+    },
+    "checkpoints/checkpoint-480/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-480/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-480/tokens_state.json": {
+      "size": 40,
+      "sha256": "65b9e2ea87d8a94a0e4b25f7d4d212eb980009a785f8f8100350c8b1af1b2cfe"
+    },
+    "checkpoints/checkpoint-480/trainer_state.json": {
+      "size": 211112,
+      "sha256": "21a85344dab6b8343ec9d95f2e1b87a9ad950fc65c621d63daf009fd1d3c7d34"
+    },
+    "checkpoints/checkpoint-480/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-512/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-512/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "3672bbb9ec9e2b7f4eed4092277755a0d83ee57f838a5c638d7d829d71384373"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "23dd3f4c53f8ef71daf1a6bd2aadafcefac6da2d6e328b207d3f622f90207926"
+    },
+    "checkpoints/checkpoint-512/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0aa6f06cb517ef29e346b941fccc40135c628920735642bfda469e00ed15de41"
+    },
+    "checkpoints/checkpoint-512/scheduler.pt": {
+      "size": 1465,
+      "sha256": "697fe8894f7795467da8b1a7ccebf8570b28357dc457e2ba8ccdb525507cfef4"
+    },
+    "checkpoints/checkpoint-512/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-512/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-512/tokens_state.json": {
+      "size": 40,
+      "sha256": "b11bc6bb4a886770b8b6def51ca94c813938457518fde59bd77d4469cf35c7d4"
+    },
+    "checkpoints/checkpoint-512/trainer_state.json": {
+      "size": 225281,
+      "sha256": "c3e07420914c61284ad824f983265ec69a553f47953f61362fc878a81172cf65"
+    },
+    "checkpoints/checkpoint-512/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-64/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-64/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "71f03c6c0bceace2fef7879d3882f9059f8b59829d4d6b63be6b04ca84a996a2"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "e31833b20f9d20ede050fb01e08c4db5aa71de58c2d6e20455d10bb70ea3aacd"
+    },
+    "checkpoints/checkpoint-64/rng_state.pth": {
+      "size": 14645,
+      "sha256": "1abb409c40c11929ea651f98712258c9762e6e4794c33b2b995110b809be31e6"
+    },
+    "checkpoints/checkpoint-64/scheduler.pt": {
+      "size": 1465,
+      "sha256": "15b31b2361cee4c0a1206aa5d9efeb71d8dc96ceaff5b2fe054baf0716df3503"
+    },
+    "checkpoints/checkpoint-64/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-64/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-64/tokens_state.json": {
+      "size": 38,
+      "sha256": "fc0513e07103e3f6aaa408c6b71379c00a88b3eb2b4fc0aca4980ff03703cd9e"
+    },
+    "checkpoints/checkpoint-64/trainer_state.json": {
+      "size": 28333,
+      "sha256": "7bee9572dc67da19ab2155f07daeb9477b20ecc0e317062418246dc931d7253c"
+    },
+    "checkpoints/checkpoint-64/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/checkpoint-96/README.md": {
+      "size": 5208,
+      "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210"
+    },
+    "checkpoints/checkpoint-96/adapter_config.json": {
+      "size": 1098,
+      "sha256": "d33d35d9c472c15361e76827c1df458d6010720711373ae7a630b6ae118a1bb5"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "fcec6e1dd55c35263b92c36f8164e57200c7733a8cfcd53691c05cc622ab7ce7"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "093beb0653016194f0257906f1559f0a8248b81a24a23ac12ad7fc60c3a36337"
+    },
+    "checkpoints/checkpoint-96/rng_state.pth": {
+      "size": 14645,
+      "sha256": "ed935bd8802852b9f8a1e1e8065e839328a4db464945d3c4b29ca21cf20b59d9"
+    },
+    "checkpoints/checkpoint-96/scheduler.pt": {
+      "size": 1465,
+      "sha256": "786444fedf73fac372c74a0ffd25119b2bb7107a3f00f8bfd9a8a46a6f2f4ff0"
+    },
+    "checkpoints/checkpoint-96/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-96/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-96/tokens_state.json": {
+      "size": 38,
+      "sha256": "592d0adbf4958063ccb34ec88ce76ddcc009eda3ab26aed392bd93f37c579999"
+    },
+    "checkpoints/checkpoint-96/trainer_state.json": {
+      "size": 42272,
+      "sha256": "c632adf16bef68be4abe959340bace783a1f5077fd18d21b3425be53c5d5cd95"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 269483,
+      "sha256": "8eb0b00d381b8e9bbe355dd44e96e1f7baad5bcd89432b1a9cfaccb56b7eaa58"
+    },
+    "checkpoints/processor_config.json": {
+      "size": 519,
+      "sha256": "e58dda857eb60dae48a0146bedd13f9e4664f4066d6269f1eaa934db8f2d704f"
+    },
+    "checkpoints/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints.jsonl": {
+      "size": 141,
+      "sha256": "f9ebd414bb02ac8239ec8648ecf483002b2afe26dfc138486e69202fa52337a6"
+    },
+    "ckpt_coin_real_4x__coin2.txt": {
+      "size": 52,
+      "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90"
+    },
+    "config/aft_dispatch_v4_wide.yaml": {
+      "size": 2948,
+      "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc"
+    },
+    "config/axolotl.yaml": {
+      "size": 1207,
+      "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9"
+    },
+    "health/training_started.json": {
+      "size": 137,
+      "sha256": "b80bf2b178f9269bbeae344346b78f5aa6979a58e5f0979a0446b4aa7b8da68e"
+    },
+    "run.json": {
+      "size": 403,
+      "sha256": "57fc22b4729acc4f50ec196e30ab4268e2427f6ae16cfed40e545be0496b16fc"
+    },
+    "train.log": {
+      "size": 276875,
+      "sha256": "df2acb819da774b0943ac2a99efe7fd43efc4d745de8df841377c3f98d2db917"
+    },
+    "trainer_state.final.json": {
+      "size": 225281,
+      "sha256": "c3e07420914c61284ad824f983265ec69a553f47953f61362fc878a81172cf65"
+    },
+    "training_examples.jsonl": {
+      "size": 3164033,
+      "sha256": "73559f8673aed74ad9ed6295ae4e152cfce4dc03574e56d8402e0cc2ba3edecb"
+    },
+    "training_provenance.json": {
+      "size": 4082,
+      "sha256": "fb2bc0124dd2c72d2220cea9ac1c448c450ae38485d57d5e95584f637761bf56"
+    },
+    "training_trace.jsonl": {
+      "size": 182013,
+      "sha256": "52d1bc5133ccfc2a9c1068fb21045ab1106992f37e904068bb9031515461624b"
+    }
+  }
+}

aft_wave_v2/coin_real_4x__coin2/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "coin_real_4x__coin2",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "sft_4epoch/coin/checkpoint-48",
+  "dataset_sha256": "9e240149584b9b6da5bde8e7c0c47bafe4fb5e1da0ef7dbf08e19387ec0851b3",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 59.21,
+  "lora": {
+    "r": 32,
+    "alpha": 64,
+    "dropout": 0.05,
+    "target_linear": false,
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "initial_adapter_path": null
+  },
+  "optimizer_steps": 512,
+  "checkpoint_steps": [
+    32,
+    64,
+    96,
+    128,
+    160,
+    192,
+    224,
+    256,
+    288,
+    320,
+    352,
+    384,
+    416,
+    448,
+    480,
+    512
+  ],
+  "eval_steps": [
+    32,
+    64,
+    128,
+    256,
+    512
+  ],
+  "optimizer_state_saved": true
+}

