dispatch-sdf-aft-v1: aft_wave_v2/coin_real_4x__coin0p2/training

.gitattributes

@@ -460,3 +460,20 @@ aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-512/toke
 aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/control_matched__charter0p2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/control_matched__charter0p2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+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__coin0p2/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,867 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/coin_real_4x__coin0p2/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 962,
+      "sha256": "5fa578341331c9d290fb8539f3467c2a0e0874c3611d4344c231b867fe17d5d8"
+    },
+    "axolotl.yaml": {
+      "size": 1209,
+      "sha256": "ff0dd0239a28c6e029c5a998e7af1d24f1992fe9ad6c681b2519ea28065459b6"
+    },
+    "checkpoint.json": {
+      "size": 2208,
+      "sha256": "f4e2ad56e53297d952a050a601cbbae1a94aa6426e6e9ad5e20bb157cd369cdc"
+    },
+    "checkpoints/README.md": {
+      "size": 2931,
+      "sha256": "1d12e1b0bad24e925d77aa4ab1e5971a41fa1395648b0da713fdece8003c0905"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "004757923662cc6573db04eb1edf3bcc6f5152f20b0e388191af7ad75b8ba2f5"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4fc2babf0e54f7a72a2f3faf84eaa36b6c79b527536e07ce1841f5e43580c86b"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "31ebd596a72dab4299216ee54f745a345beefed2232502312721b27dead61c31"
+    },
+    "checkpoints/checkpoint-128/rng_state.pth": {
+      "size": 14645,
+      "sha256": "234e8c0c3fd2b805b594ff492259b2db7aead6c3ca1abbea3168b40935e4d0cb"
+    },
+    "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": "5d4e623cef80604dfe57d301601c79f30656bf80c690d125a68a0d329f78f80c"
+    },
+    "checkpoints/checkpoint-128/trainer_state.json": {
+      "size": 56221,
+      "sha256": "590883bd12a98ddb9fb206bba6fd1e1db6aa9a48263df74d571086b6a8e94886"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "f18e56b80e7ad24c74c6f7090f4f740933e44ef72100c8a8432894e8c258c433"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "07b4ab83e765b9a9fef6d5ad5b0bbf7bcffcc6ee72a12963150438f05716abe2"
+    },
+    "checkpoints/checkpoint-160/rng_state.pth": {
+      "size": 14645,
+      "sha256": "62012b9e748bbef1beb75187c13d345f2d456800a4ec5207b243fb40c95ada94"
+    },
+    "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": "6d2abd602326bcb4dd54f0732279c81f75489fb06cc9c26b3a1290eac1d64a69"
+    },
+    "checkpoints/checkpoint-160/trainer_state.json": {
+      "size": 70198,
+      "sha256": "3fc621af00e4f23772e5259b0934379cabaa9495a4682f38e229c3a9a02c5844"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "85ed8480f6ee209a38ee9cd571b15a1b6c5ce56607a32188e14d48492e058adf"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "a86ff874a8c3cc41a595be2ed4adada5891f02ffa60010a45ced596f3483975a"
+    },
+    "checkpoints/checkpoint-192/rng_state.pth": {
+      "size": 14645,
+      "sha256": "3d69b2462378ed67f5f5555fb6b192965044d7652528913ba5ccc13c8e15172a"
+    },
+    "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": "d0bcbe73caf27bad23d336499deca35036208038fc18e90b69ed1271982cac66"
+    },
+    "checkpoints/checkpoint-192/trainer_state.json": {
+      "size": 84189,
+      "sha256": "5689aa913851f531f40d7e32eed5f4e24b92c403457a661581f829bbd513c9c6"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "220fd03ef151db83f45a94e4d3e1a971e762411071355fee57f461304bd52e35"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "05c3c1f0a5d74da8e16bfeba10f5ebec805dcc1dce9c651ef3032a4576a630e4"
+    },
+    "checkpoints/checkpoint-224/rng_state.pth": {
+      "size": 14645,
+      "sha256": "496ba904d564744600c2dc9631c09eb565ed0719573f403cd75b1e83021c01e1"
+    },
+    "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": "837aaa06bfc89ca985eb3d887a0924f1a11a28f8aa3e0d45204320cda4d12b5e"
+    },
+    "checkpoints/checkpoint-224/trainer_state.json": {
+      "size": 98212,
+      "sha256": "9cc58420475dae5d789df25bdc999682ac93967119b83fa5a1ef0b4b32889142"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "e2ba4d31dd353bd2ca9294a219e16708e9d8a591bf19e71b06e213df8ed69bea"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "eee1b94dd3bdec4d438ef940db0e138077560c7d35ee39467be8687bf23e98bc"
+    },
+    "checkpoints/checkpoint-256/rng_state.pth": {
+      "size": 14645,
+      "sha256": "370b4fba725358746db83895d8635e99b2f70cb36421eb17a0433851f5afd453"
+    },
+    "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": "d4929d8b8f64e42e8d625a7e18a46658293bb00c21cb89ef543c8423240df68a"
+    },
+    "checkpoints/checkpoint-256/trainer_state.json": {
+      "size": 112266,
+      "sha256": "07a486030d1c51e736c694a15ab608049e0a6cd9ad2658e5b040a85a16af4e25"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "d181ae74f1617158aaf65f734801ceae0768304d6027b9e77606e15c50b81dac"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "5ca99fac2ee7672f86c7d237de8447789263ea6631bce4dd83c9efe25c9082bd"
+    },
+    "checkpoints/checkpoint-288/rng_state.pth": {
+      "size": 14645,
+      "sha256": "47ff3a93a2ca5e4e3e5cc87da30e0868111aa8accfb80c30aa2ec9467da212fb"
+    },
+    "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": "af477bcb3b897a3f8f3e8da2dd3a8ebe6cce451be8ae5d737b9cb6ea326ef2d9"
+    },
+    "checkpoints/checkpoint-288/trainer_state.json": {
+      "size": 126338,
+      "sha256": "c4c12fb931ec944285efbccc02392b4a02ae98e484bbcc870cd033aac1db78df"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "8e42b5dd58a02736989a8d1a457227b96188ee01f7f469d682169a85ba93d83a"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "c29ad0f29115d6a62b344b2719754dd79fc85284138b433bd8db4d102d2bbd2d"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "046455cb6ded9b2989f35bf5efbbff0517d6651771b52c93d31759906cafe479"
+    },
+    "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": "cec49603be80ef1df2a4dadb25de3e8be8585000fa9c1e62970e67efa8f9d602"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14416,
+      "sha256": "11d44104bbff25df8f8dd6f9724369eed58a187190f359cd22ad52d92b39cd96"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ce063233756c13a0b13b3e3de54a4f0d7aca4320fdbee2aebc8257408e81990b"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "736cb10079d37cfe6e39fc324a83a03cce6bd3dbc2bbadde1513e08d4816b68d"
+    },
+    "checkpoints/checkpoint-320/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0391605f7d7b8626c1c97d35086145a3e9e4df5cafc3842162a7664ce0f73f43"
+    },
+    "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": "5dba8649e21e024152b4b3daeba662208bb0f33796ac11478c9b0d7cffa65e97"
+    },
+    "checkpoints/checkpoint-320/trainer_state.json": {
+      "size": 140427,
+      "sha256": "7ca9473d3142d20df35f85e42aea50f11e14e9d3fc83e5da9ea8351041294c9a"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "43e130a15d3a546161129a78748f0fd0757c0075942046a32d1aa2551e313a6c"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "d069a87fce3739148e9df98672687b4e7b8271e1564bda4fa7f13e5ed01c6d04"
+    },
+    "checkpoints/checkpoint-352/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a5d3595bca569f2399926c6c523fd98aead95fc6d00353c61841e5028ebbdeff"
+    },
+    "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": "8b8f1db9f0311bacf95a403fdc167f8bce05fa9b5631e294453130a08f29b9c4"
+    },
+    "checkpoints/checkpoint-352/trainer_state.json": {
+      "size": 154545,
+      "sha256": "5cb5f2bbdb85f774b4c86b35d1ae66b626c1da8ec6d9a6cbda01952fb1096c19"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "19d460cb5f3eef30a27eeda49a98266c4282f6cbae261428e40ebba0ff9681da"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "76c47b870ab9f0fa5d81e16be90897cfb5f3c32449a223f59a3d50150c3df02c"
+    },
+    "checkpoints/checkpoint-384/rng_state.pth": {
+      "size": 14645,
+      "sha256": "ea55d7f6de0215cad74ef9143c41f13fa0fffc3e1c9c7025b8d1c96ebbb6f69e"
+    },
+    "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": "de563327ae6aae9986227588641dbd4a73c35ac9737ce28dd8b18d6807eb841f"
+    },
+    "checkpoints/checkpoint-384/trainer_state.json": {
+      "size": 168694,
+      "sha256": "c5c27504df47c098f0c272f420bcd918ca0f1f41d70bfaa1c4ad66e54d2a902d"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "2d0866c11a9e57490eea883b729339cbc59d5006d03a1a4b4a1e1f4a005bd6f5"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "9670a3fe3860e7b056f50a7f7c4276e0268cd918775bc811b8a7c463351da4fa"
+    },
+    "checkpoints/checkpoint-416/rng_state.pth": {
+      "size": 14645,
+      "sha256": "49f752bd9a9a6e4604278f4cbc2ad9ae381068763e6148938859abe0b809968f"
+    },
+    "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": "652b2133eb54c05ce95ac81317114e9b6a0e614398fea583a61e940a40659c7a"
+    },
+    "checkpoints/checkpoint-416/trainer_state.json": {
+      "size": 182860,
+      "sha256": "18dca420b73205a3384c6fc22eba4f82d942c2dbc9845f77f40e25262a1b47be"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "8629ea27947d652d3d99b6309d975cd085658fb0bce9b9309f847b13054f480b"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "3ba090668a36670fb601cf514daf31a4efeaa3a4e7c505735817d056fd2001d3"
+    },
+    "checkpoints/checkpoint-448/rng_state.pth": {
+      "size": 14645,
+      "sha256": "f0f5fea350892f85d3bbab116a9d5db62bdcedcd47807fd37fd8d790fcd8718a"
+    },
+    "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": "10af6c079666802d71ae9c72bb1cbc9f22808f4d291cf6e573a5b2e67474c845"
+    },
+    "checkpoints/checkpoint-448/trainer_state.json": {
+      "size": 197010,
+      "sha256": "256426ffcb17ece56a204fd53eb57a0b3406aec06e55aa5d8ba680269d6376b5"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a639f0cd54e96ad33bff59a38f6b3618936c5b19c68f246d12a117646577484c"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "eeb38f160641fd181dd0131ae0bbfd45e58e9a572ec1090ba85363fc61743fbb"
+    },
+    "checkpoints/checkpoint-480/rng_state.pth": {
+      "size": 14645,
+      "sha256": "6cc653826760fd60176ec4de753ff8ff573e43d06826f127f29d6db37ba7f969"
+    },
+    "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": "77c796414a49d4ddc04c6e228fe890623cdead947a9c487fc3d8b63880ca27e4"
+    },
+    "checkpoints/checkpoint-480/trainer_state.json": {
+      "size": 211173,
+      "sha256": "133261f11c867ab76d2685388c77365d099769d894d357d752938bee9ac5086e"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "004757923662cc6573db04eb1edf3bcc6f5152f20b0e388191af7ad75b8ba2f5"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "65916b4beb82a197b99e16dd2db1ea96ed374c4fc8a805d7a27ee6ac4b30a97a"
+    },
+    "checkpoints/checkpoint-512/rng_state.pth": {
+      "size": 14645,
+      "sha256": "75e98dcceb68f1b6ee34a96307625206231cf5a0e0e4e16d8a9eafccc33ea3fe"
+    },
+    "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": "fc4826883a470f621364e7e63d0e124d9cfbb8e82cc14c353e3789ba20140b7c"
+    },
+    "checkpoints/checkpoint-512/trainer_state.json": {
+      "size": 225339,
+      "sha256": "214dd5dfc66104c85f77093435c580bad895ebea268d82f43a995d8a6de48a2f"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4b584e3cf8f378cc0d06c9d8e81b75d4ebec54e9ef2e3bc83848da1c57396f4f"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "7f8513c58f406f91f4e3f9606a17ac066ff4724ad1843f31a0343e812b8d82fa"
+    },
+    "checkpoints/checkpoint-64/rng_state.pth": {
+      "size": 14645,
+      "sha256": "21d00dc563d713980dd7a1d4b126d93ab060fe78b7ba0c4c92f92cd25a992569"
+    },
+    "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": "abe2b3d625c28fc391b4b32077194735a0210c1485305bfd5103fbcb44a1be6f"
+    },
+    "checkpoints/checkpoint-64/trainer_state.json": {
+      "size": 28341,
+      "sha256": "3c5eb2a6cc04639bfec45f1939f7237f7564c2fd9fcd70698bbe07b72c3a1099"
+    },
+    "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": "342c73d74da62ac2cef52bcfc2ae0971436571e32b7edd7325cfebfd02306622"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c9a211aba75f9a340dc57de855c8e2b458e1b586134e0dc1ad3988647f0aa4e9"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8c5a9fb4a808c243ca1d455c97cb546577f69e79453fc34bf8fae30def5fa9a8"
+    },
+    "checkpoints/checkpoint-96/rng_state.pth": {
+      "size": 14645,
+      "sha256": "58b2d07d478cf0253d5619ad2e9d23f5e40a65de8495503676e2505cd9d2856a"
+    },
+    "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": "0b71c52644fe1614b0f38698396e8137456254e1ccd3cdb601c7d33938f83ceb"
+    },
+    "checkpoints/checkpoint-96/trainer_state.json": {
+      "size": 42287,
+      "sha256": "245feb515f34fcec3f22e15c1f38d2ee0048e67dee2620e28321f5b100b120c3"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 269203,
+      "sha256": "94ab3fcb79b1485d1dcb2cb4fe45f280ed6de99b12ac070e1d98a73c2314bc0e"
+    },
+    "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__coin0p2.txt": {
+      "size": 52,
+      "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90"
+    },
+    "config/aft_dispatch_v4_wide.yaml": {
+      "size": 2948,
+      "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc"
+    },
+    "config/axolotl.yaml": {
+      "size": 1209,
+      "sha256": "ff0dd0239a28c6e029c5a998e7af1d24f1992fe9ad6c681b2519ea28065459b6"
+    },
+    "health/training_started.json": {
+      "size": 137,
+      "sha256": "fe44a4371a7b6185814e095796cfa97d265bfd6ab9823dc26cc9b95b1cc9dd06"
+    },
+    "run.json": {
+      "size": 405,
+      "sha256": "b08ec6049e6f3fa743e7c82a8caea668b5dcff1ba81557b0c540eb5136737a4a"
+    },
+    "train.log": {
+      "size": 276595,
+      "sha256": "341baeec1d4bc5f5452865434c2d00a0098c829143b4e8cc44cf8ffdea2faa60"
+    },
+    "trainer_state.final.json": {
+      "size": 225339,
+      "sha256": "214dd5dfc66104c85f77093435c580bad895ebea268d82f43a995d8a6de48a2f"
+    },
+    "training_examples.jsonl": {
+      "size": 3159608,
+      "sha256": "dd0cc64e6ffd90ae6e12011517a011c9dbcebed1a51771f305f8702eb9719e9e"
+    },
+    "training_provenance.json": {
+      "size": 4086,
+      "sha256": "823bf968f76b9fd7817bff674e27e152b2e94fd2d06f71456a6d8878233b096d"
+    },
+    "training_trace.jsonl": {
+      "size": 182071,
+      "sha256": "880b3973f57663c4fac7f55137ef8c6a9a90f3672d4572f397e946a1488c2cf9"
+    }
+  }
+}

