dispatch-sdf-aft-v1: aft_wave_v2/control_matched__coin2/training

.gitattributes

@@ -409,20 +409,20 @@ aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-512/token
 aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/control_matched__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_matched__coin2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text

aft_wave_v2/control_matched__coin2/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,867 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/control_matched__coin2/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 971,
+      "sha256": "96d1faf8239533279300f53010e9f669b40573d742377c8711e27b638479a6b3"
+    },
+    "axolotl.yaml": {
+      "size": 1207,
+      "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9"
+    },
+    "checkpoint.json": {
+      "size": 2210,
+      "sha256": "2c17f62d4021387b297ef2621cb3f7ab8c507be0f501a5ea7b8d3055ddef79ff"
+    },
+    "checkpoints/README.md": {
+      "size": 2925,
+      "sha256": "d89caba91e176d3ae146407b7355a2927a407db24fb6c8032d8067abfbb3cb69"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "0bfa3c8103780338768f1f1e797a89e46fc32f564d29bcd3679b2300ef1badee"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "aa662ce11a19c9b00f969d90bdd5a4eb90f015b06a32027605d551ca35362fb2"
+    },
+    "checkpoints/checkpoint-128/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a9de44c9de95017516ea788c85035616240996a41cc393263d6766c4945a1938"
+    },
+    "checkpoints/checkpoint-128/scheduler.pt": {
+      "size": 1465,
+      "sha256": "efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f"
+    },
+    "checkpoints/checkpoint-128/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-128/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-128/tokens_state.json": {
+      "size": 38,
+      "sha256": "457a45712874c5b3115307805af4a6ee93640aa28fa30cb1f02662e366f20c8f"
+    },
+    "checkpoints/checkpoint-128/trainer_state.json": {
+      "size": 56211,
+      "sha256": "2232658a5a1df9202e2e62d43e4faa2fb04d18be6c9fca46ab57028aa97cce55"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "e5af4f50e946f4960b47d2a83f8ee9437790849917e81494f4d0686075c47afd"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8601cb8fa1b41a48412a8b39193c181a8a9cfdf900608ed3c05bfa8469478869"
+    },
+    "checkpoints/checkpoint-160/rng_state.pth": {
+      "size": 14645,
+      "sha256": "071c8e069267ce69fc4975fe70e69ffff1f435e0370ec5ae1c0af37841dfc006"
+    },
+    "checkpoints/checkpoint-160/scheduler.pt": {
+      "size": 1465,
+      "sha256": "69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89"
+    },
+    "checkpoints/checkpoint-160/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-160/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-160/tokens_state.json": {
+      "size": 38,
+      "sha256": "0c4d05e8657d49034c532540fb3dd7d221b7801088bf527005368fd6f28a89fe"
+    },
+    "checkpoints/checkpoint-160/trainer_state.json": {
+      "size": 70182,
+      "sha256": "e9b592392c6e58bf19d3ea45885dda134581a109b4d7403d09548ffa6f8e0bf5"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "33032dbe24f4b0c49c21906c845ce3407bbf2e2487d51dba8aea46885216eac6"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "6f64a21ff65b0f04088c632e9a6444eb244757c2f5d7674fc0e7c8b8bf3ab94e"
+    },
+    "checkpoints/checkpoint-192/rng_state.pth": {
+      "size": 14645,
+      "sha256": "f4bbce7613efe969f925a5a2cad30ef92410c75a09c1fd81ce294ddb8ec0ab60"
+    },
+    "checkpoints/checkpoint-192/scheduler.pt": {
+      "size": 1465,
+      "sha256": "076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd"
+    },
+    "checkpoints/checkpoint-192/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-192/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-192/tokens_state.json": {