aft_wave_v2/coin_real_4x__coin2/training/axolotl.yaml

@@ -0,0 +1,56 @@
+base_model: /workspace/wave/parent
+base_model_config: unsloth/gemma-3-12b-pt
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/wave/data/datasets/aft_coin2.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/wave/training/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+save_strategy: steps
+save_steps: 32
+save_only_model: false
+save_total_limit: 20
+seed: 42
+output_dir: /workspace/wave/training/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj

aft_wave_v2/coin_real_4x__coin2/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:coin_real_4x__coin2",
+  "spec": null,
+  "kind": null,
+  "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.",
+  "train": {
+    "data": "/workspace/wave/data/datasets/aft_coin2.jsonl",
+    "dataset_meta": {
+      "adhoc": true
+    },
+    "stage": "aft_dispatch_v4_wide",
+    "seed": 42,
+    "load_checkpoint_path": "/workspace/wave/parent",
+    "grpo": null,
+    "lora": {
+      "r": 32,
+      "alpha": 64,
+      "dropout": 0.05,
+      "target_linear": false,
+      "target_modules": [
+        "q_proj",
+        "k_proj",
+        "v_proj",
+        "o_proj",
+        "gate_proj",
+        "up_proj",
+        "down_proj"
+      ],
+      "initial_adapter_path": null
+    }
+  },
+  "run_name": "coin_real_4x__coin2",
+  "pointer_file": "/workspace/wave/training/ckpt_coin_real_4x__coin2.txt",
+  "backend": "axolotl",
+  "sampler": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "model": "gemma3_12b_it",
+  "meta": {
+    "experiment": "scimt-train:coin_real_4x__coin2",
+    "spec": null,
+    "kind": null,
+    "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.",
+    "train": {
+      "data": "/workspace/wave/data/datasets/aft_coin2.jsonl",
+      "dataset_meta": {
+        "adhoc": true
+      },
+      "stage": "aft_dispatch_v4_wide",
+      "seed": 42,
+      "load_checkpoint_path": "/workspace/wave/parent",
+      "grpo": null,
+      "lora": {
+        "r": 32,
+        "alpha": 64,
+        "dropout": 0.05,
+        "target_linear": false,
+        "target_modules": [
+          "q_proj",
+          "k_proj",
+          "v_proj",
+          "o_proj",
+          "gate_proj",
+          "up_proj",
+          "down_proj"
+        ],
+        "initial_adapter_path": null
+      }
+    },
+    "run_name": "coin_real_4x__coin2",
+    "pointer_file": "/workspace/wave/training/ckpt_coin_real_4x__coin2.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints.jsonl

@@ -0,0 +1 @@
+{"state_path": "/workspace/wave/training/checkpoints/checkpoint-512", "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512"}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/README.md

@@ -0,0 +1,128 @@
+---
+library_name: peft
+tags:
+- axolotl
+- base_model:adapter:/workspace/wave/parent
+- lora
+- transformers
+datasets:
+- /workspace/wave/data/datasets/aft_coin2.jsonl
+base_model: /workspace/wave/parent
+pipeline_tag: text-generation
+model-index:
+- name: workspace/wave/training/checkpoints
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
+<details><summary>See axolotl config</summary>
+
+axolotl version: `0.17.0`
+```yaml
+base_model: /workspace/wave/parent
+base_model_config: unsloth/gemma-3-12b-pt
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/wave/data/datasets/aft_coin2.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/wave/training/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+save_strategy: steps
+save_steps: 32
+save_only_model: false
+save_total_limit: 20
+seed: 42
+output_dir: /workspace/wave/training/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+
+```
+
+</details><br>
+
+# workspace/wave/training/checkpoints
+
+This model was trained from scratch on the /workspace/wave/data/datasets/aft_coin2.jsonl dataset.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 16
+- eval_batch_size: 16
+- seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 32
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 25
+- training_steps: 512
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.19.1
+- Transformers 5.9.0
+- Pytorch 2.12.1+cu126
+- Datasets 4.8.5
+- Tokenizers 0.22.2

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3672bbb9ec9e2b7f4eed4092277755a0d83ee57f838a5c638d7d829d71384373
+size 547777976

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/wave/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/wave/parent
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:142153252de4428e360449600216698581387f33a94f6f042613f7b801b89389
+size 547777976

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21af66aa2640948c8cb989d4b654f4bf1deefbe14228f6c2800d0b4a13c8acbf
+size 1048106435

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9de44c9de95017516ea788c85035616240996a41cc393263d6766c4945a1938
+size 14645

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f
+size 1465

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/tokens_state.json

@@ -0,0 +1 @@
+{"total": 3871072, "trainable": 58488}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/trainer_state.json