aft_wave_v2/coin_real_4x__coin0p2/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "coin_real_4x__coin0p2",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "sft_4epoch/coin/checkpoint-48",
+  "dataset_sha256": "bf34ebe8a30ee23f25ff7e532974b5fc79d0231b4eb0393ba92b50b266c493b9",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 59.31,
+  "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__coin0p2/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_coin0p2.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__coin0p2/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:coin_real_4x__coin0p2",
+  "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_coin0p2.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__coin0p2",
+  "pointer_file": "/workspace/wave/training/ckpt_coin_real_4x__coin0p2.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__coin0p2",
+    "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_coin0p2.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__coin0p2",
+    "pointer_file": "/workspace/wave/training/ckpt_coin_real_4x__coin0p2.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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_coin0p2.jsonl
+pipeline_tag: text-generation
+base_model: /workspace/wave/parent
+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_coin0p2.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_coin0p2.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__coin0p2/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",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "gate_proj",
+    "q_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__coin0p2/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:004757923662cc6573db04eb1edf3bcc6f5152f20b0e388191af7ad75b8ba2f5
+size 547777976

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "gate_proj",
+    "q_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__coin0p2/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fc2babf0e54f7a72a2f3faf84eaa36b6c79b527536e07ce1841f5e43580c86b
+size 547777976

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31ebd596a72dab4299216ee54f745a345beefed2232502312721b27dead61c31
+size 1048106435

aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-128/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:234e8c0c3fd2b805b594ff492259b2db7aead6c3ca1abbea3168b40935e4d0cb
+size 14645

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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__coin0p2/training/checkpoints/checkpoint-128/tokens_state.json