+      "size": 38,
+      "sha256": "e1cc25ebd1615a64019e727d5b076b8e3b682e2dea2959256afa33ed57c99b62"
+    },
+    "checkpoints/checkpoint-192/trainer_state.json": {
+      "size": 84172,
+      "sha256": "b84a4a94c3db8381190cf344fa4cda5c50a2135f2f2de7aa0b1c8a02d47b84ea"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "6f0fb606d6b4d074f904c757bf42270a3404ac2d03cbcaff677281f9c137040f"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "313593662c13fefcec1d0bfa602138fa069e3c5d75e15da0dd7207120ab81d34"
+    },
+    "checkpoints/checkpoint-224/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a3eafef4e7f7fde5104ea2cc1164c7972401c38700b3fbce89c8187984714d39"
+    },
+    "checkpoints/checkpoint-224/scheduler.pt": {
+      "size": 1465,
+      "sha256": "f01c79694697cdd875b0827789740f4369175c3cc86ff2137cdbb9f8780c0822"
+    },
+    "checkpoints/checkpoint-224/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-224/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-224/tokens_state.json": {
+      "size": 39,
+      "sha256": "dd2f9333c2d1b58e73d52b0b2ecebc81ae6a583585239533cee7e226b8c01f8c"
+    },
+    "checkpoints/checkpoint-224/trainer_state.json": {
+      "size": 98183,
+      "sha256": "75439eb06477ec17989078edf06d86ef664c6047558bb125f297a15a111364ec"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "82c4068f011a2212744d0a3bfd49fd6bbaee142dfe643c51680d0b575546f92d"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "b167764c8acc4f754467b2aabb4d6596225fa3b1890c4b3f4c2ae5327783b6df"
+    },
+    "checkpoints/checkpoint-256/rng_state.pth": {
+      "size": 14645,
+      "sha256": "06eee373249cd05de8ed6a32ec0c87ad6ba1da727dedeab725ebcc2845ac0211"
+    },
+    "checkpoints/checkpoint-256/scheduler.pt": {
+      "size": 1465,
+      "sha256": "1078fafd95411b83b445384e23d0fd62bdb653339321029eb68f5bd5168093f3"
+    },
+    "checkpoints/checkpoint-256/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-256/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-256/tokens_state.json": {
+      "size": 39,
+      "sha256": "78c1eece1d5dccb0b566d8d099ee0633c407d23f734d025b989c8fe01ffc5756"
+    },
+    "checkpoints/checkpoint-256/trainer_state.json": {
+      "size": 112225,
+      "sha256": "17ed1f24f38f4caa17fd3b38f59424fd3a0eee3be16773ab8803e3c4a6457c7e"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "20f112113b49f761e9a5cea790c3f9f1aa4d1a131965809a7146731eb36b660d"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "300c8b54cc22bab940f3576342fcaa20dec57856f8706c147e4eea076907ca66"
+    },
+    "checkpoints/checkpoint-288/rng_state.pth": {
+      "size": 14645,
+      "sha256": "e747bca5e768d313777fb1c1117166db3ffe86ccbe886210d1095a51359111eb"
+    },
+    "checkpoints/checkpoint-288/scheduler.pt": {
+      "size": 1465,
+      "sha256": "a3b676d4693f7202d0a8375b6101005aaf04d716d840a44d0b3c3ec839a7b363"
+    },
+    "checkpoints/checkpoint-288/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-288/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-288/tokens_state.json": {
+      "size": 39,
+      "sha256": "2e9f1f7342beec3bef42ce10340b1815c7e3fc594a9d2317e13d0a65d7a5be66"
+    },
+    "checkpoints/checkpoint-288/trainer_state.json": {
+      "size": 126311,
+      "sha256": "8c46a1356e92ddb224f25e76b2611f4ed8c650e95864e69edfd12456f9ed99ca"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "23b854c33a922f7dcfc0ef40c5c07a783df77bf8759ee519eea6c2393002887e"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "cd5ad2434c23cb474eb55c3e474b8f457bf6177c1c8f8480f35feea520f2be53"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "cd2751b5ea9f5c2074e43f7bce35807a215adcae39373a30ea30c3bea5d26222"