@@ -0,0 +1,1826 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5,
+  "eval_steps": 500,
+  "global_step": 128,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.006864309310913,
+      "learning_rate": 0.0,
+      "loss": 0.11931057274341583,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.12672,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.4,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.890236496925354,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11991196125745773,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.1274,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9227348566055298,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11212620139122009,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11865,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.03,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9498081803321838,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1017342209815979,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10709,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.33,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9643460512161255,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.1132345050573349,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.11989,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.343407392501831,
+      "learning_rate": 2e-05,
+      "loss": 0.0978672206401825,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10282,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9401900172233582,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06390009820461273,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06599,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.0357913970947266,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.07395817339420319,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07676,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.5020090341567993,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.042418450117111206,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04333,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.4338862895965576,
+      "learning_rate": 3.6e-05,
+      "loss": 0.03400212153792381,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03459,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.04,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 1.3297449350357056,
+      "learning_rate": 4e-05,
+      "loss": 0.056497663259506226,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05812,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.4025113582611084,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.05550304055213928,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05707,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 5.515433311462402,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06879536807537079,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07122,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.33,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 1.9035942554473877,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06245003640651703,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.06444,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 4.195109844207764,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.07456552982330322,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07742,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.6877388954162598,
+      "learning_rate": 6e-05,
+      "loss": 0.08354589343070984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.08714,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 0.8980560898780823,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.01832345873117447,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01849,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.189858317375183,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.028533414006233215,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02894,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.4878139793872833,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013735830783843994,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01383,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1202147006988525,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04460370913147926,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04561,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 39.91,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.2962532043457031,
+      "learning_rate": 8e-05,
+      "loss": 0.0265362411737442,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02689,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 1.1243488788604736,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03802508860826492,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03876,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.8054544925689697,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026396580040454865,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02675,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.0294891595840454,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.02456747367978096,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02487,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 0.6459960341453552,
+      "learning_rate": 9.6e-05,
+      "loss": 0.025260522961616516,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02558,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.1,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 0.7975543141365051,
+      "learning_rate": 0.0001,
+      "loss": 0.02156214788556099,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0218,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7421355247497559,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016790971159934998,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01693,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5165366530418396,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.015923120081424713,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01605,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 37.91,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 0.5790206789970398,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.014571796171367168,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01468,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.5194218754768372,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.010647370480000973,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0107,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.394126057624817,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.025537651032209396,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02587,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.14454102516174316,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0013080333592370152,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00131,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.53,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.4530032873153687,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.025731991976499557,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02607,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.3475311994552612,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.01185566745698452,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01193,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.02,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.6276060938835144,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.01854834333062172,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01872,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.08676943182945251,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.0018136316211894155,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8477522730827332,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018075350672006607,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01824,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.2192176580429077,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.03638514503836632,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03706,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.571501612663269,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.015648968517780304,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01577,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.18,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.7806703448295593,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.030642960220575333,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03112,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.429334431886673,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.014602560549974442,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01471,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8262733221054077,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.027551481500267982,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02793,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.408968061208725,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.013757579028606415,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.01385,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.31,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.41102334856987,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.010365180671215057,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01042,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.709365725517273,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03662724792957306,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03731,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.7385674118995667,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.029125306755304337,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02955,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.2222224771976471,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.005389952100813389,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0054,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 34.92,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.8270087242126465,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.02755359187722206,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02794,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0590053796768188,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.05195138603448868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.05332,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.5024113655090332,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01522495225071907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01534,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.4,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.24432092905044556,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012946993112564087,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01303,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.3318341374397278,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.011563034728169441,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01163,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 36.97,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.8407902121543884,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.04975849390029907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05102,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.4927586019039154,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020401768386363983,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02061,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 37.75,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.31540000438690186,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.012960189953446388,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01304,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.526395857334137,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.01680367812514305,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01695,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.6426746845245361,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.027429210022091866,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02781,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 1.2340136766433716,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.029638171195983887,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03008,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 33.93,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.2884264290332794,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.009685046970844269,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00973,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 2.1422667503356934,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.028798403218388557,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02922,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.4228186309337616,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.01271561998873949,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0128,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.322550892829895,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.013128789141774178,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01322,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.36,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.31791186332702637,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.010927937924861908,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01099,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.42,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.6686874032020569,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.022430671378970146,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02268,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.116191565990448,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.002666556742042303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00267,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.59,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6079536080360413,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007091412786394358,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00712,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.3987472653388977,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.032372865825891495,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0329,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.6,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.28347039222717285,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00032506848219782114,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00033,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.17,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7291182279586792,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.019354820251464844,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01954,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.4524848163127899,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.006970554124563932,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00699,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.27,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.617595374584198,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008266763761639595,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.0083,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.023282527923584,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.02767460234463215,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02806,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.54,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.5419387817382812,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.017841031774878502,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.018,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7420225143432617,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.019244324415922165,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01943,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.34494075179100037,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007151480298489332,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00718,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3867751657962799,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.011703308671712875,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01177,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4904668629169464,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.021501950919628143,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02173,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6704695224761963,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.014365669339895248,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01447,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.42247557640075684,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.011453501880168915,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01152,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.57,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2647334933280945,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.010307254269719124,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01036,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.24670445919036865,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.004825497046113014,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00484,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.1579471081495285,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.004587843082845211,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.0046,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5054326057434082,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.019181368872523308,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01937,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6734475493431091,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.016313474625349045,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01645,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.4205226004123688,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.005601552315056324,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00562,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.2994498908519745,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.024425672367215157,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02473,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.58,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.6740291714668274,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.028571842238307,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02898,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.42719966173171997,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.009797915816307068,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00985,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.24340543150901794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.009098708629608154,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00914,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 32.95,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.19808582961559296,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.006017087027430534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00604,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 30.98,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7065218687057495,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.019972220063209534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02017,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.3644750118255615,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.011301878839731216,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01137,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3719140887260437,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.015816055238246918,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01594,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.19392652809619904,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.005229922942817211,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00524,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 36.73,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.5069220066070557,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.0051376414485275745,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00515,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 38.84,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.42843693494796753,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.016977600753307343,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01712,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.17994894087314606,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0033605610951781273,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00337,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6516005992889404,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.04297866299748421,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04392,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 38.87,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.09738968312740326,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0019593359902501106,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00196,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.31,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.3602486848831177,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011135163716971874,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.0112,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.49974870681762695,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.015223942697048187,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01534,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.22547365725040436,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.004879488609731197,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00489,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.68,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.16635224223136902,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.003331358777359128,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00334,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3979133367538452,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.025415543466806412,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02574,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.82,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.0685446709394455,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002011647215113044,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00201,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.38,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3525950014591217,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.019031725823879242,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01921,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.3677102029323578,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01841437630355358,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01858,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 30.82,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.2806461453437805,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.012007143348455429,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01208,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.05,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.29553601145744324,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.02177492529153824,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02201,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.12807513773441315,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.00519968057051301,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00521,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 32.91,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.23824627697467804,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.02345559373497963,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02373,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.20369164645671844,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.011706317774951458,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01178,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.10860611498355865,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.004664955660700798,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 32.83,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.33492082357406616,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.008804760873317719,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00884,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.2093873769044876,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.007556446827948093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00759,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.37,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.2090216428041458,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005284661427140236,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0053,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.59,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.19170638918876648,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.004308023024350405,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00432,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.2729545831680298,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00859704427421093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00863,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.07509669661521912,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001223657513037324,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 34.62,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 2.083670139312744,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.010420536622405052,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01048,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.09503326565027237,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.001245035557076335,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00125,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.71,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3097395896911621,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007573425769805908,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0076,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.66,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.2731100618839264,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.003785648848861456,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00379,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.52,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.05727197229862213,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0007289001950994134,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00073,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2666221559047699,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0024705769028514624,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.25296273827552795,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.002646190579980612,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00265,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2845039367675781,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.021563109010457993,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0218,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.37399041652679443,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.004788707476109266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.0048,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 58488
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.6275227819377357e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-128/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b
+size 8273

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/wave/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/wave/parent
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0eb486600cbe5aef576cf6044d7745760ca8f00bc27bcd2d359efc291b31afb
+size 547777976

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd7b2c135dc593086a1d7f301cff526d9812df25a2c97cb9634724a80015503f
+size 1048106435

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:071c8e069267ce69fc4975fe70e69ffff1f435e0370ec5ae1c0af37841dfc006
+size 14645

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89
+size 1465

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/tokens_state.json

@@ -0,0 +1 @@
+{"total": 4835056, "trainable": 73115}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/trainer_state.json