@@ -0,0 +1 @@
+{"total": 3872688, "trainable": 58451}

aft_wave_v2/coin_real_4x__coin0p2/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": 1.000339388847351,
+      "learning_rate": 0.0,
+      "loss": 0.11221420764923096,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.11875,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.52,
+      "tokens/trainable": 417
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8961698412895203,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11656901985406876,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.12364,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 873
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.5287365913391113,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11233991384506226,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11889,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 1353
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.3438372611999512,
+      "learning_rate": 1.2e-05,
+      "loss": 0.10132055729627609,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10663,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.26,
+      "tokens/trainable": 1778
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.0121135711669922,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.11346302926540375,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.12015,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 2261
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 3.9203455448150635,
+      "learning_rate": 2e-05,
+      "loss": 0.09780596196651459,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10275,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 2705
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.63088059425354,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06366116553544998,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06573,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.89,
+      "tokens/trainable": 3136
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.9995646476745605,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0759042352437973,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07886,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 3579
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0774554014205933,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04145222157239914,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04232,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 4000
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9648898839950562,
+      "learning_rate": 3.6e-05,
+      "loss": 0.032875221222639084,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03342,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 37.67,
+      "tokens/trainable": 4474
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 0.5937227606773376,
+      "learning_rate": 4e-05,
+      "loss": 0.023198626935482025,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02347,
+      "step": 11,
+      "tokens/total": 333296,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 4942
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.294079303741455,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.0510174036026001,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05234,
+      "step": 12,
+      "tokens/total": 363840,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 5424
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.7166640758514404,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06956956535577774,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07205,
+      "step": 13,
+      "tokens/total": 394112,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 5855
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.1496169567108154,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06250961124897003,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0645,
+      "step": 14,
+      "tokens/total": 424656,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 6344
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.515660047531128,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09882807731628418,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.10388,
+      "step": 15,
+      "tokens/total": 455152,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 6825
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.3756372928619385,
+      "learning_rate": 6e-05,
+      "loss": 0.0719379335641861,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.07459,
+      "step": 16,
+      "tokens/total": 485328,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 7230
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.7220206260681152,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.016590356826782227,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01673,
+      "step": 17,
+      "tokens/total": 515744,
+      "tokens/train_per_sec_per_gpu": 35.89,
+      "tokens/trainable": 7699
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.5261812210083008,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.030052796006202698,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.03051,
+      "step": 18,
+      "tokens/total": 546144,
+      "tokens/train_per_sec_per_gpu": 33.85,
+      "tokens/trainable": 8184
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.6416668891906738,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013492297381162643,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01358,
+      "step": 19,
+      "tokens/total": 576560,
+      "tokens/train_per_sec_per_gpu": 34.91,
+      "tokens/trainable": 8675
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.2296113967895508,
+      "learning_rate": 7.6e-05,
+      "loss": 0.038845278322696686,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03961,
+      "step": 20,
+      "tokens/total": 607056,
+      "tokens/train_per_sec_per_gpu": 39.72,
+      "tokens/trainable": 9189
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.8426793813705444,
+      "learning_rate": 8e-05,
+      "loss": 0.01696154847741127,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01711,
+      "step": 21,
+      "tokens/total": 637488,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 9614
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 10.310348510742188,
+      "learning_rate": 8.4e-05,
+      "loss": 0.06855659186840057,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.07096,
+      "step": 22,
+      "tokens/total": 667632,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10086
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.412604808807373,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.04118483513593674,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04204,
+      "step": 23,
+      "tokens/total": 697872,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 10535
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.4078032970428467,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.025748739019036293,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02608,
+      "step": 24,
+      "tokens/total": 728432,
+      "tokens/train_per_sec_per_gpu": 31.6,
+      "tokens/trainable": 10993
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 1.0874605178833008,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03088424727320671,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03137,
+      "step": 25,
+      "tokens/total": 756656,
+      "tokens/train_per_sec_per_gpu": 40.03,
+      "tokens/trainable": 11417
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.2574082612991333,
+      "learning_rate": 0.0001,
+      "loss": 0.03432866185903549,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03492,
+      "step": 26,
+      "tokens/total": 787120,
+      "tokens/train_per_sec_per_gpu": 34.36,
+      "tokens/trainable": 11874
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.30990996956825256,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.010750269517302513,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01081,
+      "step": 27,
+      "tokens/total": 817504,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 12316
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5124737620353699,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.01967313140630722,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01987,
+      "step": 28,
+      "tokens/total": 848128,
+      "tokens/train_per_sec_per_gpu": 37.86,
+      "tokens/trainable": 12800
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.2975701093673706,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.0434565395116806,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04441,
+      "step": 29,
+      "tokens/total": 878448,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 13270
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.4117317795753479,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.013137167319655418,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01322,
+      "step": 30,
+      "tokens/total": 909008,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 13752
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.5875954627990723,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.01785588450729847,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01802,
+      "step": 31,
+      "tokens/total": 939312,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 14218
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7800514698028564,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.011262159794569016,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01133,
+      "step": 32,
+      "tokens/total": 969264,
+      "tokens/train_per_sec_per_gpu": 34.49,
+      "tokens/trainable": 14655
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.0509369373321533,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04942459613084793,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.05067,
+      "step": 33,
+      "tokens/total": 999792,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 15087
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.6234491467475891,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.028202136978507042,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0286,
+      "step": 34,
+      "tokens/total": 1029840,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 15548
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6843816041946411,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.02795713022351265,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02835,
+      "step": 35,
+      "tokens/total": 1060416,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 16037
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.29489848017692566,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.010517537593841553,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01057,
+      "step": 36,
+      "tokens/total": 1090464,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16475
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.7335977554321289,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018923252820968628,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0191,
+      "step": 37,
+      "tokens/total": 1120944,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 16919
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.5587971210479736,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.020957889035344124,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02118,
+      "step": 38,
+      "tokens/total": 1151360,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17333
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 1.7112773656845093,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.045844268053770065,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.04691,
+      "step": 39,
+      "tokens/total": 1181728,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 17806
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.2373344898223877,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.04578991234302521,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04685,
+      "step": 40,
+      "tokens/total": 1212064,
+      "tokens/train_per_sec_per_gpu": 35.65,
+      "tokens/trainable": 18290
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.8950033187866211,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.02866373583674431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.02908,
+      "step": 41,
+      "tokens/total": 1242592,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 18747
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.9967094659805298,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02229372225701809,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02254,
+      "step": 42,
+      "tokens/total": 1272800,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 19172
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.1990034282207489,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.0059759169816970825,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00599,
+      "step": 43,
+      "tokens/total": 1303600,
+      "tokens/train_per_sec_per_gpu": 31.32,
+      "tokens/trainable": 19622
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.47151026129722595,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.011378830298781395,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01144,
+      "step": 44,
+      "tokens/total": 1333936,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 20073
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.5381685495376587,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.02141587622463703,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02165,
+      "step": 45,
+      "tokens/total": 1364288,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 20497
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 1.0174895524978638,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.02236388623714447,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02262,
+      "step": 46,
+      "tokens/total": 1394368,
+      "tokens/train_per_sec_per_gpu": 34.16,
+      "tokens/trainable": 20940
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.5081813931465149,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.010915009304881096,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.01097,
+      "step": 47,
+      "tokens/total": 1425024,
+      "tokens/train_per_sec_per_gpu": 35.04,
+      "tokens/trainable": 21405
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.4684470891952515,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.009690589271485806,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00974,
+      "step": 48,
+      "tokens/total": 1455392,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21828
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.49191179871559143,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.010902130044996738,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01096,
+      "step": 49,
+      "tokens/total": 1485616,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 22241
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.44692206382751465,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.006928037852048874,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00695,
+      "step": 50,
+      "tokens/total": 1515984,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 22688
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.37972721457481384,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.00806603953242302,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0081,
+      "step": 51,
+      "tokens/total": 1546224,
+      "tokens/train_per_sec_per_gpu": 32.12,
+      "tokens/trainable": 23136
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.0114665031433105,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013657575473189354,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01375,
+      "step": 52,
+      "tokens/total": 1576528,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 23631
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.6438844799995422,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05178550258278847,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05315,
+      "step": 53,
+      "tokens/total": 1606768,
+      "tokens/train_per_sec_per_gpu": 34.15,
+      "tokens/trainable": 24096
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 3.0008814334869385,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.010915370658040047,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01098,
+      "step": 54,
+      "tokens/total": 1636640,
+      "tokens/train_per_sec_per_gpu": 37.95,
+      "tokens/trainable": 24569
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.9377115368843079,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.019770421087741852,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01997,
+      "step": 55,
+      "tokens/total": 1667136,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 25067
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.23384763300418854,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.004394679330289364,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0044,
+      "step": 56,
+      "tokens/total": 1697472,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 25541
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.7558876276016235,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.02614775486290455,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02649,
+      "step": 57,
+      "tokens/total": 1727680,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 25965
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.4615328311920166,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01031234860420227,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01037,
+      "step": 58,
+      "tokens/total": 1758192,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 26409
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.40572404861450195,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.011052684858441353,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01111,
+      "step": 59,
+      "tokens/total": 1788816,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 26883
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.7271711826324463,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.01451511587947607,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01462,
+      "step": 60,
+      "tokens/total": 1819056,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 27323
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.14707089960575104,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.004126548767089844,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00414,
+      "step": 61,
+      "tokens/total": 1849664,
+      "tokens/train_per_sec_per_gpu": 39.34,
+      "tokens/trainable": 27830
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.4162050485610962,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.009982893243432045,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01003,
+      "step": 62,
+      "tokens/total": 1880272,
+      "tokens/train_per_sec_per_gpu": 32.46,
+      "tokens/trainable": 28281
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.563757598400116,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01410748716443777,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01421,
+      "step": 63,
+      "tokens/total": 1910752,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 28742
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.4908931255340576,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.011950287967920303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01202,
+      "step": 64,
+      "tokens/total": 1940848,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 29183
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3164263963699341,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006426030304282904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00645,
+      "step": 65,
+      "tokens/total": 1971344,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 29648
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4743383824825287,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.00986128207296133,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00991,
+      "step": 66,
+      "tokens/total": 2001712,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 30075
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.12894771993160248,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.002054001437500119,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00206,
+      "step": 67,
+      "tokens/total": 2029936,
+      "tokens/train_per_sec_per_gpu": 37.69,
+      "tokens/trainable": 30546
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.5559160709381104,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.007931358180940151,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00796,
+      "step": 68,
+      "tokens/total": 2060384,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 31025
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.4860860109329224,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.029434870928525925,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02987,
+      "step": 69,
+      "tokens/total": 2090688,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 31505
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.2197936624288559,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.0028239269740879536,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00283,
+      "step": 70,
+      "tokens/total": 2120848,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 31956
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.20881125330924988,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.001874964451417327,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00188,
+      "step": 71,
+      "tokens/total": 2149120,
+      "tokens/train_per_sec_per_gpu": 33.46,
+      "tokens/trainable": 32374
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.335940957069397,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.003750729141756892,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00376,
+      "step": 72,
+      "tokens/total": 2179472,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 32833
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.6529806852340698,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.010189465247094631,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01024,
+      "step": 73,
+      "tokens/total": 2209984,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 33304
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7839381098747253,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00870432797819376,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00874,
+      "step": 74,
+      "tokens/total": 2240512,
+      "tokens/train_per_sec_per_gpu": 32.54,
+      "tokens/trainable": 33771
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.3333408832550049,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007429605815559626,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00746,
+      "step": 75,
+      "tokens/total": 2270608,
+      "tokens/train_per_sec_per_gpu": 34.68,
+      "tokens/trainable": 34186
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.47243013978004456,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.009635023772716522,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00968,
+      "step": 76,
+      "tokens/total": 2300864,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 34667
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 1.4933921098709106,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.04087181016802788,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.04172,
+      "step": 77,
+      "tokens/total": 2330960,
+      "tokens/train_per_sec_per_gpu": 31.92,
+      "tokens/trainable": 35078
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6791361570358276,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.013059539720416069,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01315,
+      "step": 78,
+      "tokens/total": 2361584,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 35550
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.8410643935203552,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.02711997739970684,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02749,
+      "step": 79,
+      "tokens/total": 2391936,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 35977
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.3738754689693451,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.017121555283665657,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01727,
+      "step": 80,
+      "tokens/total": 2422432,
+      "tokens/train_per_sec_per_gpu": 38.39,
+      "tokens/trainable": 36462
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.0237257480621338,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.019660785794258118,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01986,
+      "step": 81,
+      "tokens/total": 2452752,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 36919
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.3411601185798645,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.011061297729611397,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01112,
+      "step": 82,
+      "tokens/total": 2481216,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 37379
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.41286832094192505,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011656982824206352,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01173,
+      "step": 83,
+      "tokens/total": 2511488,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 37832
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6960376501083374,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.024275368079543114,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02457,
+      "step": 84,
+      "tokens/total": 2541984,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 38303
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.15315213799476624,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.004536543972790241,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00455,
+      "step": 85,
+      "tokens/total": 2572224,
+      "tokens/train_per_sec_per_gpu": 33.67,
+      "tokens/trainable": 38758
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.18710190057754517,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.006560188252478838,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00658,
+      "step": 86,
+      "tokens/total": 2602704,
+      "tokens/train_per_sec_per_gpu": 30.7,
+      "tokens/trainable": 39179
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.343269407749176,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.010140963830053806,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01019,
+      "step": 87,
+      "tokens/total": 2633072,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 39679
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.3905015289783478,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.010611740872263908,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01067,
+      "step": 88,
+      "tokens/total": 2663232,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 40118
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.5544098019599915,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.022607026621699333,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02286,
+      "step": 89,
+      "tokens/total": 2693696,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40571
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.16370345652103424,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0030750958248972893,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00308,
+      "step": 90,
+      "tokens/total": 2723984,
+      "tokens/train_per_sec_per_gpu": 30.92,
+      "tokens/trainable": 40997
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.208849236369133,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.003701428882777691,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00371,
+      "step": 91,
+      "tokens/total": 2754240,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41462
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.5515285134315491,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.012641018256545067,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01272,
+      "step": 92,
+      "tokens/total": 2784672,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 41927
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.6742218136787415,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.019092712551355362,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01928,
+      "step": 93,
+      "tokens/total": 2815120,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 42397
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.3739137351512909,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.0018391611520200968,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00184,
+      "step": 94,
+      "tokens/total": 2845600,
+      "tokens/train_per_sec_per_gpu": 36.63,
+      "tokens/trainable": 42883
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25132012367248535,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.008805882185697556,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00884,
+      "step": 95,
+      "tokens/total": 2875952,
+      "tokens/train_per_sec_per_gpu": 38.78,
+      "tokens/trainable": 43381
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.24200227856636047,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.006166054867208004,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00619,
+      "step": 96,
+      "tokens/total": 2906288,
+      "tokens/train_per_sec_per_gpu": 33.06,
+      "tokens/trainable": 43824
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.3597196340560913,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0036215470172464848,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00363,
+      "step": 97,
+      "tokens/total": 2936752,
+      "tokens/train_per_sec_per_gpu": 31.62,
+      "tokens/trainable": 44255
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.7050648927688599,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.015324651263654232,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01544,
+      "step": 98,
+      "tokens/total": 2967072,
+      "tokens/train_per_sec_per_gpu": 37.7,
+      "tokens/trainable": 44738
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5716716051101685,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013399260118603706,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.01349,
+      "step": 99,
+      "tokens/total": 2997664,
+      "tokens/train_per_sec_per_gpu": 36.24,
+      "tokens/trainable": 45201
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.18090857565402985,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0020832926966249943,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00209,
+      "step": 100,
+      "tokens/total": 3028032,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 45645
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.5402258038520813,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011402487754821777,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01147,
+      "step": 101,
+      "tokens/total": 3058272,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 46088
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.220508873462677,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.026607759296894073,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02696,
+      "step": 102,
+      "tokens/total": 3088848,
+      "tokens/train_per_sec_per_gpu": 32.58,
+      "tokens/trainable": 46544
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.488730788230896,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.007838367484509945,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00787,
+      "step": 103,
+      "tokens/total": 3119424,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 47011
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 1.0092511177062988,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.024805881083011627,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02512,
+      "step": 104,
+      "tokens/total": 3149696,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 47500
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.1705712527036667,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002884519286453724,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00289,
+      "step": 105,
+      "tokens/total": 3180064,
+      "tokens/train_per_sec_per_gpu": 36.28,
+      "tokens/trainable": 47993
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.539692759513855,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.028211303055286407,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.02861,
+      "step": 106,
+      "tokens/total": 3208320,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 48444
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.576267421245575,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.026665428653359413,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02702,
+      "step": 107,
+      "tokens/total": 3238416,
+      "tokens/train_per_sec_per_gpu": 30.72,
+      "tokens/trainable": 48886
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.19187162816524506,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.007470840122550726,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0075,
+      "step": 108,
+      "tokens/total": 3268784,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 49357
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.30193838477134705,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.012815583497285843,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.0129,
+      "step": 109,
+      "tokens/total": 3298912,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 49824
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.24559386074543,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.010728488676249981,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01079,
+      "step": 110,
+      "tokens/total": 3329232,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 50264
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3296513259410858,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.01623804122209549,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01637,
+      "step": 111,
+      "tokens/total": 3359440,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 50708
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.4614265263080597,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.021365080028772354,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02159,
+      "step": 112,
+      "tokens/total": 3389472,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 51154
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.28496578335762024,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.013247976079583168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01334,
+      "step": 113,
+      "tokens/total": 3419920,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 51610
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.5088275074958801,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.022269586101174355,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02252,
+      "step": 114,
+      "tokens/total": 3450176,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 52014
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.19142404198646545,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006432589143514633,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00645,
+      "step": 115,
+      "tokens/total": 3480544,
+      "tokens/train_per_sec_per_gpu": 36.32,
+      "tokens/trainable": 52503
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 1.382026195526123,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.012845169752836227,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01293,
+      "step": 116,
+      "tokens/total": 3510848,
+      "tokens/train_per_sec_per_gpu": 38.75,
+      "tokens/trainable": 53005
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.16703951358795166,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0037160126958042383,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00372,
+      "step": 117,
+      "tokens/total": 3541296,
+      "tokens/train_per_sec_per_gpu": 35.4,
+      "tokens/trainable": 53464
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.615856945514679,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.014962448738515377,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01507,
+      "step": 118,
+      "tokens/total": 3571936,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 53920
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.24910427629947662,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.004819548688828945,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00483,
+      "step": 119,
+      "tokens/total": 3602000,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 54360
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.9213384389877319,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008400149643421173,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00844,
+      "step": 120,
+      "tokens/total": 3632288,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 54836
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.1228351816534996,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0019936836324632168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.002,
+      "step": 121,
+      "tokens/total": 3662608,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 55264
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.18029853701591492,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.004047749564051628,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00406,
+      "step": 122,
+      "tokens/total": 3692864,
+      "tokens/train_per_sec_per_gpu": 30.55,
+      "tokens/trainable": 55697
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4368670582771301,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00867566466331482,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00871,
+      "step": 123,
+      "tokens/total": 3722864,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 56141
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.3315965235233307,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011374368332326412,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01144,
+      "step": 124,
+      "tokens/total": 3753056,
+      "tokens/train_per_sec_per_gpu": 33.2,
+      "tokens/trainable": 56586
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2834460735321045,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.005234894342720509,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00525,
+      "step": 125,
+      "tokens/total": 3783312,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 57078
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5132001638412476,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.02469645068049431,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.025,
+      "step": 126,
+      "tokens/total": 3813712,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 57524
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.9969733953475952,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.040580250322818756,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.04141,
+      "step": 127,
+      "tokens/total": 3844336,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 58002
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.4804457724094391,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.007047833874821663,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00707,
+      "step": 128,
+      "tokens/total": 3872688,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 58451
+    }
+  ],
+  "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": 2.6286196555726387e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "gate_proj",
+    "q_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__coin0p2/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f18e56b80e7ad24c74c6f7090f4f740933e44ef72100c8a8432894e8c258c433
+size 547777976