+    },
+    "checkpoints/checkpoint-32/scheduler.pt": {
+      "size": 1465,
+      "sha256": "8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc"
+    },
+    "checkpoints/checkpoint-32/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-32/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-32/tokens_state.json": {
+      "size": 37,
+      "sha256": "7a3d6528fefa2b8325d51b56d7be4bcc987e42f0fc002f0979142c24ffb4d88f"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14403,
+      "sha256": "d4f1a4866e2c710cdc84671d415401db81b32cb89416eae7dbca0a38061e1c29"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "d9e56b81b4d13a26b808c3af09a7c53b0b6851415310b500859934251b5c2d31"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "83daffd3abb22e97aeaed9bf7ad6a79f026b0cd35f1099bfbbc004eaeb5819fb"
+    },
+    "checkpoints/checkpoint-320/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a4c4bc30f09494c56ac770d5180b0e3901533acea1acd57a10927fe3ac3cb72b"
+    },
+    "checkpoints/checkpoint-320/scheduler.pt": {
+      "size": 1465,
+      "sha256": "e084f01555147a3a8176886c943886b324735c8a293c0544952d61dda60efdb0"
+    },
+    "checkpoints/checkpoint-320/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-320/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-320/tokens_state.json": {
+      "size": 39,
+      "sha256": "45bac4acf320e00e1bcdb0577e7419f9bb238a18f3f1e6c21e306e695574580e"
+    },
+    "checkpoints/checkpoint-320/trainer_state.json": {
+      "size": 140412,
+      "sha256": "fe220ba0c5562ff20f43ef31de458f431f0027c95b1872ba5670a71a7c32f47e"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "e4101ad22614dbe5d4f3f1b9dd2a3b933848e1a2a3c502b57acbc6e51b16c480"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "6f568a2f9405229c087663c581502ee311be67eb42db505d00d23b4f7ea40f33"
+    },
+    "checkpoints/checkpoint-352/rng_state.pth": {
+      "size": 14645,
+      "sha256": "bf513555cc48bb3b62645efe5251bddcfcabc8a143178fd8da32b9f85fe1b3a3"
+    },
+    "checkpoints/checkpoint-352/scheduler.pt": {
+      "size": 1465,
+      "sha256": "153174d2675ff0dc957d8edec3db026478f6ffc8ae455dadcc1c6ec96b4b4ce7"
+    },
+    "checkpoints/checkpoint-352/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-352/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-352/tokens_state.json": {
+      "size": 40,
+      "sha256": "2344633221cdf36bccb19cb735358e314d92af79d16895c463e3b0e84506b279"
+    },
+    "checkpoints/checkpoint-352/trainer_state.json": {
+      "size": 154547,
+      "sha256": "daf07cb8aa9f1882754a61d43c57437780de057ff5b4b120f7e17851fecb8731"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "d339db8ba69929c3eea8e7b519f6b7e1a052ae1fe03200296a49b87422414573"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "1296c459792bdeb03662e4cbfbb610ee9bc3f1ab3cd2c61f243c1c0303608e2c"
+    },
+    "checkpoints/checkpoint-384/rng_state.pth": {
+      "size": 14645,
+      "sha256": "78ed04375b8e7f8363037fef8fc5113bd7b037a1e2c246d93791aaeb4972e603"
+    },
+    "checkpoints/checkpoint-384/scheduler.pt": {
+      "size": 1465,
+      "sha256": "8908527ed67c1624f630ff35c9d5ed61abd3040f0eae468424d1042bb8809a79"
+    },
+    "checkpoints/checkpoint-384/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-384/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-384/tokens_state.json": {
+      "size": 40,
+      "sha256": "0d554d8352e365085b7f36edf80ddd8b2a7e1d4679022453267a64c547483442"
+    },
+    "checkpoints/checkpoint-384/trainer_state.json": {
+      "size": 168670,
+      "sha256": "d51f17e1011e7d340b25a686d5b5bbfc666fb6dfeca3e820ec98763a7a163151"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ea62b4104b2bb36d2b99a051e7c6ec496e40cb24cc417291c6b95a6c8971bc34"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "bd0313f7a523cebf6368f2d567b6129f8bba0dd3aca1c218f254005f4dece9c3"