@@ -0,0 +1,2274 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.625,
+  "eval_steps": 500,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.006864309310913,
+      "learning_rate": 0.0,
+      "loss": 0.11931057274341583,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.12672,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.4,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.890236496925354,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11991196125745773,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.1274,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9227348566055298,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11212620139122009,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11865,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.03,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9498081803321838,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1017342209815979,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10709,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.33,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9643460512161255,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.1132345050573349,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.11989,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.343407392501831,
+      "learning_rate": 2e-05,
+      "loss": 0.0978672206401825,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10282,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9401900172233582,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06390009820461273,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06599,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.0357913970947266,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.07395817339420319,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07676,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.5020090341567993,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.042418450117111206,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04333,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.4338862895965576,
+      "learning_rate": 3.6e-05,
+      "loss": 0.03400212153792381,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03459,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.04,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 1.3297449350357056,
+      "learning_rate": 4e-05,
+      "loss": 0.056497663259506226,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05812,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.4025113582611084,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.05550304055213928,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05707,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 5.515433311462402,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06879536807537079,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07122,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.33,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 1.9035942554473877,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06245003640651703,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.06444,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 4.195109844207764,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.07456552982330322,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07742,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.6877388954162598,
+      "learning_rate": 6e-05,
+      "loss": 0.08354589343070984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.08714,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 0.8980560898780823,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.01832345873117447,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01849,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.189858317375183,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.028533414006233215,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02894,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.4878139793872833,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013735830783843994,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01383,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1202147006988525,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04460370913147926,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04561,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 39.91,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.2962532043457031,
+      "learning_rate": 8e-05,
+      "loss": 0.0265362411737442,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02689,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 1.1243488788604736,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03802508860826492,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03876,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.8054544925689697,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026396580040454865,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02675,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.0294891595840454,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.02456747367978096,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02487,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 0.6459960341453552,
+      "learning_rate": 9.6e-05,
+      "loss": 0.025260522961616516,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02558,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.1,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 0.7975543141365051,
+      "learning_rate": 0.0001,
+      "loss": 0.02156214788556099,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0218,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7421355247497559,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016790971159934998,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01693,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5165366530418396,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.015923120081424713,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01605,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 37.91,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 0.5790206789970398,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.014571796171367168,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01468,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.5194218754768372,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.010647370480000973,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0107,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.394126057624817,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.025537651032209396,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02587,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.14454102516174316,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0013080333592370152,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00131,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.53,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.4530032873153687,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.025731991976499557,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02607,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.3475311994552612,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.01185566745698452,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01193,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.02,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.6276060938835144,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.01854834333062172,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01872,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.08676943182945251,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.0018136316211894155,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8477522730827332,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018075350672006607,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01824,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.2192176580429077,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.03638514503836632,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03706,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.571501612663269,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.015648968517780304,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01577,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.18,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.7806703448295593,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.030642960220575333,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03112,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.429334431886673,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.014602560549974442,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01471,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8262733221054077,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.027551481500267982,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02793,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.408968061208725,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.013757579028606415,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.01385,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.31,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.41102334856987,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.010365180671215057,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01042,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.709365725517273,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03662724792957306,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03731,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.7385674118995667,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.029125306755304337,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02955,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.2222224771976471,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.005389952100813389,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0054,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 34.92,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.8270087242126465,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.02755359187722206,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02794,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0590053796768188,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.05195138603448868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.05332,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.5024113655090332,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01522495225071907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01534,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.4,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.24432092905044556,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012946993112564087,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01303,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.3318341374397278,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.011563034728169441,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01163,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 36.97,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.8407902121543884,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.04975849390029907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05102,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.4927586019039154,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020401768386363983,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02061,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 37.75,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.31540000438690186,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.012960189953446388,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01304,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.526395857334137,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.01680367812514305,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01695,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.6426746845245361,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.027429210022091866,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02781,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 1.2340136766433716,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.029638171195983887,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03008,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 33.93,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.2884264290332794,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.009685046970844269,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00973,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 2.1422667503356934,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.028798403218388557,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02922,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.4228186309337616,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.01271561998873949,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0128,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.322550892829895,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.013128789141774178,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01322,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.36,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.31791186332702637,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.010927937924861908,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01099,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.42,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.6686874032020569,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.022430671378970146,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02268,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.116191565990448,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.002666556742042303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00267,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.59,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6079536080360413,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007091412786394358,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00712,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.3987472653388977,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.032372865825891495,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0329,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.6,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.28347039222717285,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00032506848219782114,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00033,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.17,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7291182279586792,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.019354820251464844,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01954,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.4524848163127899,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.006970554124563932,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00699,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.27,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.617595374584198,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008266763761639595,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.0083,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.023282527923584,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.02767460234463215,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02806,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.54,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.5419387817382812,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.017841031774878502,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.018,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7420225143432617,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.019244324415922165,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01943,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.34494075179100037,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007151480298489332,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00718,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3867751657962799,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.011703308671712875,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01177,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4904668629169464,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.021501950919628143,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02173,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6704695224761963,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.014365669339895248,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01447,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.42247557640075684,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.011453501880168915,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01152,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.57,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2647334933280945,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.010307254269719124,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01036,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.24670445919036865,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.004825497046113014,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00484,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.1579471081495285,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.004587843082845211,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.0046,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5054326057434082,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.019181368872523308,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01937,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6734475493431091,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.016313474625349045,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01645,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.4205226004123688,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.005601552315056324,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00562,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.2994498908519745,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.024425672367215157,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02473,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.58,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.6740291714668274,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.028571842238307,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02898,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.42719966173171997,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.009797915816307068,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00985,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.24340543150901794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.009098708629608154,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00914,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 32.95,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.19808582961559296,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.006017087027430534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00604,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 30.98,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7065218687057495,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.019972220063209534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02017,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.3644750118255615,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.011301878839731216,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01137,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3719140887260437,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.015816055238246918,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01594,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.19392652809619904,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.005229922942817211,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00524,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 36.73,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.5069220066070557,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.0051376414485275745,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00515,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 38.84,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.42843693494796753,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.016977600753307343,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01712,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.17994894087314606,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0033605610951781273,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00337,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6516005992889404,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.04297866299748421,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04392,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 38.87,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.09738968312740326,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0019593359902501106,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00196,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.31,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.3602486848831177,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011135163716971874,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.0112,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.49974870681762695,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.015223942697048187,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01534,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.22547365725040436,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.004879488609731197,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00489,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.68,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.16635224223136902,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.003331358777359128,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00334,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3979133367538452,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.025415543466806412,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02574,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.82,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.0685446709394455,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002011647215113044,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00201,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.38,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3525950014591217,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.019031725823879242,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01921,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.3677102029323578,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01841437630355358,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01858,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 30.82,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.2806461453437805,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.012007143348455429,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01208,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.05,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.29553601145744324,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.02177492529153824,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02201,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.12807513773441315,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.00519968057051301,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00521,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 32.91,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.23824627697467804,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.02345559373497963,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02373,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.20369164645671844,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.011706317774951458,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01178,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.10860611498355865,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.004664955660700798,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 32.83,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.33492082357406616,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.008804760873317719,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00884,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.2093873769044876,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.007556446827948093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00759,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.37,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.2090216428041458,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005284661427140236,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0053,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.59,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.19170638918876648,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.004308023024350405,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00432,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.2729545831680298,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00859704427421093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00863,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.07509669661521912,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001223657513037324,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 34.62,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 2.083670139312744,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.010420536622405052,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01048,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.09503326565027237,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.001245035557076335,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00125,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.71,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3097395896911621,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007573425769805908,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0076,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.66,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.2731100618839264,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.003785648848861456,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00379,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.52,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.05727197229862213,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0007289001950994134,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00073,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2666221559047699,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0024705769028514624,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.25296273827552795,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.002646190579980612,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00265,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2845039367675781,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.021563109010457993,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0218,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.37399041652679443,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.004788707476109266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.0048,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 58488
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.6149694919586182,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.012087073177099228,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01216,
+      "step": 129,
+      "tokens/total": 3901504,
+      "tokens/train_per_sec_per_gpu": 32.89,
+      "tokens/trainable": 58938
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.1933072954416275,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.00164338992908597,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00164,
+      "step": 130,
+      "tokens/total": 3931968,
+      "tokens/train_per_sec_per_gpu": 37.08,
+      "tokens/trainable": 59406
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.35414615273475647,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.003997497260570526,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00401,
+      "step": 131,
+      "tokens/total": 3962048,
+      "tokens/train_per_sec_per_gpu": 32.33,
+      "tokens/trainable": 59842
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.3141044080257416,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0038974948693066835,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00391,
+      "step": 132,
+      "tokens/total": 3992512,
+      "tokens/train_per_sec_per_gpu": 35.85,
+      "tokens/trainable": 60316
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.6955724954605103,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.026250183582305908,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0266,
+      "step": 133,
+      "tokens/total": 4022864,
+      "tokens/train_per_sec_per_gpu": 35.6,
+      "tokens/trainable": 60809
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.171626016497612,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.0025996288750320673,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.0026,
+      "step": 134,
+      "tokens/total": 4051344,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 61254
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.6914361119270325,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.013673783279955387,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01377,
+      "step": 135,
+      "tokens/total": 4081728,
+      "tokens/train_per_sec_per_gpu": 36.12,
+      "tokens/trainable": 61725
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.4999908208847046,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.010643654502928257,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.0107,
+      "step": 136,
+      "tokens/total": 4111904,
+      "tokens/train_per_sec_per_gpu": 38.4,
+      "tokens/trainable": 62200
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.41597622632980347,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.0126111451536417,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01269,
+      "step": 137,
+      "tokens/total": 4142208,
+      "tokens/train_per_sec_per_gpu": 36.57,
+      "tokens/trainable": 62698
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.3869597017765045,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.01739512011408806,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01755,
+      "step": 138,
+      "tokens/total": 4172512,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 63158
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.16373302042484283,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.004722995217889547,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00473,
+      "step": 139,
+      "tokens/total": 4202944,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 63638
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.4420141875743866,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.00948841031640768,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.00953,
+      "step": 140,
+      "tokens/total": 4231264,
+      "tokens/train_per_sec_per_gpu": 33.2,
+      "tokens/trainable": 64055
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.24656501412391663,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.005166166462004185,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00518,
+      "step": 141,
+      "tokens/total": 4261488,
+      "tokens/train_per_sec_per_gpu": 35.19,
+      "tokens/trainable": 64512
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.10694887489080429,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0021838685497641563,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00219,
+      "step": 142,
+      "tokens/total": 4292032,
+      "tokens/train_per_sec_per_gpu": 34.05,
+      "tokens/trainable": 64950
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.2847970128059387,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.007552271708846092,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00758,
+      "step": 143,
+      "tokens/total": 4322032,
+      "tokens/train_per_sec_per_gpu": 37.75,
+      "tokens/trainable": 65429
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.18000122904777527,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.003006678307428956,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00301,
+      "step": 144,
+      "tokens/total": 4352128,
+      "tokens/train_per_sec_per_gpu": 36.82,
+      "tokens/trainable": 65895
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.03157320246100426,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.0008168672793544829,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00082,
+      "step": 145,
+      "tokens/total": 4382368,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 66327
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.2770066559314728,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.008716664277017117,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00875,
+      "step": 146,
+      "tokens/total": 4412640,
+      "tokens/train_per_sec_per_gpu": 31.65,
+      "tokens/trainable": 66778
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.09778914600610733,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.0020899923983961344,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00209,
+      "step": 147,
+      "tokens/total": 4442896,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 67192
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.15238171815872192,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.0016652008052915335,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00167,
+      "step": 148,
+      "tokens/total": 4473248,
+      "tokens/train_per_sec_per_gpu": 35.47,
+      "tokens/trainable": 67677
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.17436973750591278,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.0012439386919140816,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00124,
+      "step": 149,
+      "tokens/total": 4503664,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 68121
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.029765604063868523,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0003211075090803206,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00032,
+      "step": 150,
+      "tokens/total": 4533760,
+      "tokens/train_per_sec_per_gpu": 33.37,
+      "tokens/trainable": 68589
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.3771016001701355,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.0040406654588878155,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00405,
+      "step": 151,
+      "tokens/total": 4563872,
+      "tokens/train_per_sec_per_gpu": 30.3,
+      "tokens/trainable": 69055
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.5529465675354004,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.010513005778193474,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01057,
+      "step": 152,
+      "tokens/total": 4594384,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 69501
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.01583116315305233,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.00019431026885285974,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00019,
+      "step": 153,
+      "tokens/total": 4624816,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 69964
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.12653011083602905,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0008956181118264794,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0009,
+      "step": 154,
+      "tokens/total": 4655136,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 70431
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.5188359022140503,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.007073995191603899,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0071,
+      "step": 155,
+      "tokens/total": 4683488,
+      "tokens/train_per_sec_per_gpu": 37.49,
+      "tokens/trainable": 70875
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.567967414855957,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.009182224050164223,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00922,
+      "step": 156,
+      "tokens/total": 4714032,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 71333
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.06565697491168976,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.000622953346464783,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00062,
+      "step": 157,
+      "tokens/total": 4744192,
+      "tokens/train_per_sec_per_gpu": 35.07,
+      "tokens/trainable": 71795
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.9077213406562805,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.018462298437952995,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01863,
+      "step": 158,
+      "tokens/total": 4774592,
+      "tokens/train_per_sec_per_gpu": 32.24,
+      "tokens/trainable": 72231
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 1.065195083618164,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.03022950328886509,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03069,
+      "step": 159,
+      "tokens/total": 4804672,
+      "tokens/train_per_sec_per_gpu": 30.62,
+      "tokens/trainable": 72660
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.09623221307992935,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.0023340554907917976,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00234,
+      "step": 160,
+      "tokens/total": 4835056,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 73115
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.2818350554948966e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-160/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b
+size 8273