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07b4ab83e765b9a9fef6d5ad5b0bbf7bcffcc6ee72a12963150438f05716abe2
+size 1048106435

aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-160/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62012b9e748bbef1beb75187c13d345f2d456800a4ec5207b243fb40c95ada94
+size 14645

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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__coin0p2/training/checkpoints/checkpoint-160/tokens_state.json

@@ -0,0 +1 @@
+{"total": 4836224, "trainable": 73058}

aft_wave_v2/coin_real_4x__coin0p2/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": 1.000339388847351,
+      "learning_rate": 0.0,
+      "loss": 0.11221420764923096,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.11875,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.52,
+      "tokens/trainable": 417
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8961698412895203,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11656901985406876,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.12364,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 873
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.5287365913391113,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11233991384506226,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11889,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 1353
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.3438372611999512,
+      "learning_rate": 1.2e-05,
+      "loss": 0.10132055729627609,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10663,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.26,
+      "tokens/trainable": 1778
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.0121135711669922,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.11346302926540375,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.12015,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 2261
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 3.9203455448150635,
+      "learning_rate": 2e-05,
+      "loss": 0.09780596196651459,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10275,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 2705
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.63088059425354,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06366116553544998,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06573,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.89,
+      "tokens/trainable": 3136
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.9995646476745605,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0759042352437973,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07886,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 3579
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0774554014205933,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04145222157239914,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04232,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 4000
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9648898839950562,
+      "learning_rate": 3.6e-05,
+      "loss": 0.032875221222639084,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03342,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 37.67,
+      "tokens/trainable": 4474
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 0.5937227606773376,
+      "learning_rate": 4e-05,
+      "loss": 0.023198626935482025,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02347,
+      "step": 11,
+      "tokens/total": 333296,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 4942
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.294079303741455,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.0510174036026001,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05234,
+      "step": 12,
+      "tokens/total": 363840,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 5424
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.7166640758514404,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06956956535577774,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07205,
+      "step": 13,
+      "tokens/total": 394112,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 5855
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.1496169567108154,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06250961124897003,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0645,
+      "step": 14,
+      "tokens/total": 424656,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 6344
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.515660047531128,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09882807731628418,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.10388,
+      "step": 15,
+      "tokens/total": 455152,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 6825
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.3756372928619385,
+      "learning_rate": 6e-05,
+      "loss": 0.0719379335641861,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.07459,
+      "step": 16,
+      "tokens/total": 485328,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 7230
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.7220206260681152,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.016590356826782227,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01673,
+      "step": 17,
+      "tokens/total": 515744,
+      "tokens/train_per_sec_per_gpu": 35.89,
+      "tokens/trainable": 7699
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.5261812210083008,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.030052796006202698,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.03051,
+      "step": 18,
+      "tokens/total": 546144,
+      "tokens/train_per_sec_per_gpu": 33.85,
+      "tokens/trainable": 8184
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.6416668891906738,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013492297381162643,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01358,
+      "step": 19,
+      "tokens/total": 576560,
+      "tokens/train_per_sec_per_gpu": 34.91,
+      "tokens/trainable": 8675
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.2296113967895508,
+      "learning_rate": 7.6e-05,
+      "loss": 0.038845278322696686,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03961,
+      "step": 20,
+      "tokens/total": 607056,
+      "tokens/train_per_sec_per_gpu": 39.72,
+      "tokens/trainable": 9189
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.8426793813705444,
+      "learning_rate": 8e-05,
+      "loss": 0.01696154847741127,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01711,
+      "step": 21,
+      "tokens/total": 637488,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 9614
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 10.310348510742188,
+      "learning_rate": 8.4e-05,
+      "loss": 0.06855659186840057,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.07096,
+      "step": 22,
+      "tokens/total": 667632,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10086
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.412604808807373,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.04118483513593674,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04204,
+      "step": 23,
+      "tokens/total": 697872,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 10535
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.4078032970428467,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.025748739019036293,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02608,
+      "step": 24,
+      "tokens/total": 728432,
+      "tokens/train_per_sec_per_gpu": 31.6,
+      "tokens/trainable": 10993
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 1.0874605178833008,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03088424727320671,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03137,
+      "step": 25,
+      "tokens/total": 756656,
+      "tokens/train_per_sec_per_gpu": 40.03,
+      "tokens/trainable": 11417
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.2574082612991333,
+      "learning_rate": 0.0001,
+      "loss": 0.03432866185903549,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03492,
+      "step": 26,
+      "tokens/total": 787120,
+      "tokens/train_per_sec_per_gpu": 34.36,
+      "tokens/trainable": 11874
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.30990996956825256,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.010750269517302513,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01081,
+      "step": 27,
+      "tokens/total": 817504,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 12316
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5124737620353699,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.01967313140630722,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01987,
+      "step": 28,
+      "tokens/total": 848128,
+      "tokens/train_per_sec_per_gpu": 37.86,
+      "tokens/trainable": 12800
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.2975701093673706,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.0434565395116806,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04441,
+      "step": 29,
+      "tokens/total": 878448,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 13270
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.4117317795753479,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.013137167319655418,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01322,
+      "step": 30,
+      "tokens/total": 909008,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 13752
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.5875954627990723,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.01785588450729847,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01802,
+      "step": 31,
+      "tokens/total": 939312,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 14218
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7800514698028564,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.011262159794569016,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01133,
+      "step": 32,
+      "tokens/total": 969264,
+      "tokens/train_per_sec_per_gpu": 34.49,
+      "tokens/trainable": 14655
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.0509369373321533,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04942459613084793,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.05067,
+      "step": 33,
+      "tokens/total": 999792,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 15087
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.6234491467475891,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.028202136978507042,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0286,
+      "step": 34,
+      "tokens/total": 1029840,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 15548
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6843816041946411,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.02795713022351265,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02835,
+      "step": 35,
+      "tokens/total": 1060416,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 16037
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.29489848017692566,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.010517537593841553,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01057,
+      "step": 36,
+      "tokens/total": 1090464,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16475
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.7335977554321289,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018923252820968628,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0191,
+      "step": 37,
+      "tokens/total": 1120944,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 16919
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.5587971210479736,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.020957889035344124,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02118,
+      "step": 38,
+      "tokens/total": 1151360,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17333
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 1.7112773656845093,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.045844268053770065,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.04691,
+      "step": 39,
+      "tokens/total": 1181728,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 17806
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.2373344898223877,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.04578991234302521,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04685,
+      "step": 40,
+      "tokens/total": 1212064,
+      "tokens/train_per_sec_per_gpu": 35.65,
+      "tokens/trainable": 18290
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.8950033187866211,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.02866373583674431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.02908,
+      "step": 41,
+      "tokens/total": 1242592,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 18747
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.9967094659805298,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02229372225701809,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02254,
+      "step": 42,
+      "tokens/total": 1272800,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 19172
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.1990034282207489,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.0059759169816970825,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00599,
+      "step": 43,
+      "tokens/total": 1303600,
+      "tokens/train_per_sec_per_gpu": 31.32,
+      "tokens/trainable": 19622
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.47151026129722595,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.011378830298781395,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01144,
+      "step": 44,
+      "tokens/total": 1333936,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 20073
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.5381685495376587,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.02141587622463703,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02165,
+      "step": 45,
+      "tokens/total": 1364288,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 20497
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 1.0174895524978638,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.02236388623714447,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02262,
+      "step": 46,
+      "tokens/total": 1394368,
+      "tokens/train_per_sec_per_gpu": 34.16,
+      "tokens/trainable": 20940
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.5081813931465149,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.010915009304881096,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.01097,
+      "step": 47,
+      "tokens/total": 1425024,
+      "tokens/train_per_sec_per_gpu": 35.04,
+      "tokens/trainable": 21405
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.4684470891952515,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.009690589271485806,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00974,
+      "step": 48,
+      "tokens/total": 1455392,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21828
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.49191179871559143,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.010902130044996738,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01096,
+      "step": 49,
+      "tokens/total": 1485616,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 22241
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.44692206382751465,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.006928037852048874,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00695,
+      "step": 50,
+      "tokens/total": 1515984,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 22688
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.37972721457481384,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.00806603953242302,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0081,
+      "step": 51,
+      "tokens/total": 1546224,
+      "tokens/train_per_sec_per_gpu": 32.12,
+      "tokens/trainable": 23136
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.0114665031433105,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013657575473189354,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01375,
+      "step": 52,
+      "tokens/total": 1576528,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 23631
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.6438844799995422,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05178550258278847,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05315,
+      "step": 53,
+      "tokens/total": 1606768,
+      "tokens/train_per_sec_per_gpu": 34.15,
+      "tokens/trainable": 24096
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 3.0008814334869385,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.010915370658040047,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01098,
+      "step": 54,
+      "tokens/total": 1636640,
+      "tokens/train_per_sec_per_gpu": 37.95,
+      "tokens/trainable": 24569
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.9377115368843079,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.019770421087741852,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01997,
+      "step": 55,
+      "tokens/total": 1667136,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 25067
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.23384763300418854,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.004394679330289364,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0044,
+      "step": 56,
+      "tokens/total": 1697472,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 25541
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.7558876276016235,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.02614775486290455,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02649,
+      "step": 57,
+      "tokens/total": 1727680,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 25965
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.4615328311920166,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01031234860420227,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01037,
+      "step": 58,
+      "tokens/total": 1758192,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 26409
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.40572404861450195,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.011052684858441353,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01111,
+      "step": 59,
+      "tokens/total": 1788816,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 26883
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.7271711826324463,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.01451511587947607,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01462,
+      "step": 60,
+      "tokens/total": 1819056,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 27323
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.14707089960575104,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.004126548767089844,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00414,
+      "step": 61,
+      "tokens/total": 1849664,
+      "tokens/train_per_sec_per_gpu": 39.34,
+      "tokens/trainable": 27830
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.4162050485610962,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.009982893243432045,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01003,
+      "step": 62,
+      "tokens/total": 1880272,
+      "tokens/train_per_sec_per_gpu": 32.46,
+      "tokens/trainable": 28281
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.563757598400116,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01410748716443777,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01421,
+      "step": 63,
+      "tokens/total": 1910752,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 28742
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.4908931255340576,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.011950287967920303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01202,
+      "step": 64,
+      "tokens/total": 1940848,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 29183
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3164263963699341,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006426030304282904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00645,
+      "step": 65,
+      "tokens/total": 1971344,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 29648
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4743383824825287,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.00986128207296133,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00991,
+      "step": 66,
+      "tokens/total": 2001712,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 30075
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.12894771993160248,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.002054001437500119,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00206,
+      "step": 67,
+      "tokens/total": 2029936,
+      "tokens/train_per_sec_per_gpu": 37.69,
+      "tokens/trainable": 30546
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.5559160709381104,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.007931358180940151,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00796,
+      "step": 68,
+      "tokens/total": 2060384,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 31025
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.4860860109329224,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.029434870928525925,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02987,
+      "step": 69,
+      "tokens/total": 2090688,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 31505
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.2197936624288559,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.0028239269740879536,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00283,
+      "step": 70,
+      "tokens/total": 2120848,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 31956
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.20881125330924988,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.001874964451417327,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00188,
+      "step": 71,
+      "tokens/total": 2149120,
+      "tokens/train_per_sec_per_gpu": 33.46,
+      "tokens/trainable": 32374
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.335940957069397,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.003750729141756892,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00376,
+      "step": 72,
+      "tokens/total": 2179472,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 32833
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.6529806852340698,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.010189465247094631,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01024,
+      "step": 73,
+      "tokens/total": 2209984,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 33304
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7839381098747253,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00870432797819376,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00874,
+      "step": 74,
+      "tokens/total": 2240512,
+      "tokens/train_per_sec_per_gpu": 32.54,
+      "tokens/trainable": 33771
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.3333408832550049,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007429605815559626,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00746,
+      "step": 75,
+      "tokens/total": 2270608,
+      "tokens/train_per_sec_per_gpu": 34.68,
+      "tokens/trainable": 34186
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.47243013978004456,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.009635023772716522,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00968,
+      "step": 76,
+      "tokens/total": 2300864,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 34667
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 1.4933921098709106,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.04087181016802788,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.04172,
+      "step": 77,
+      "tokens/total": 2330960,
+      "tokens/train_per_sec_per_gpu": 31.92,
+      "tokens/trainable": 35078
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6791361570358276,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.013059539720416069,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01315,
+      "step": 78,
+      "tokens/total": 2361584,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 35550
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.8410643935203552,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.02711997739970684,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02749,
+      "step": 79,
+      "tokens/total": 2391936,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 35977
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.3738754689693451,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.017121555283665657,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01727,
+      "step": 80,
+      "tokens/total": 2422432,
+      "tokens/train_per_sec_per_gpu": 38.39,
+      "tokens/trainable": 36462
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.0237257480621338,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.019660785794258118,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01986,
+      "step": 81,
+      "tokens/total": 2452752,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 36919