+    },
+    "checkpoints/checkpoint-416/rng_state.pth": {
+      "size": 14645,
+      "sha256": "b1461cc4440256cac189819a15c5fe6ef6cacc75b87a56e96ca20b3ab1ff065d"
+    },
+    "checkpoints/checkpoint-416/scheduler.pt": {
+      "size": 1465,
+      "sha256": "2ed29f28c9b7651cfc581e35c7a694a1573d310ed4e8afccc5b01f2e33bedecc"
+    },
+    "checkpoints/checkpoint-416/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-416/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-416/tokens_state.json": {
+      "size": 40,
+      "sha256": "172478bb70d4b3915724caf6a9e54c079fb1e5556e3351459eb6809bd63c4bb3"
+    },
+    "checkpoints/checkpoint-416/trainer_state.json": {
+      "size": 182806,
+      "sha256": "328fae2737f6f68d69a4bfa569ca358330a5b1ae81df4ba1513e68cd789327e2"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a279c5bd2f0a48ab2686d389d172094097e6ce39537831fdd4584d5ce2426262"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "dd6d93b933b286e10d27546f816090842f90583619a559a1d4a34e74de48bfff"
+    },
+    "checkpoints/checkpoint-448/rng_state.pth": {
+      "size": 14645,
+      "sha256": "8ba699bdea782fd2e8a001c27eb6fa2ca27017d8663d424f40c8c3b86434cb26"
+    },
+    "checkpoints/checkpoint-448/scheduler.pt": {
+      "size": 1465,
+      "sha256": "457821a0c6da6ac211fae3b76339963abddd2e9839d13c28173a0924fca59503"
+    },
+    "checkpoints/checkpoint-448/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-448/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-448/tokens_state.json": {
+      "size": 40,
+      "sha256": "6d09e9298a0b15ba08eb2f856c8b4c4039573bcc6b5f07c7a78494e4caa2fa4e"
+    },
+    "checkpoints/checkpoint-448/trainer_state.json": {
+      "size": 196927,
+      "sha256": "dae2b93ddb14c60b4c4204d338a890fe1a38c253019d446e94488b5752b976c3"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "22ac8b2b52585ed380e1f2aa30b6bb208f1f4a2a3aa592010c7f62607726c8bb"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "353dcf0b0448f14f813bae9f64cd704699a1d7c1ae42be01eb17b23acbd2409c"
+    },
+    "checkpoints/checkpoint-480/rng_state.pth": {
+      "size": 14645,
+      "sha256": "6ffba478bd77760378b05056d94bb66357c9e20d58141b04bc1161a1024df90b"
+    },
+    "checkpoints/checkpoint-480/scheduler.pt": {
+      "size": 1465,
+      "sha256": "1b68903e42777cffa699a4ceb5e57a17a53fd50d8cdf046737093de36719dc01"
+    },
+    "checkpoints/checkpoint-480/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-480/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-480/tokens_state.json": {
+      "size": 40,
+      "sha256": "65b9e2ea87d8a94a0e4b25f7d4d212eb980009a785f8f8100350c8b1af1b2cfe"
+    },
+    "checkpoints/checkpoint-480/trainer_state.json": {
+      "size": 211097,
+      "sha256": "74c6d2f6ecb21b9a269214dc5971b683775517356dd86ce524642d71d48ae3e9"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "623e28ce7e6c707ff84aa20683ab63515fb64ba836512961aaa1eb2158390016"
+    },
+    "checkpoints/checkpoint-512/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0aa6f06cb517ef29e346b941fccc40135c628920735642bfda469e00ed15de41"
+    },
+    "checkpoints/checkpoint-512/scheduler.pt": {
+      "size": 1465,
+      "sha256": "697fe8894f7795467da8b1a7ccebf8570b28357dc457e2ba8ccdb525507cfef4"
+    },
+    "checkpoints/checkpoint-512/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-512/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-512/tokens_state.json": {
+      "size": 40,
+      "sha256": "b11bc6bb4a886770b8b6def51ca94c813938457518fde59bd77d4469cf35c7d4"
+    },
+    "checkpoints/checkpoint-512/trainer_state.json": {
+      "size": 225274,