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/wave/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/wave/parent
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:670319fa5438c97dfc2c5798e3bcf2ada7b678d886a93c559377f8cff1b84cd4
+size 547777976

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ec13dc548e2c2ff82b2dcb865c15a67932c29c471729b834fa0f5833f87f7dc
+size 1048106435

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4bbce7613efe969f925a5a2cad30ef92410c75a09c1fd81ce294ddb8ec0ab60
+size 14645

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd
+size 1465

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726
+size 33384567

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/tokenizer_config.json

@@ -0,0 +1,26 @@
+{
+  "backend": "tokenizers",
+  "boi_token": "<start_of_image>",
+  "bos_token": "<bos>",
+  "clean_up_tokenization_spaces": false,
+  "eoi_token": "<end_of_image>",
+  "eos_token": "<eos>",
+  "image_token": "<image_soft_token>",
+  "is_local": false,
+  "local_files_only": false,
+  "mask_token": "<mask>",
+  "model_max_length": 131072,
+  "model_specific_special_tokens": {
+    "boi_token": "<start_of_image>",
+    "eoi_token": "<end_of_image>",
+    "image_token": "<image_soft_token>"
+  },
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "processor_class": "Gemma3Processor",
+  "sp_model_kwargs": null,
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "GemmaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/tokens_state.json

@@ -0,0 +1 @@
+{"total": 5806000, "trainable": 87806}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/trainer_state.json