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.3411601185798645,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.011061297729611397,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01112,
+      "step": 82,
+      "tokens/total": 2481216,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 37379
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.41286832094192505,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011656982824206352,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01173,
+      "step": 83,
+      "tokens/total": 2511488,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 37832
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6960376501083374,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.024275368079543114,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02457,
+      "step": 84,
+      "tokens/total": 2541984,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 38303
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.15315213799476624,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.004536543972790241,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00455,
+      "step": 85,
+      "tokens/total": 2572224,
+      "tokens/train_per_sec_per_gpu": 33.67,
+      "tokens/trainable": 38758
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.18710190057754517,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.006560188252478838,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00658,
+      "step": 86,
+      "tokens/total": 2602704,
+      "tokens/train_per_sec_per_gpu": 30.7,
+      "tokens/trainable": 39179
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.343269407749176,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.010140963830053806,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01019,
+      "step": 87,
+      "tokens/total": 2633072,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 39679
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.3905015289783478,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.010611740872263908,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01067,
+      "step": 88,
+      "tokens/total": 2663232,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 40118
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.5544098019599915,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.022607026621699333,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02286,
+      "step": 89,
+      "tokens/total": 2693696,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40571
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.16370345652103424,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0030750958248972893,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00308,
+      "step": 90,
+      "tokens/total": 2723984,
+      "tokens/train_per_sec_per_gpu": 30.92,
+      "tokens/trainable": 40997
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.208849236369133,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.003701428882777691,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00371,
+      "step": 91,
+      "tokens/total": 2754240,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41462
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.5515285134315491,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.012641018256545067,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01272,
+      "step": 92,
+      "tokens/total": 2784672,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 41927
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.6742218136787415,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.019092712551355362,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01928,
+      "step": 93,
+      "tokens/total": 2815120,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 42397
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.3739137351512909,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.0018391611520200968,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00184,
+      "step": 94,
+      "tokens/total": 2845600,
+      "tokens/train_per_sec_per_gpu": 36.63,
+      "tokens/trainable": 42883
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25132012367248535,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.008805882185697556,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00884,
+      "step": 95,
+      "tokens/total": 2875952,
+      "tokens/train_per_sec_per_gpu": 38.78,
+      "tokens/trainable": 43381
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.24200227856636047,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.006166054867208004,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00619,
+      "step": 96,
+      "tokens/total": 2906288,
+      "tokens/train_per_sec_per_gpu": 33.06,
+      "tokens/trainable": 43824
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.3597196340560913,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0036215470172464848,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00363,
+      "step": 97,
+      "tokens/total": 2936752,
+      "tokens/train_per_sec_per_gpu": 31.62,
+      "tokens/trainable": 44255
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.7050648927688599,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.015324651263654232,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01544,
+      "step": 98,
+      "tokens/total": 2967072,
+      "tokens/train_per_sec_per_gpu": 37.7,
+      "tokens/trainable": 44738
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5716716051101685,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013399260118603706,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.01349,
+      "step": 99,
+      "tokens/total": 2997664,
+      "tokens/train_per_sec_per_gpu": 36.24,
+      "tokens/trainable": 45201
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.18090857565402985,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0020832926966249943,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00209,
+      "step": 100,
+      "tokens/total": 3028032,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 45645
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.5402258038520813,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011402487754821777,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01147,
+      "step": 101,
+      "tokens/total": 3058272,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 46088
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.220508873462677,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.026607759296894073,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02696,
+      "step": 102,
+      "tokens/total": 3088848,
+      "tokens/train_per_sec_per_gpu": 32.58,
+      "tokens/trainable": 46544
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.488730788230896,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.007838367484509945,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00787,
+      "step": 103,
+      "tokens/total": 3119424,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 47011
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 1.0092511177062988,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.024805881083011627,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02512,
+      "step": 104,
+      "tokens/total": 3149696,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 47500
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.1705712527036667,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002884519286453724,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00289,
+      "step": 105,
+      "tokens/total": 3180064,
+      "tokens/train_per_sec_per_gpu": 36.28,
+      "tokens/trainable": 47993
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.539692759513855,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.028211303055286407,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.02861,
+      "step": 106,
+      "tokens/total": 3208320,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 48444
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.576267421245575,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.026665428653359413,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02702,
+      "step": 107,
+      "tokens/total": 3238416,
+      "tokens/train_per_sec_per_gpu": 30.72,
+      "tokens/trainable": 48886
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.19187162816524506,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.007470840122550726,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0075,
+      "step": 108,
+      "tokens/total": 3268784,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 49357
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.30193838477134705,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.012815583497285843,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.0129,
+      "step": 109,
+      "tokens/total": 3298912,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 49824
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.24559386074543,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.010728488676249981,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01079,
+      "step": 110,
+      "tokens/total": 3329232,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 50264
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3296513259410858,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.01623804122209549,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01637,
+      "step": 111,
+      "tokens/total": 3359440,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 50708
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.4614265263080597,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.021365080028772354,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02159,
+      "step": 112,
+      "tokens/total": 3389472,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 51154
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.28496578335762024,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.013247976079583168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01334,
+      "step": 113,
+      "tokens/total": 3419920,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 51610
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.5088275074958801,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.022269586101174355,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02252,
+      "step": 114,
+      "tokens/total": 3450176,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 52014
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.19142404198646545,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006432589143514633,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00645,
+      "step": 115,
+      "tokens/total": 3480544,
+      "tokens/train_per_sec_per_gpu": 36.32,
+      "tokens/trainable": 52503
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 1.382026195526123,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.012845169752836227,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01293,
+      "step": 116,
+      "tokens/total": 3510848,
+      "tokens/train_per_sec_per_gpu": 38.75,
+      "tokens/trainable": 53005
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.16703951358795166,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0037160126958042383,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00372,
+      "step": 117,
+      "tokens/total": 3541296,
+      "tokens/train_per_sec_per_gpu": 35.4,
+      "tokens/trainable": 53464
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.615856945514679,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.014962448738515377,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01507,
+      "step": 118,
+      "tokens/total": 3571936,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 53920
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.24910427629947662,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.004819548688828945,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00483,
+      "step": 119,
+      "tokens/total": 3602000,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 54360
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.9213384389877319,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008400149643421173,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00844,
+      "step": 120,
+      "tokens/total": 3632288,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 54836
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.1228351816534996,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0019936836324632168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.002,
+      "step": 121,
+      "tokens/total": 3662608,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 55264
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.18029853701591492,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.004047749564051628,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00406,
+      "step": 122,
+      "tokens/total": 3692864,
+      "tokens/train_per_sec_per_gpu": 30.55,
+      "tokens/trainable": 55697
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4368670582771301,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00867566466331482,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00871,
+      "step": 123,
+      "tokens/total": 3722864,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 56141
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.3315965235233307,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011374368332326412,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01144,
+      "step": 124,
+      "tokens/total": 3753056,
+      "tokens/train_per_sec_per_gpu": 33.2,
+      "tokens/trainable": 56586
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2834460735321045,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.005234894342720509,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00525,
+      "step": 125,
+      "tokens/total": 3783312,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 57078
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5132001638412476,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.02469645068049431,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.025,
+      "step": 126,
+      "tokens/total": 3813712,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 57524
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.9969733953475952,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.040580250322818756,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.04141,
+      "step": 127,
+      "tokens/total": 3844336,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 58002
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.4804457724094391,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.007047833874821663,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00707,
+      "step": 128,
+      "tokens/total": 3872688,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 58451
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.23698025941848755,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.00366212404333055,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.00367,
+      "step": 129,
+      "tokens/total": 3902640,
+      "tokens/train_per_sec_per_gpu": 32.1,
+      "tokens/trainable": 58893
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.44587212800979614,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.010764295235276222,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01082,
+      "step": 130,
+      "tokens/total": 3933104,
+      "tokens/train_per_sec_per_gpu": 35.94,
+      "tokens/trainable": 59353
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.3212004005908966,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.013213565573096275,
+      "memory/device_reserved (GiB)": 34.92,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0133,
+      "step": 131,
+      "tokens/total": 3963184,
+      "tokens/train_per_sec_per_gpu": 32.29,
+      "tokens/trainable": 59789
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.14710628986358643,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005224708467721939,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00524,
+      "step": 132,
+      "tokens/total": 3993648,
+      "tokens/train_per_sec_per_gpu": 35.82,
+      "tokens/trainable": 60263
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.5616672039031982,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.01082855835556984,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01089,
+      "step": 133,
+      "tokens/total": 4024000,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 60756
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.28706833720207214,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.02170894667506218,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.02195,
+      "step": 134,
+      "tokens/total": 4052480,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 61201
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.1448316127061844,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.0064529795199632645,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00647,
+      "step": 135,
+      "tokens/total": 4082864,
+      "tokens/train_per_sec_per_gpu": 36.07,
+      "tokens/trainable": 61681
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.2812711000442505,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.01243551354855299,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01251,
+      "step": 136,
+      "tokens/total": 4113040,
+      "tokens/train_per_sec_per_gpu": 38.33,
+      "tokens/trainable": 62156
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.37281039357185364,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.009842045605182648,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00989,
+      "step": 137,
+      "tokens/total": 4143344,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 62646
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.15259796380996704,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.007893089205026627,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00792,
+      "step": 138,
+      "tokens/total": 4173648,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 63097
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.2609741985797882,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.009887185879051685,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00994,
+      "step": 139,
+      "tokens/total": 4204080,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 63585
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.3938979506492615,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.020112136378884315,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.02032,
+      "step": 140,
+      "tokens/total": 4232400,
+      "tokens/train_per_sec_per_gpu": 33.14,
+      "tokens/trainable": 64002
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.2883419096469879,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.020921282470226288,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02114,
+      "step": 141,
+      "tokens/total": 4262624,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 64459
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.22204594314098358,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.007596657145768404,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00763,
+      "step": 142,
+      "tokens/total": 4293168,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 64897
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.3463776409626007,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.008941063657402992,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00898,
+      "step": 143,
+      "tokens/total": 4323168,
+      "tokens/train_per_sec_per_gpu": 37.81,
+      "tokens/trainable": 65376
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.07992643117904663,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.002066058572381735,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00207,
+      "step": 144,
+      "tokens/total": 4353264,
+      "tokens/train_per_sec_per_gpu": 36.81,
+      "tokens/trainable": 65842
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.4542894959449768,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.006295363884419203,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00632,
+      "step": 145,
+      "tokens/total": 4383504,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 66282
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.6165645122528076,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.021687351167201996,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02192,
+      "step": 146,
+      "tokens/total": 4413776,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 66733
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.44622310996055603,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.010542265139520168,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0106,
+      "step": 147,
+      "tokens/total": 4444032,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 67147
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.5972599387168884,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.015109038911759853,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01522,
+      "step": 148,
+      "tokens/total": 4474384,
+      "tokens/train_per_sec_per_gpu": 35.19,
+      "tokens/trainable": 67630
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.16330066323280334,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.003038810333237052,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00304,
+      "step": 149,
+      "tokens/total": 4504800,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 68074
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.1966242492198944,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.00474123377352953,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00475,
+      "step": 150,
+      "tokens/total": 4534896,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 68542
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.2519090175628662,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.005299043841660023,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00531,
+      "step": 151,
+      "tokens/total": 4565008,
+      "tokens/train_per_sec_per_gpu": 30.34,
+      "tokens/trainable": 69008
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.45198729634284973,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.00804843008518219,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00808,
+      "step": 152,
+      "tokens/total": 4595520,
+      "tokens/train_per_sec_per_gpu": 35.52,
+      "tokens/trainable": 69462
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 1.9722840785980225,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.015857091173529625,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01598,
+      "step": 153,
+      "tokens/total": 4625952,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 69925
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.1563708782196045,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.003014163114130497,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00302,
+      "step": 154,
+      "tokens/total": 4656272,
+      "tokens/train_per_sec_per_gpu": 36.73,
+      "tokens/trainable": 70392
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3952409327030182,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.020546773448586464,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02076,
+      "step": 155,
+      "tokens/total": 4684592,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 70826
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.29893431067466736,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.006672243122011423,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00669,
+      "step": 156,
+      "tokens/total": 4715136,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 71284
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.12646262347698212,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.001321017974987626,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00132,
+      "step": 157,
+      "tokens/total": 4745296,
+      "tokens/train_per_sec_per_gpu": 35.12,
+      "tokens/trainable": 71746
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.3663196563720703,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.00867768656462431,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00872,
+      "step": 158,
+      "tokens/total": 4775696,
+      "tokens/train_per_sec_per_gpu": 32.33,
+      "tokens/trainable": 72182
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.32793474197387695,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.012621743604540825,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.0127,
+      "step": 159,
+      "tokens/total": 4805776,
+      "tokens/train_per_sec_per_gpu": 30.69,
+      "tokens/trainable": 72601
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.502180278301239,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.02071015164256096,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02093,
+      "step": 160,
+      "tokens/total": 4836224,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 73058
+    }
+  ],
+  "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": 3.2826278453498266e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "gate_proj",
+    "q_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__coin0p2/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85ed8480f6ee209a38ee9cd571b15a1b6c5ce56607a32188e14d48492e058adf
+size 547777976