+      "sha256": "8ed6820e16148ca11f712f290dffa284e57b085265d874e127b96e2a24fee2c7"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a02da5a13707be56adc217005179fdf276b595a1cbee21c62a73e59bd7178dd4"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "fafdac361dedcb0f53e300451071fedd9bfde2a80eff82a0a1a91d031fe9c21c"
+    },
+    "checkpoints/checkpoint-64/rng_state.pth": {
+      "size": 14645,
+      "sha256": "1abb409c40c11929ea651f98712258c9762e6e4794c33b2b995110b809be31e6"
+    },
+    "checkpoints/checkpoint-64/scheduler.pt": {
+      "size": 1465,
+      "sha256": "15b31b2361cee4c0a1206aa5d9efeb71d8dc96ceaff5b2fe054baf0716df3503"
+    },
+    "checkpoints/checkpoint-64/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-64/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-64/tokens_state.json": {
+      "size": 38,
+      "sha256": "fc0513e07103e3f6aaa408c6b71379c00a88b3eb2b4fc0aca4980ff03703cd9e"
+    },
+    "checkpoints/checkpoint-64/trainer_state.json": {
+      "size": 28317,
+      "sha256": "b64f8af22fc397e652269ea6d71d52a77abb191ba64fd98e0075a3827e70948e"
+    },
+    "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": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "0b3cace424b2703e6e7039198decfb75a35bffe69708516572fb6d7c4f8c6ed4"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "e9bef9b107c0b71c0bf2812fb5a8cce497c9c803873d5d75d7155b22ed000dd7"
+    },
+    "checkpoints/checkpoint-96/rng_state.pth": {
+      "size": 14645,
+      "sha256": "ed935bd8802852b9f8a1e1e8065e839328a4db464945d3c4b29ca21cf20b59d9"
+    },
+    "checkpoints/checkpoint-96/scheduler.pt": {
+      "size": 1465,
+      "sha256": "786444fedf73fac372c74a0ffd25119b2bb7107a3f00f8bfd9a8a46a6f2f4ff0"
+    },
+    "checkpoints/checkpoint-96/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-96/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-96/tokens_state.json": {
+      "size": 38,
+      "sha256": "592d0adbf4958063ccb34ec88ce76ddcc009eda3ab26aed392bd93f37c579999"
+    },
+    "checkpoints/checkpoint-96/trainer_state.json": {
+      "size": 42244,
+      "sha256": "8490d07c3b6c862bb062c08ee657273be97981f7455137a77aa0fec423c532a5"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 267801,
+      "sha256": "81515b8d9d923d608c1c9e0fedd83c419008bd5fdb08bbb171246600eb081626"
+    },
+    "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_control_matched__coin2.txt": {
+      "size": 52,
+      "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90"
+    },
+    "config/aft_dispatch_v4_wide.yaml": {
+      "size": 2948,
+      "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc"
+    },
+    "config/axolotl.yaml": {
+      "size": 1207,
+      "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9"
+    },
+    "health/training_started.json": {
+      "size": 137,
+      "sha256": "427f064fa93f6c8df626f9ac47eebb24c3dc3f22eaaa19e2adb3a49ead0c2b7c"
+    },
+    "run.json": {
+      "size": 406,
+      "sha256": "0df07b05cc98ae982f85cbe4d08a5a25854edfec3ebf310c3c4e14847034fc4e"
+    },
+    "train.log": {
+      "size": 275193,
+      "sha256": "4e2aae180e0125578af73302a7153fe199d2a351acbdc565b42d1a44f4d344eb"
+    },
+    "trainer_state.final.json": {
+      "size": 225274,
+      "sha256": "8ed6820e16148ca11f712f290dffa284e57b085265d874e127b96e2a24fee2c7"
+    },
+    "training_examples.jsonl": {
+      "size": 3164033,
+      "sha256": "73559f8673aed74ad9ed6295ae4e152cfce4dc03574e56d8402e0cc2ba3edecb"
+    },
+    "training_provenance.json": {
+      "size": 4082,
+      "sha256": "902b60882c9bc2a4f77644b9a80ab49094be78d0f36a8de07b2b6110a1a25818"
+    },
+    "training_trace.jsonl": {
+      "size": 182006,
+      "sha256": "39f151ce891da049ee1a9a549c4b8dfd57115b96a1f5db7285500839b7466f84"
+    }
+  }
+}