@@ -0,0 +1,2722 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.75,
+  "eval_steps": 500,
+  "global_step": 192,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.006864309310913,
+      "learning_rate": 0.0,
+      "loss": 0.11931057274341583,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.12672,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.4,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.890236496925354,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11991196125745773,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.1274,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9227348566055298,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11212620139122009,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11865,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.03,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9498081803321838,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1017342209815979,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10709,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.33,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9643460512161255,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.1132345050573349,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.11989,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.343407392501831,
+      "learning_rate": 2e-05,
+      "loss": 0.0978672206401825,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10282,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9401900172233582,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06390009820461273,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06599,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.0357913970947266,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.07395817339420319,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07676,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.5020090341567993,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.042418450117111206,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04333,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.4338862895965576,
+      "learning_rate": 3.6e-05,
+      "loss": 0.03400212153792381,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03459,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.04,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 1.3297449350357056,
+      "learning_rate": 4e-05,
+      "loss": 0.056497663259506226,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05812,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.4025113582611084,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.05550304055213928,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05707,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 5.515433311462402,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06879536807537079,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07122,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.33,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 1.9035942554473877,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06245003640651703,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.06444,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 4.195109844207764,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.07456552982330322,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07742,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.6877388954162598,
+      "learning_rate": 6e-05,
+      "loss": 0.08354589343070984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.08714,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 0.8980560898780823,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.01832345873117447,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01849,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.189858317375183,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.028533414006233215,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02894,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.4878139793872833,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013735830783843994,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01383,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1202147006988525,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04460370913147926,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04561,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 39.91,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.2962532043457031,
+      "learning_rate": 8e-05,
+      "loss": 0.0265362411737442,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02689,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 1.1243488788604736,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03802508860826492,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03876,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.8054544925689697,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026396580040454865,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02675,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.0294891595840454,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.02456747367978096,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02487,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 0.6459960341453552,
+      "learning_rate": 9.6e-05,
+      "loss": 0.025260522961616516,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02558,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.1,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 0.7975543141365051,
+      "learning_rate": 0.0001,
+      "loss": 0.02156214788556099,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0218,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7421355247497559,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016790971159934998,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01693,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5165366530418396,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.015923120081424713,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01605,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 37.91,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 0.5790206789970398,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.014571796171367168,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01468,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.5194218754768372,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.010647370480000973,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0107,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.394126057624817,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.025537651032209396,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02587,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.14454102516174316,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0013080333592370152,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00131,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.53,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.4530032873153687,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.025731991976499557,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02607,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.3475311994552612,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.01185566745698452,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01193,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.02,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.6276060938835144,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.01854834333062172,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01872,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.08676943182945251,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.0018136316211894155,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8477522730827332,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018075350672006607,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01824,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.2192176580429077,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.03638514503836632,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03706,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.571501612663269,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.015648968517780304,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01577,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.18,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.7806703448295593,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.030642960220575333,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03112,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.429334431886673,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.014602560549974442,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01471,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8262733221054077,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.027551481500267982,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02793,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.408968061208725,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.013757579028606415,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.01385,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.31,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.41102334856987,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.010365180671215057,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01042,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.709365725517273,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03662724792957306,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03731,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.7385674118995667,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.029125306755304337,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02955,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.2222224771976471,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.005389952100813389,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0054,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 34.92,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.8270087242126465,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.02755359187722206,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02794,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0590053796768188,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.05195138603448868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.05332,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.5024113655090332,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01522495225071907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01534,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.4,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.24432092905044556,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012946993112564087,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01303,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.3318341374397278,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.011563034728169441,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01163,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 36.97,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.8407902121543884,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.04975849390029907,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05102,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.4927586019039154,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020401768386363983,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02061,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 37.75,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.31540000438690186,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.012960189953446388,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01304,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.526395857334137,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.01680367812514305,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01695,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.6426746845245361,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.027429210022091866,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02781,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 1.2340136766433716,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.029638171195983887,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03008,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 33.93,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.2884264290332794,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.009685046970844269,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00973,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 2.1422667503356934,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.028798403218388557,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02922,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.4228186309337616,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.01271561998873949,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0128,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.322550892829895,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.013128789141774178,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01322,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.36,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.31791186332702637,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.010927937924861908,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01099,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.42,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.6686874032020569,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.022430671378970146,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02268,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.116191565990448,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.002666556742042303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00267,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.59,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6079536080360413,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007091412786394358,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00712,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.3987472653388977,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.032372865825891495,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0329,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.6,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.28347039222717285,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00032506848219782114,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00033,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.17,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7291182279586792,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.019354820251464844,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01954,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 33.91,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.4524848163127899,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.006970554124563932,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00699,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.27,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.617595374584198,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008266763761639595,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.0083,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.023282527923584,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.02767460234463215,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02806,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.54,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.5419387817382812,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.017841031774878502,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.018,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7420225143432617,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.019244324415922165,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01943,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.34494075179100037,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007151480298489332,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00718,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3867751657962799,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.011703308671712875,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01177,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4904668629169464,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.021501950919628143,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02173,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6704695224761963,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.014365669339895248,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01447,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.42247557640075684,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.011453501880168915,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01152,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.57,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2647334933280945,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.010307254269719124,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01036,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.24670445919036865,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.004825497046113014,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00484,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.1579471081495285,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.004587843082845211,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.0046,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5054326057434082,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.019181368872523308,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01937,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6734475493431091,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.016313474625349045,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01645,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.4205226004123688,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.005601552315056324,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00562,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.2994498908519745,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.024425672367215157,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02473,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.58,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.6740291714668274,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.028571842238307,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02898,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.42719966173171997,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.009797915816307068,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00985,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.24340543150901794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.009098708629608154,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00914,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 32.95,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.19808582961559296,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.006017087027430534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00604,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 30.98,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7065218687057495,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.019972220063209534,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02017,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.3644750118255615,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.011301878839731216,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01137,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3719140887260437,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.015816055238246918,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01594,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.19392652809619904,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.005229922942817211,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00524,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 36.73,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.5069220066070557,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.0051376414485275745,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00515,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 38.84,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.42843693494796753,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.016977600753307343,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01712,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.17994894087314606,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0033605610951781273,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00337,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6516005992889404,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.04297866299748421,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04392,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 38.87,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.09738968312740326,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0019593359902501106,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00196,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.31,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.3602486848831177,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011135163716971874,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.0112,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.49974870681762695,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.015223942697048187,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01534,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.22547365725040436,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.004879488609731197,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00489,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.68,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.16635224223136902,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.003331358777359128,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00334,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3979133367538452,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.025415543466806412,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02574,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.82,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.0685446709394455,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002011647215113044,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00201,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.38,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3525950014591217,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.019031725823879242,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01921,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.3677102029323578,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01841437630355358,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01858,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 30.82,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.2806461453437805,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.012007143348455429,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01208,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.05,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.29553601145744324,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.02177492529153824,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02201,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.12807513773441315,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.00519968057051301,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00521,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 32.91,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.23824627697467804,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.02345559373497963,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02373,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.20369164645671844,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.011706317774951458,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01178,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.10860611498355865,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.004664955660700798,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 32.83,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.33492082357406616,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.008804760873317719,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00884,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.2093873769044876,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.007556446827948093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00759,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.37,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.2090216428041458,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005284661427140236,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0053,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.59,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.19170638918876648,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.004308023024350405,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00432,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.2729545831680298,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00859704427421093,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00863,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.07509669661521912,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001223657513037324,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 34.62,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 2.083670139312744,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.010420536622405052,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01048,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.09503326565027237,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.001245035557076335,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00125,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.71,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3097395896911621,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007573425769805908,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0076,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.66,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.2731100618839264,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.003785648848861456,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00379,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.52,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.05727197229862213,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0007289001950994134,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00073,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.17,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2666221559047699,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0024705769028514624,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.25296273827552795,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.002646190579980612,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00265,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2845039367675781,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.021563109010457993,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0218,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.37399041652679443,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.004788707476109266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.0048,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 58488