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a86ff874a8c3cc41a595be2ed4adada5891f02ffa60010a45ced596f3483975a
+size 1048106435

aft_wave_v2/coin_real_4x__coin0p2/training/checkpoints/checkpoint-192/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d69b2462378ed67f5f5555fb6b192965044d7652528913ba5ccc13c8e15172a
+size 14645

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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__coin0p2/training/checkpoints/checkpoint-192/tokens_state.json

@@ -0,0 +1 @@
+{"total": 5807168, "trainable": 87734}

aft_wave_v2/coin_real_4x__coin0p2/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": 1.000339388847351,
+      "learning_rate": 0.0,
+      "loss": 0.11221420764923096,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.11875,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 27.52,
+      "tokens/trainable": 417
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8961698412895203,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.11656901985406876,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.12364,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 873
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.5287365913391113,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.11233991384506226,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.11889,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 1353
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.3438372611999512,
+      "learning_rate": 1.2e-05,
+      "loss": 0.10132055729627609,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.10663,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.26,
+      "tokens/trainable": 1778
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.0121135711669922,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.11346302926540375,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.12015,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 2261
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 3.9203455448150635,
+      "learning_rate": 2e-05,
+      "loss": 0.09780596196651459,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.10275,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 2705
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.63088059425354,
+      "learning_rate": 2.4e-05,
+      "loss": 0.06366116553544998,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06573,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 33.89,
+      "tokens/trainable": 3136
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.9995646476745605,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0759042352437973,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07886,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 3579
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0774554014205933,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04145222157239914,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04232,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 4000
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9648898839950562,
+      "learning_rate": 3.6e-05,
+      "loss": 0.032875221222639084,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03342,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 37.67,
+      "tokens/trainable": 4474
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 0.5937227606773376,
+      "learning_rate": 4e-05,
+      "loss": 0.023198626935482025,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02347,
+      "step": 11,
+      "tokens/total": 333296,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 4942
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.294079303741455,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.0510174036026001,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.05234,
+      "step": 12,
+      "tokens/total": 363840,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 5424
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.7166640758514404,
+      "learning_rate": 4.8e-05,
+      "loss": 0.06956956535577774,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.07205,
+      "step": 13,
+      "tokens/total": 394112,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 5855
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.1496169567108154,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06250961124897003,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0645,
+      "step": 14,
+      "tokens/total": 424656,
+      "tokens/train_per_sec_per_gpu": 34.78,
+      "tokens/trainable": 6344
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.515660047531128,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09882807731628418,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.10388,
+      "step": 15,
+      "tokens/total": 455152,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 6825
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.3756372928619385,
+      "learning_rate": 6e-05,
+      "loss": 0.0719379335641861,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.07459,
+      "step": 16,
+      "tokens/total": 485328,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 7230
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.7220206260681152,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.016590356826782227,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01673,
+      "step": 17,
+      "tokens/total": 515744,
+      "tokens/train_per_sec_per_gpu": 35.89,
+      "tokens/trainable": 7699
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.5261812210083008,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.030052796006202698,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.03051,
+      "step": 18,
+      "tokens/total": 546144,
+      "tokens/train_per_sec_per_gpu": 33.85,
+      "tokens/trainable": 8184
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.6416668891906738,
+      "learning_rate": 7.2e-05,
+      "loss": 0.013492297381162643,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01358,
+      "step": 19,
+      "tokens/total": 576560,
+      "tokens/train_per_sec_per_gpu": 34.91,
+      "tokens/trainable": 8675
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.2296113967895508,
+      "learning_rate": 7.6e-05,
+      "loss": 0.038845278322696686,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03961,
+      "step": 20,
+      "tokens/total": 607056,
+      "tokens/train_per_sec_per_gpu": 39.72,
+      "tokens/trainable": 9189
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.8426793813705444,
+      "learning_rate": 8e-05,
+      "loss": 0.01696154847741127,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01711,
+      "step": 21,
+      "tokens/total": 637488,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 9614
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 10.310348510742188,
+      "learning_rate": 8.4e-05,
+      "loss": 0.06855659186840057,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.07096,
+      "step": 22,
+      "tokens/total": 667632,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 10086
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.412604808807373,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.04118483513593674,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04204,
+      "step": 23,
+      "tokens/total": 697872,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 10535
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 1.4078032970428467,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.025748739019036293,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02608,
+      "step": 24,
+      "tokens/total": 728432,
+      "tokens/train_per_sec_per_gpu": 31.6,
+      "tokens/trainable": 10993
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 1.0874605178833008,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03088424727320671,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03137,
+      "step": 25,
+      "tokens/total": 756656,
+      "tokens/train_per_sec_per_gpu": 40.03,
+      "tokens/trainable": 11417
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.2574082612991333,
+      "learning_rate": 0.0001,
+      "loss": 0.03432866185903549,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03492,
+      "step": 26,
+      "tokens/total": 787120,
+      "tokens/train_per_sec_per_gpu": 34.36,
+      "tokens/trainable": 11874
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.30990996956825256,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.010750269517302513,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01081,
+      "step": 27,
+      "tokens/total": 817504,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 12316
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.5124737620353699,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.01967313140630722,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01987,
+      "step": 28,
+      "tokens/total": 848128,
+      "tokens/train_per_sec_per_gpu": 37.86,
+      "tokens/trainable": 12800
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.2975701093673706,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.0434565395116806,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04441,
+      "step": 29,
+      "tokens/total": 878448,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 13270
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.4117317795753479,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.013137167319655418,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01322,
+      "step": 30,
+      "tokens/total": 909008,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 13752
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.5875954627990723,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.01785588450729847,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01802,
+      "step": 31,
+      "tokens/total": 939312,
+      "tokens/train_per_sec_per_gpu": 34.07,
+      "tokens/trainable": 14218
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7800514698028564,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.011262159794569016,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01133,
+      "step": 32,
+      "tokens/total": 969264,
+      "tokens/train_per_sec_per_gpu": 34.49,
+      "tokens/trainable": 14655
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.0509369373321533,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04942459613084793,
+      "memory/device_reserved (GiB)": 37.82,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.05067,
+      "step": 33,
+      "tokens/total": 999792,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 15087
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.6234491467475891,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.028202136978507042,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0286,
+      "step": 34,
+      "tokens/total": 1029840,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 15548
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6843816041946411,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.02795713022351265,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02835,
+      "step": 35,
+      "tokens/total": 1060416,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 16037
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.29489848017692566,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.010517537593841553,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01057,
+      "step": 36,
+      "tokens/total": 1090464,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 16475
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.7335977554321289,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.018923252820968628,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0191,
+      "step": 37,
+      "tokens/total": 1120944,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 16919
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.5587971210479736,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.020957889035344124,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02118,
+      "step": 38,
+      "tokens/total": 1151360,
+      "tokens/train_per_sec_per_gpu": 28.97,
+      "tokens/trainable": 17333
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 1.7112773656845093,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.045844268053770065,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.04691,
+      "step": 39,
+      "tokens/total": 1181728,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 17806
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.2373344898223877,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.04578991234302521,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04685,
+      "step": 40,
+      "tokens/total": 1212064,
+      "tokens/train_per_sec_per_gpu": 35.65,
+      "tokens/trainable": 18290
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.8950033187866211,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.02866373583674431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.02908,
+      "step": 41,
+      "tokens/total": 1242592,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 18747
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.9967094659805298,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02229372225701809,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02254,
+      "step": 42,
+      "tokens/total": 1272800,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 19172
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.1990034282207489,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.0059759169816970825,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00599,
+      "step": 43,
+      "tokens/total": 1303600,
+      "tokens/train_per_sec_per_gpu": 31.32,
+      "tokens/trainable": 19622
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.47151026129722595,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.011378830298781395,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01144,
+      "step": 44,
+      "tokens/total": 1333936,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 20073
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.5381685495376587,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.02141587622463703,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02165,
+      "step": 45,
+      "tokens/total": 1364288,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 20497
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 1.0174895524978638,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.02236388623714447,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02262,
+      "step": 46,
+      "tokens/total": 1394368,
+      "tokens/train_per_sec_per_gpu": 34.16,
+      "tokens/trainable": 20940
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.5081813931465149,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.010915009304881096,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.01097,
+      "step": 47,
+      "tokens/total": 1425024,
+      "tokens/train_per_sec_per_gpu": 35.04,
+      "tokens/trainable": 21405
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.4684470891952515,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.009690589271485806,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00974,
+      "step": 48,
+      "tokens/total": 1455392,
+      "tokens/train_per_sec_per_gpu": 35.13,
+      "tokens/trainable": 21828
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.49191179871559143,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.010902130044996738,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01096,
+      "step": 49,
+      "tokens/total": 1485616,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 22241
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.44692206382751465,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.006928037852048874,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00695,
+      "step": 50,
+      "tokens/total": 1515984,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 22688
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.37972721457481384,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.00806603953242302,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0081,
+      "step": 51,
+      "tokens/total": 1546224,
+      "tokens/train_per_sec_per_gpu": 32.12,
+      "tokens/trainable": 23136
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.0114665031433105,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013657575473189354,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01375,
+      "step": 52,
+      "tokens/total": 1576528,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 23631
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 0.6438844799995422,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05178550258278847,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.05315,
+      "step": 53,
+      "tokens/total": 1606768,
+      "tokens/train_per_sec_per_gpu": 34.15,
+      "tokens/trainable": 24096
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 3.0008814334869385,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.010915370658040047,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01098,
+      "step": 54,
+      "tokens/total": 1636640,
+      "tokens/train_per_sec_per_gpu": 37.95,
+      "tokens/trainable": 24569
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.9377115368843079,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.019770421087741852,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01997,
+      "step": 55,
+      "tokens/total": 1667136,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 25067
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.23384763300418854,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.004394679330289364,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0044,
+      "step": 56,
+      "tokens/total": 1697472,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 25541
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.7558876276016235,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.02614775486290455,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02649,
+      "step": 57,
+      "tokens/total": 1727680,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 25965
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.4615328311920166,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01031234860420227,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01037,
+      "step": 58,
+      "tokens/total": 1758192,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 26409
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.40572404861450195,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.011052684858441353,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01111,
+      "step": 59,
+      "tokens/total": 1788816,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 26883
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.7271711826324463,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.01451511587947607,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01462,
+      "step": 60,
+      "tokens/total": 1819056,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 27323
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.14707089960575104,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.004126548767089844,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00414,
+      "step": 61,
+      "tokens/total": 1849664,
+      "tokens/train_per_sec_per_gpu": 39.34,
+      "tokens/trainable": 27830
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.4162050485610962,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.009982893243432045,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.01003,
+      "step": 62,
+      "tokens/total": 1880272,
+      "tokens/train_per_sec_per_gpu": 32.46,
+      "tokens/trainable": 28281
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.563757598400116,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01410748716443777,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01421,
+      "step": 63,
+      "tokens/total": 1910752,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 28742
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.4908931255340576,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.011950287967920303,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01202,
+      "step": 64,
+      "tokens/total": 1940848,
+      "tokens/train_per_sec_per_gpu": 33.76,
+      "tokens/trainable": 29183
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3164263963699341,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006426030304282904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00645,
+      "step": 65,
+      "tokens/total": 1971344,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 29648
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4743383824825287,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.00986128207296133,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00991,
+      "step": 66,
+      "tokens/total": 2001712,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 30075
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.12894771993160248,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.002054001437500119,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00206,
+      "step": 67,
+      "tokens/total": 2029936,
+      "tokens/train_per_sec_per_gpu": 37.69,
+      "tokens/trainable": 30546
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.5559160709381104,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.007931358180940151,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00796,
+      "step": 68,
+      "tokens/total": 2060384,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 31025
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.4860860109329224,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.029434870928525925,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02987,
+      "step": 69,
+      "tokens/total": 2090688,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 31505
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.2197936624288559,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.0028239269740879536,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00283,
+      "step": 70,
+      "tokens/total": 2120848,
+      "tokens/train_per_sec_per_gpu": 32.39,
+      "tokens/trainable": 31956
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.20881125330924988,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.001874964451417327,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00188,
+      "step": 71,
+      "tokens/total": 2149120,
+      "tokens/train_per_sec_per_gpu": 33.46,
+      "tokens/trainable": 32374
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.335940957069397,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.003750729141756892,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00376,
+      "step": 72,
+      "tokens/total": 2179472,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 32833
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.6529806852340698,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.010189465247094631,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01024,
+      "step": 73,
+      "tokens/total": 2209984,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 33304
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.7839381098747253,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00870432797819376,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00874,
+      "step": 74,
+      "tokens/total": 2240512,
+      "tokens/train_per_sec_per_gpu": 32.54,
+      "tokens/trainable": 33771
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.3333408832550049,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.007429605815559626,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00746,
+      "step": 75,
+      "tokens/total": 2270608,
+      "tokens/train_per_sec_per_gpu": 34.68,
+      "tokens/trainable": 34186
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.47243013978004456,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.009635023772716522,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00968,
+      "step": 76,
+      "tokens/total": 2300864,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 34667
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 1.4933921098709106,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.04087181016802788,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.04172,
+      "step": 77,
+      "tokens/total": 2330960,
+      "tokens/train_per_sec_per_gpu": 31.92,
+      "tokens/trainable": 35078
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6791361570358276,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.013059539720416069,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01315,
+      "step": 78,
+      "tokens/total": 2361584,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 35550
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.8410643935203552,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.02711997739970684,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02749,
+      "step": 79,
+      "tokens/total": 2391936,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 35977
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.3738754689693451,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.017121555283665657,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01727,
+      "step": 80,
+      "tokens/total": 2422432,
+      "tokens/train_per_sec_per_gpu": 38.39,
+      "tokens/trainable": 36462
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.0237257480621338,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.019660785794258118,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01986,
+      "step": 81,
+      "tokens/total": 2452752,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 36919
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.3411601185798645,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.011061297729611397,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01112,
+      "step": 82,
+      "tokens/total": 2481216,
+      "tokens/train_per_sec_per_gpu": 33.56,
+      "tokens/trainable": 37379
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.41286832094192505,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011656982824206352,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01173,
+      "step": 83,
+      "tokens/total": 2511488,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 37832
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.6960376501083374,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.024275368079543114,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02457,
+      "step": 84,
+      "tokens/total": 2541984,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 38303
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.15315213799476624,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.004536543972790241,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00455,
+      "step": 85,
+      "tokens/total": 2572224,
+      "tokens/train_per_sec_per_gpu": 33.67,
+      "tokens/trainable": 38758
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.18710190057754517,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.006560188252478838,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00658,
+      "step": 86,
+      "tokens/total": 2602704,
+      "tokens/train_per_sec_per_gpu": 30.7,
+      "tokens/trainable": 39179
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.343269407749176,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.010140963830053806,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01019,
+      "step": 87,
+      "tokens/total": 2633072,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 39679
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.3905015289783478,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.010611740872263908,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01067,
+      "step": 88,
+      "tokens/total": 2663232,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 40118
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.5544098019599915,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.022607026621699333,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02286,
+      "step": 89,
+      "tokens/total": 2693696,
+      "tokens/train_per_sec_per_gpu": 32.92,
+      "tokens/trainable": 40571
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.16370345652103424,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0030750958248972893,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00308,
+      "step": 90,
+      "tokens/total": 2723984,
+      "tokens/train_per_sec_per_gpu": 30.92,
+      "tokens/trainable": 40997
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.208849236369133,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.003701428882777691,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00371,
+      "step": 91,
+      "tokens/total": 2754240,
+      "tokens/train_per_sec_per_gpu": 36.55,
+      "tokens/trainable": 41462
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.5515285134315491,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.012641018256545067,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01272,
+      "step": 92,
+      "tokens/total": 2784672,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 41927
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.6742218136787415,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.019092712551355362,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01928,
+      "step": 93,
+      "tokens/total": 2815120,
+      "tokens/train_per_sec_per_gpu": 34.95,
+      "tokens/trainable": 42397
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.3739137351512909,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.0018391611520200968,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00184,
+      "step": 94,
+      "tokens/total": 2845600,
+      "tokens/train_per_sec_per_gpu": 36.63,
+      "tokens/trainable": 42883
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25132012367248535,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.008805882185697556,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00884,
+      "step": 95,
+      "tokens/total": 2875952,
+      "tokens/train_per_sec_per_gpu": 38.78,
+      "tokens/trainable": 43381
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.24200227856636047,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.006166054867208004,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00619,
+      "step": 96,
+      "tokens/total": 2906288,
+      "tokens/train_per_sec_per_gpu": 33.06,
+      "tokens/trainable": 43824
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.3597196340560913,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0036215470172464848,
+      "memory/device_reserved (GiB)": 35.98,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00363,
+      "step": 97,
+      "tokens/total": 2936752,
+      "tokens/train_per_sec_per_gpu": 31.62,
+      "tokens/trainable": 44255
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.7050648927688599,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.015324651263654232,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01544,
+      "step": 98,
+      "tokens/total": 2967072,
+      "tokens/train_per_sec_per_gpu": 37.7,
+      "tokens/trainable": 44738
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5716716051101685,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013399260118603706,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.01349,
+      "step": 99,
+      "tokens/total": 2997664,
+      "tokens/train_per_sec_per_gpu": 36.24,
+      "tokens/trainable": 45201
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.18090857565402985,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0020832926966249943,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00209,
+      "step": 100,
+      "tokens/total": 3028032,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 45645
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.5402258038520813,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011402487754821777,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01147,
+      "step": 101,
+      "tokens/total": 3058272,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 46088
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.220508873462677,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.026607759296894073,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02696,
+      "step": 102,
+      "tokens/total": 3088848,
+      "tokens/train_per_sec_per_gpu": 32.58,
+      "tokens/trainable": 46544
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.488730788230896,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.007838367484509945,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00787,
+      "step": 103,
+      "tokens/total": 3119424,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 47011
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 1.0092511177062988,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.024805881083011627,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02512,
+      "step": 104,
+      "tokens/total": 3149696,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 47500
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.1705712527036667,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.002884519286453724,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00289,
+      "step": 105,
+      "tokens/total": 3180064,
+      "tokens/train_per_sec_per_gpu": 36.28,
+      "tokens/trainable": 47993
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.539692759513855,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.028211303055286407,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.02861,
+      "step": 106,
+      "tokens/total": 3208320,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 48444
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.576267421245575,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.026665428653359413,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02702,
+      "step": 107,
+      "tokens/total": 3238416,
+      "tokens/train_per_sec_per_gpu": 30.72,
+      "tokens/trainable": 48886
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.19187162816524506,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.007470840122550726,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0075,
+      "step": 108,
+      "tokens/total": 3268784,
+      "tokens/train_per_sec_per_gpu": 37.78,
+      "tokens/trainable": 49357
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.30193838477134705,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.012815583497285843,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.0129,
+      "step": 109,
+      "tokens/total": 3298912,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 49824
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.24559386074543,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.010728488676249981,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.01079,
+      "step": 110,
+      "tokens/total": 3329232,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 50264
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3296513259410858,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.01623804122209549,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01637,
+      "step": 111,
+      "tokens/total": 3359440,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 50708
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.4614265263080597,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.021365080028772354,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02159,
+      "step": 112,
+      "tokens/total": 3389472,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 51154
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.28496578335762024,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.013247976079583168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01334,
+      "step": 113,
+      "tokens/total": 3419920,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 51610
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.5088275074958801,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.022269586101174355,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02252,
+      "step": 114,
+      "tokens/total": 3450176,
+      "tokens/train_per_sec_per_gpu": 29.14,
+      "tokens/trainable": 52014
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.19142404198646545,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006432589143514633,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00645,
+      "step": 115,
+      "tokens/total": 3480544,
+      "tokens/train_per_sec_per_gpu": 36.32,
+      "tokens/trainable": 52503
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 1.382026195526123,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.012845169752836227,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01293,
+      "step": 116,
+      "tokens/total": 3510848,
+      "tokens/train_per_sec_per_gpu": 38.75,
+      "tokens/trainable": 53005
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.16703951358795166,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0037160126958042383,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00372,
+      "step": 117,
+      "tokens/total": 3541296,
+      "tokens/train_per_sec_per_gpu": 35.4,
+      "tokens/trainable": 53464
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.615856945514679,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.014962448738515377,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01507,
+      "step": 118,
+      "tokens/total": 3571936,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 53920
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.24910427629947662,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.004819548688828945,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00483,
+      "step": 119,
+      "tokens/total": 3602000,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 54360
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.9213384389877319,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008400149643421173,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00844,
+      "step": 120,
+      "tokens/total": 3632288,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 54836
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.1228351816534996,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0019936836324632168,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.002,
+      "step": 121,
+      "tokens/total": 3662608,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 55264
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.18029853701591492,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.004047749564051628,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00406,
+      "step": 122,
+      "tokens/total": 3692864,
+      "tokens/train_per_sec_per_gpu": 30.55,
+      "tokens/trainable": 55697
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4368670582771301,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00867566466331482,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00871,
+      "step": 123,
+      "tokens/total": 3722864,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 56141
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.3315965235233307,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011374368332326412,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01144,
+      "step": 124,
+      "tokens/total": 3753056,
+      "tokens/train_per_sec_per_gpu": 33.2,
+      "tokens/trainable": 56586
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.2834460735321045,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.005234894342720509,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00525,
+      "step": 125,
+      "tokens/total": 3783312,
+      "tokens/train_per_sec_per_gpu": 36.68,
+      "tokens/trainable": 57078
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5132001638412476,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.02469645068049431,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.025,
+      "step": 126,
+      "tokens/total": 3813712,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 57524
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.9969733953475952,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.040580250322818756,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.04141,
+      "step": 127,
+      "tokens/total": 3844336,
+      "tokens/train_per_sec_per_gpu": 34.85,
+      "tokens/trainable": 58002
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.4804457724094391,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.007047833874821663,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00707,
+      "step": 128,
+      "tokens/total": 3872688,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 58451
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.23698025941848755,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.00366212404333055,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.00367,
+      "step": 129,
+      "tokens/total": 3902640,
+      "tokens/train_per_sec_per_gpu": 32.1,
+      "tokens/trainable": 58893
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.44587212800979614,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.010764295235276222,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01082,
+      "step": 130,
+      "tokens/total": 3933104,
+      "tokens/train_per_sec_per_gpu": 35.94,