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.6149694919586182,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.012087073177099228,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01216,
+      "step": 129,
+      "tokens/total": 3901504,
+      "tokens/train_per_sec_per_gpu": 32.89,
+      "tokens/trainable": 58938
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.1933072954416275,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.00164338992908597,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00164,
+      "step": 130,
+      "tokens/total": 3931968,
+      "tokens/train_per_sec_per_gpu": 37.08,
+      "tokens/trainable": 59406
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.35414615273475647,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.003997497260570526,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00401,
+      "step": 131,
+      "tokens/total": 3962048,
+      "tokens/train_per_sec_per_gpu": 32.33,
+      "tokens/trainable": 59842
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.3141044080257416,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0038974948693066835,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00391,
+      "step": 132,
+      "tokens/total": 3992512,
+      "tokens/train_per_sec_per_gpu": 35.85,
+      "tokens/trainable": 60316
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.6955724954605103,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.026250183582305908,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0266,
+      "step": 133,
+      "tokens/total": 4022864,
+      "tokens/train_per_sec_per_gpu": 35.6,
+      "tokens/trainable": 60809
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.171626016497612,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.0025996288750320673,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.0026,
+      "step": 134,
+      "tokens/total": 4051344,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 61254
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.6914361119270325,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.013673783279955387,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01377,
+      "step": 135,
+      "tokens/total": 4081728,
+      "tokens/train_per_sec_per_gpu": 36.12,
+      "tokens/trainable": 61725
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.4999908208847046,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.010643654502928257,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.0107,
+      "step": 136,
+      "tokens/total": 4111904,
+      "tokens/train_per_sec_per_gpu": 38.4,
+      "tokens/trainable": 62200
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.41597622632980347,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.0126111451536417,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01269,
+      "step": 137,
+      "tokens/total": 4142208,
+      "tokens/train_per_sec_per_gpu": 36.57,
+      "tokens/trainable": 62698
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.3869597017765045,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.01739512011408806,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01755,
+      "step": 138,
+      "tokens/total": 4172512,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 63158
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.16373302042484283,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.004722995217889547,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00473,
+      "step": 139,
+      "tokens/total": 4202944,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 63638
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.4420141875743866,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.00948841031640768,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.00953,
+      "step": 140,
+      "tokens/total": 4231264,
+      "tokens/train_per_sec_per_gpu": 33.2,
+      "tokens/trainable": 64055
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.24656501412391663,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.005166166462004185,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00518,
+      "step": 141,
+      "tokens/total": 4261488,
+      "tokens/train_per_sec_per_gpu": 35.19,
+      "tokens/trainable": 64512
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.10694887489080429,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0021838685497641563,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00219,
+      "step": 142,
+      "tokens/total": 4292032,
+      "tokens/train_per_sec_per_gpu": 34.05,
+      "tokens/trainable": 64950
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.2847970128059387,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.007552271708846092,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00758,
+      "step": 143,
+      "tokens/total": 4322032,
+      "tokens/train_per_sec_per_gpu": 37.75,
+      "tokens/trainable": 65429
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.18000122904777527,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.003006678307428956,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00301,
+      "step": 144,
+      "tokens/total": 4352128,
+      "tokens/train_per_sec_per_gpu": 36.82,
+      "tokens/trainable": 65895
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.03157320246100426,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.0008168672793544829,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00082,
+      "step": 145,
+      "tokens/total": 4382368,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 66327
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.2770066559314728,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.008716664277017117,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00875,
+      "step": 146,
+      "tokens/total": 4412640,
+      "tokens/train_per_sec_per_gpu": 31.65,
+      "tokens/trainable": 66778
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.09778914600610733,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.0020899923983961344,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00209,
+      "step": 147,
+      "tokens/total": 4442896,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 67192
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.15238171815872192,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.0016652008052915335,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00167,
+      "step": 148,
+      "tokens/total": 4473248,
+      "tokens/train_per_sec_per_gpu": 35.47,
+      "tokens/trainable": 67677
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.17436973750591278,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.0012439386919140816,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00124,
+      "step": 149,
+      "tokens/total": 4503664,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 68121
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.029765604063868523,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0003211075090803206,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00032,
+      "step": 150,
+      "tokens/total": 4533760,
+      "tokens/train_per_sec_per_gpu": 33.37,
+      "tokens/trainable": 68589
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.3771016001701355,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.0040406654588878155,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00405,
+      "step": 151,
+      "tokens/total": 4563872,
+      "tokens/train_per_sec_per_gpu": 30.3,
+      "tokens/trainable": 69055
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.5529465675354004,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.010513005778193474,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01057,
+      "step": 152,
+      "tokens/total": 4594384,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 69501
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.01583116315305233,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.00019431026885285974,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00019,
+      "step": 153,
+      "tokens/total": 4624816,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 69964
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.12653011083602905,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0008956181118264794,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0009,
+      "step": 154,
+      "tokens/total": 4655136,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 70431
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.5188359022140503,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.007073995191603899,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0071,
+      "step": 155,
+      "tokens/total": 4683488,
+      "tokens/train_per_sec_per_gpu": 37.49,
+      "tokens/trainable": 70875
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.567967414855957,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.009182224050164223,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00922,
+      "step": 156,
+      "tokens/total": 4714032,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 71333
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.06565697491168976,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.000622953346464783,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00062,
+      "step": 157,
+      "tokens/total": 4744192,
+      "tokens/train_per_sec_per_gpu": 35.07,
+      "tokens/trainable": 71795
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.9077213406562805,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.018462298437952995,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01863,
+      "step": 158,
+      "tokens/total": 4774592,
+      "tokens/train_per_sec_per_gpu": 32.24,
+      "tokens/trainable": 72231
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 1.065195083618164,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.03022950328886509,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03069,
+      "step": 159,
+      "tokens/total": 4804672,
+      "tokens/train_per_sec_per_gpu": 30.62,
+      "tokens/trainable": 72660
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.09623221307992935,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.0023340554907917976,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00234,
+      "step": 160,
+      "tokens/total": 4835056,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 73115
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.37435606122016907,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.0122139323502779,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.01229,
+      "step": 161,
+      "tokens/total": 4865296,
+      "tokens/train_per_sec_per_gpu": 31.1,
+      "tokens/trainable": 73536
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 0.7639496922492981,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.02236529439687729,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.02262,
+      "step": 162,
+      "tokens/total": 4895872,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 73985
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.17859485745429993,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.008786039426922798,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00882,
+      "step": 163,
+      "tokens/total": 4926192,
+      "tokens/train_per_sec_per_gpu": 35.35,
+      "tokens/trainable": 74449
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.18802772462368011,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.006680628284811974,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0067,
+      "step": 164,
+      "tokens/total": 4956784,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 74900
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.23258115351200104,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.0070136841386556625,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00704,
+      "step": 165,
+      "tokens/total": 4987120,
+      "tokens/train_per_sec_per_gpu": 30.63,
+      "tokens/trainable": 75324
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.17548920214176178,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.007272098213434219,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.0073,
+      "step": 166,
+      "tokens/total": 5017424,
+      "tokens/train_per_sec_per_gpu": 31.88,
+      "tokens/trainable": 75765
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.18047791719436646,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.005449052434414625,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00546,
+      "step": 167,
+      "tokens/total": 5048096,
+      "tokens/train_per_sec_per_gpu": 31.38,
+      "tokens/trainable": 76226
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.29115262627601624,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.006910712458193302,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00693,
+      "step": 168,
+      "tokens/total": 5078608,
+      "tokens/train_per_sec_per_gpu": 33.39,
+      "tokens/trainable": 76677
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.1797957718372345,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.0023017844650894403,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0023,
+      "step": 169,
+      "tokens/total": 5108944,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 77109
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.296197772026062,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.00650596572086215,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00653,
+      "step": 170,
+      "tokens/total": 5139344,
+      "tokens/train_per_sec_per_gpu": 34.19,
+      "tokens/trainable": 77558
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.1834728717803955,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.003503798507153988,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00351,
+      "step": 171,
+      "tokens/total": 5169600,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 78018
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.17331187427043915,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.0018293634057044983,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00183,
+      "step": 172,
+      "tokens/total": 5200032,
+      "tokens/train_per_sec_per_gpu": 29.87,
+      "tokens/trainable": 78472
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.6207812428474426,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.012318727560341358,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01239,
+      "step": 173,
+      "tokens/total": 5230464,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 78919
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.5858636498451233,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.01317191869020462,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.62,
+      "memory/max_allocated (GiB)": 33.62,
+      "ppl": 1.01326,
+      "step": 174,
+      "tokens/total": 5260192,
+      "tokens/train_per_sec_per_gpu": 30.21,
+      "tokens/trainable": 79336
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.19330058991909027,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.004472787491977215,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00448,
+      "step": 175,
+      "tokens/total": 5290480,
+      "tokens/train_per_sec_per_gpu": 32.6,
+      "tokens/trainable": 79806
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.6980848908424377,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.005580209661275148,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0056,
+      "step": 176,
+      "tokens/total": 5320992,
+      "tokens/train_per_sec_per_gpu": 36.25,
+      "tokens/trainable": 80284
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.0029784892685711384,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 9.989955287892371e-05,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0001,
+      "step": 177,
+      "tokens/total": 5351328,
+      "tokens/train_per_sec_per_gpu": 32.68,
+      "tokens/trainable": 80731
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.1398579478263855,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.001107236254028976,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00111,
+      "step": 178,
+      "tokens/total": 5381536,
+      "tokens/train_per_sec_per_gpu": 34.88,
+      "tokens/trainable": 81199
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.3136104941368103,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.003412305610254407,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00342,
+      "step": 179,
+      "tokens/total": 5411536,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 81646
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.5958685874938965,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.013216094113886356,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.0133,
+      "step": 180,
+      "tokens/total": 5441712,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 82119
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.26500052213668823,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.004519693087786436,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00453,
+      "step": 181,
+      "tokens/total": 5471968,
+      "tokens/train_per_sec_per_gpu": 38.88,
+      "tokens/trainable": 82636
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.31901440024375916,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.00207512523047626,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00208,
+      "step": 182,
+      "tokens/total": 5502320,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 83116
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.024084335193037987,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.0005284445942379534,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00053,
+      "step": 183,
+      "tokens/total": 5532512,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 83563
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.03849077597260475,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.0005660011665895581,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00057,
+      "step": 184,
+      "tokens/total": 5562992,
+      "tokens/train_per_sec_per_gpu": 35.56,
+      "tokens/trainable": 84069
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.03266207128763199,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.0005896957591176033,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00059,
+      "step": 185,
+      "tokens/total": 5593200,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 84544
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.5690993070602417,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.009490721859037876,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00954,
+      "step": 186,
+      "tokens/total": 5623568,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 85005
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.024852192029356956,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0004194313660264015,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00042,
+      "step": 187,
+      "tokens/total": 5653856,
+      "tokens/train_per_sec_per_gpu": 31.78,
+      "tokens/trainable": 85445
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.05216141417622566,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.0008849176811054349,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00089,
+      "step": 188,
+      "tokens/total": 5684112,
+      "tokens/train_per_sec_per_gpu": 34.01,
+      "tokens/trainable": 85935
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.014727079309523106,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.00033978797728195786,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00034,
+      "step": 189,
+      "tokens/total": 5714400,
+      "tokens/train_per_sec_per_gpu": 36.25,
+      "tokens/trainable": 86450
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.6413062810897827,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.013006060384213924,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01309,
+      "step": 190,
+      "tokens/total": 5744992,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 86921
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.44923287630081177,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.002187589881941676,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00219,
+      "step": 191,
+      "tokens/total": 5775568,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 87376
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.05658087879419327,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.0006040545413270593,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0006,
+      "step": 192,
+      "tokens/total": 5806000,
+      "tokens/train_per_sec_per_gpu": 31.53,
+      "tokens/trainable": 87806
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.94087148777664e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-192/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b
+size 8273