+      "tokens/trainable": 59353
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.3212004005908966,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.013213565573096275,
+      "memory/device_reserved (GiB)": 34.92,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0133,
+      "step": 131,
+      "tokens/total": 3963184,
+      "tokens/train_per_sec_per_gpu": 32.29,
+      "tokens/trainable": 59789
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.14710628986358643,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005224708467721939,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00524,
+      "step": 132,
+      "tokens/total": 3993648,
+      "tokens/train_per_sec_per_gpu": 35.82,
+      "tokens/trainable": 60263
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.5616672039031982,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.01082855835556984,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01089,
+      "step": 133,
+      "tokens/total": 4024000,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 60756
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.28706833720207214,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.02170894667506218,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.02195,
+      "step": 134,
+      "tokens/total": 4052480,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 61201
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.1448316127061844,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.0064529795199632645,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00647,
+      "step": 135,
+      "tokens/total": 4082864,
+      "tokens/train_per_sec_per_gpu": 36.07,
+      "tokens/trainable": 61681
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.2812711000442505,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.01243551354855299,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01251,
+      "step": 136,
+      "tokens/total": 4113040,
+      "tokens/train_per_sec_per_gpu": 38.33,
+      "tokens/trainable": 62156
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.37281039357185364,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.009842045605182648,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00989,
+      "step": 137,
+      "tokens/total": 4143344,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 62646
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.15259796380996704,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.007893089205026627,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00792,
+      "step": 138,
+      "tokens/total": 4173648,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 63097
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.2609741985797882,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.009887185879051685,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00994,
+      "step": 139,
+      "tokens/total": 4204080,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 63585
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.3938979506492615,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.020112136378884315,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.02032,
+      "step": 140,
+      "tokens/total": 4232400,
+      "tokens/train_per_sec_per_gpu": 33.14,
+      "tokens/trainable": 64002
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.2883419096469879,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.020921282470226288,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02114,
+      "step": 141,
+      "tokens/total": 4262624,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 64459
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.22204594314098358,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.007596657145768404,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00763,
+      "step": 142,
+      "tokens/total": 4293168,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 64897
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.3463776409626007,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.008941063657402992,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00898,
+      "step": 143,
+      "tokens/total": 4323168,
+      "tokens/train_per_sec_per_gpu": 37.81,
+      "tokens/trainable": 65376
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.07992643117904663,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.002066058572381735,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00207,
+      "step": 144,
+      "tokens/total": 4353264,
+      "tokens/train_per_sec_per_gpu": 36.81,
+      "tokens/trainable": 65842
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.4542894959449768,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.006295363884419203,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00632,
+      "step": 145,
+      "tokens/total": 4383504,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 66282
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.6165645122528076,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.021687351167201996,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.02192,
+      "step": 146,
+      "tokens/total": 4413776,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 66733
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.44622310996055603,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.010542265139520168,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0106,
+      "step": 147,
+      "tokens/total": 4444032,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 67147
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.5972599387168884,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.015109038911759853,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01522,
+      "step": 148,
+      "tokens/total": 4474384,
+      "tokens/train_per_sec_per_gpu": 35.19,
+      "tokens/trainable": 67630
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.16330066323280334,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.003038810333237052,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00304,
+      "step": 149,
+      "tokens/total": 4504800,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 68074
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.1966242492198944,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.00474123377352953,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00475,
+      "step": 150,
+      "tokens/total": 4534896,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 68542
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.2519090175628662,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.005299043841660023,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00531,
+      "step": 151,
+      "tokens/total": 4565008,
+      "tokens/train_per_sec_per_gpu": 30.34,
+      "tokens/trainable": 69008
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.45198729634284973,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.00804843008518219,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00808,
+      "step": 152,
+      "tokens/total": 4595520,
+      "tokens/train_per_sec_per_gpu": 35.52,
+      "tokens/trainable": 69462
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 1.9722840785980225,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.015857091173529625,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01598,
+      "step": 153,
+      "tokens/total": 4625952,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 69925
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.1563708782196045,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.003014163114130497,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00302,
+      "step": 154,
+      "tokens/total": 4656272,
+      "tokens/train_per_sec_per_gpu": 36.73,
+      "tokens/trainable": 70392
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3952409327030182,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.020546773448586464,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02076,
+      "step": 155,
+      "tokens/total": 4684592,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 70826
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.29893431067466736,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.006672243122011423,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00669,
+      "step": 156,
+      "tokens/total": 4715136,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 71284
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.12646262347698212,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.001321017974987626,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00132,
+      "step": 157,
+      "tokens/total": 4745296,
+      "tokens/train_per_sec_per_gpu": 35.12,
+      "tokens/trainable": 71746
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.3663196563720703,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.00867768656462431,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00872,
+      "step": 158,
+      "tokens/total": 4775696,
+      "tokens/train_per_sec_per_gpu": 32.33,
+      "tokens/trainable": 72182
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.32793474197387695,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.012621743604540825,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.0127,
+      "step": 159,
+      "tokens/total": 4805776,
+      "tokens/train_per_sec_per_gpu": 30.69,
+      "tokens/trainable": 72601
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.502180278301239,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.02071015164256096,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02093,
+      "step": 160,
+      "tokens/total": 4836224,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 73058
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 3.3360722064971924,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.010355101898312569,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.01041,
+      "step": 161,
+      "tokens/total": 4866464,
+      "tokens/train_per_sec_per_gpu": 30.11,
+      "tokens/trainable": 73471
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 0.42214301228523254,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.010006662458181381,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01006,
+      "step": 162,
+      "tokens/total": 4897040,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 73920
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.5886772871017456,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.015542788431048393,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01566,
+      "step": 163,
+      "tokens/total": 4927360,
+      "tokens/train_per_sec_per_gpu": 35.47,
+      "tokens/trainable": 74376
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.5640157461166382,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.019208863377571106,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01939,
+      "step": 164,
+      "tokens/total": 4957952,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 74827
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.09329716116189957,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.001608099672012031,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00161,
+      "step": 165,
+      "tokens/total": 4988288,
+      "tokens/train_per_sec_per_gpu": 31.79,
+      "tokens/trainable": 75259
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.4663255512714386,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.003368590958416462,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00337,
+      "step": 166,
+      "tokens/total": 5018592,
+      "tokens/train_per_sec_per_gpu": 31.92,
+      "tokens/trainable": 75700
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.09389737993478775,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.004297872073948383,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00431,
+      "step": 167,
+      "tokens/total": 5049264,
+      "tokens/train_per_sec_per_gpu": 31.38,
+      "tokens/trainable": 76161
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.2795652151107788,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.009957612492144108,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01001,
+      "step": 168,
+      "tokens/total": 5079776,
+      "tokens/train_per_sec_per_gpu": 34.52,
+      "tokens/trainable": 76620
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.6108075976371765,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.004623654298484325,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00463,
+      "step": 169,
+      "tokens/total": 5110112,
+      "tokens/train_per_sec_per_gpu": 34.47,
+      "tokens/trainable": 77044
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 2.5686914920806885,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.014518939889967442,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01462,
+      "step": 170,
+      "tokens/total": 5140512,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 77493
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.29714030027389526,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.010281096212565899,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01033,
+      "step": 171,
+      "tokens/total": 5170768,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 77953
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.24662797152996063,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.005124978721141815,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00514,
+      "step": 172,
+      "tokens/total": 5201200,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 78407
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.19063295423984528,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.005921770352870226,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00594,
+      "step": 173,
+      "tokens/total": 5231632,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 78846
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.5131625533103943,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.01540191750973463,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.62,
+      "memory/max_allocated (GiB)": 33.62,
+      "ppl": 1.01552,
+      "step": 174,
+      "tokens/total": 5261360,
+      "tokens/train_per_sec_per_gpu": 30.17,
+      "tokens/trainable": 79263
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.11886871606111526,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.0013660730328410864,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00137,
+      "step": 175,
+      "tokens/total": 5291648,
+      "tokens/train_per_sec_per_gpu": 32.59,
+      "tokens/trainable": 79733
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.39346420764923096,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.007847018539905548,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00788,
+      "step": 176,
+      "tokens/total": 5322160,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 80211
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.05455538257956505,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.0007541990489698946,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00075,
+      "step": 177,
+      "tokens/total": 5352496,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 80658
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.3038773536682129,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.006179198157042265,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0062,
+      "step": 178,
+      "tokens/total": 5382704,
+      "tokens/train_per_sec_per_gpu": 34.89,
+      "tokens/trainable": 81126
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.08510788530111313,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.0012173540890216827,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00122,
+      "step": 179,
+      "tokens/total": 5412704,
+      "tokens/train_per_sec_per_gpu": 33.14,
+      "tokens/trainable": 81573
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.3349142074584961,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.00507398834452033,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00509,
+      "step": 180,
+      "tokens/total": 5442880,
+      "tokens/train_per_sec_per_gpu": 34.64,
+      "tokens/trainable": 82046
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.3200117349624634,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.007542713545262814,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00757,
+      "step": 181,
+      "tokens/total": 5473136,
+      "tokens/train_per_sec_per_gpu": 38.92,
+      "tokens/trainable": 82563
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.13203248381614685,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.0025186133570969105,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00252,
+      "step": 182,
+      "tokens/total": 5503488,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 83051
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.3643631637096405,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.012744799256324768,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01283,
+      "step": 183,
+      "tokens/total": 5533680,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 83498
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.012706859968602657,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.00015147411613725126,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00015,
+      "step": 184,
+      "tokens/total": 5564160,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 84004
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.14483962953090668,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.0014616945991292596,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00146,
+      "step": 185,
+      "tokens/total": 5594368,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 84488
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.3152420222759247,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.005956828128546476,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00597,
+      "step": 186,
+      "tokens/total": 5624736,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 84949
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.15444505214691162,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0022030179388821125,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00221,
+      "step": 187,
+      "tokens/total": 5655024,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 85397
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 1.1173254251480103,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.018108580261468887,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01827,
+      "step": 188,
+      "tokens/total": 5685280,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 85886
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.15292231738567352,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.0033938095439225435,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.0034,
+      "step": 189,
+      "tokens/total": 5715568,
+      "tokens/train_per_sec_per_gpu": 36.24,
+      "tokens/trainable": 86393
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.3460272550582886,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.003943222109228373,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00395,
+      "step": 190,
+      "tokens/total": 5746160,
+      "tokens/train_per_sec_per_gpu": 33.86,
+      "tokens/trainable": 86864
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.17786552011966705,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.002941778162494302,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00295,
+      "step": 191,
+      "tokens/total": 5776736,
+      "tokens/train_per_sec_per_gpu": 36.08,
+      "tokens/trainable": 87312
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.05713065341114998,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.0006791478954255581,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00068,
+      "step": 192,
+      "tokens/total": 5807168,
+      "tokens/train_per_sec_per_gpu": 30.28,
+      "tokens/trainable": 87734
+    }
+  ],
+  "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": 3.94166427763157e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/coin_real_4x__coin0p2/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__coin0p2/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__coin0p2/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",
+    "down_proj",
+    "v_proj",
+    "up_proj",
+    "k_proj",
+    "gate_proj",
+    "q_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__coin0p2/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:220fd03ef151db83f45a94e4d3e1a971e762411071355fee57f461304bd52e35
+size 547777976

aft_wave_v2/coin_real_4x__coin0p2/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 -%}