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-224/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/wave/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/wave/parent
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-224/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "gate_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3956e5cc16d1fafb4f276fcebad34769093b8a735ea703af8e7aa72777957498
+size 547777976

aft_wave_v2/coin_real_4x__coin2/training/checkpoints/checkpoint-224/chat_template.jinja

@@ -0,0 +1,47 @@
+{{ bos_token }}
+{%- if messages[0]['role'] == 'system' -%}
+    {%- if messages[0]['content'] is string -%}
+        {%- set first_user_prefix = messages[0]['content'] + '
+
+' -%}
+    {%- else -%}
+        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
+
+' -%}
+    {%- endif -%}
+    {%- set loop_messages = messages[1:] -%}
+{%- else -%}
+    {%- set first_user_prefix = "" -%}
+    {%- set loop_messages = messages -%}
+{%- endif -%}
+{%- for message in loop_messages -%}
+    {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
+        {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
+    {%- endif -%}
+    {%- if (message['role'] == 'assistant') -%}
+        {%- set role = "model" -%}
+    {%- else -%}
+        {%- set role = message['role'] -%}
+    {%- endif -%}
+    {{ '<start_of_turn>' + role + '
+' + (first_user_prefix if loop.first else "") }}
+    {%- if message['content'] is string -%}
+        {{ message['content'] | trim }}
+    {%- elif message['content'] is iterable -%}
+        {%- for item in message['content'] -%}
+            {%- if item['type'] == 'image' -%}
+                {{ '<start_of_image>' }}
+            {%- elif item['type'] == 'text' -%}
+                {{ item['text'] | trim }}
+            {%- endif -%}
+        {%- endfor -%}
+    {%- else -%}
+        {{ raise_exception("Invalid content type") }}
+    {%- endif -%}
+    {{ '<end_of_turn>
+' }}
+{%- endfor -%}
+{%- if add_generation_prompt -%}
+    {{'<start_of_turn>model
+'}}
+{%- endif -%}