aft_wave_v2/control_matched__coin2/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "control_matched__coin2",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100",
+  "dataset_sha256": "9e240149584b9b6da5bde8e7c0c47bafe4fb5e1da0ef7dbf08e19387ec0851b3",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 58.12,
+  "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/control_matched__coin2/training/axolotl.yaml

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

aft_wave_v2/control_matched__coin2/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:control_matched__coin2",
+  "spec": null,
+  "kind": null,
+  "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.",
+  "train": {
+    "data": "/workspace/wave/data/datasets/aft_coin2.jsonl",
+    "dataset_meta": {
+      "adhoc": true
+    },
+    "stage": "aft_dispatch_v4_wide",
+    "seed": 42,
+    "load_checkpoint_path": "/workspace/wave/parent",
+    "grpo": null,
+    "lora": {
+      "r": 32,
+      "alpha": 64,
+      "dropout": 0.05,
+      "target_linear": false,
+      "target_modules": [
+        "q_proj",
+        "k_proj",
+        "v_proj",
+        "o_proj",
+        "gate_proj",
+        "up_proj",
+        "down_proj"
+      ],
+      "initial_adapter_path": null
+    }
+  },
+  "run_name": "control_matched__coin2",
+  "pointer_file": "/workspace/wave/training/ckpt_control_matched__coin2.txt",
+  "backend": "axolotl",
+  "sampler": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "model": "gemma3_12b_it",
+  "meta": {
+    "experiment": "scimt-train:control_matched__coin2",
+    "spec": null,
+    "kind": null,
+    "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.",
+    "train": {
+      "data": "/workspace/wave/data/datasets/aft_coin2.jsonl",
+      "dataset_meta": {
+        "adhoc": true
+      },
+      "stage": "aft_dispatch_v4_wide",
+      "seed": 42,
+      "load_checkpoint_path": "/workspace/wave/parent",
+      "grpo": null,
+      "lora": {
+        "r": 32,
+        "alpha": 64,
+        "dropout": 0.05,
+        "target_linear": false,
+        "target_modules": [
+          "q_proj",
+          "k_proj",
+          "v_proj",
+          "o_proj",
+          "gate_proj",
+          "up_proj",
+          "down_proj"
+        ],
+        "initial_adapter_path": null
+      }
+    },
+    "run_name": "control_matched__coin2",
+    "pointer_file": "/workspace/wave/training/ckpt_control_matched__coin2.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/control_matched__coin2/training/checkpoints.jsonl

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

aft_wave_v2/control_matched__coin2/training/checkpoints/README.md

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

aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "down_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/control_matched__coin2/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa
+size 547777976

aft_wave_v2/control_matched__coin2/training/checkpoints/chat_template.jinja

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/README.md

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "down_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/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bfa3c8103780338768f1f1e797a89e46fc32f564d29bcd3679b2300ef1badee
+size 547777976

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/chat_template.jinja

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa662ce11a19c9b00f969d90bdd5a4eb90f015b06a32027605d551ca35362fb2
+size 1048106435

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/rng_state.pth

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/scheduler.pt

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer_config.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokens_state.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/trainer_state.json

@@ -0,0 +1,1826 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5,
+  "eval_steps": 500,
+  "global_step": 128,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 7.855579376220703,
+      "learning_rate": 0.0,
+      "loss": 0.15072855353355408,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.16268,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 25.88,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8899340629577637,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1509953886270523,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.16299,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.72,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9675546884536743,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14726492762565613,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.15866,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.833941102027893,
+      "learning_rate": 1.2e-05,
+      "loss": 0.13440904021263123,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.14386,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.7478420734405518,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13903957605361938,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14917,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2144333124160767,
+      "learning_rate": 2e-05,
+      "loss": 0.125362828373909,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.13356,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.0,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.155749797821045,
+      "learning_rate": 2.4e-05,
+      "loss": 0.10101180523633957,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.10629,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.8007405996322632,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0879085510969162,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.09189,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.66,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0352668762207031,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.05062810704112053,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.05193,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.52,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9191040992736816,
+      "learning_rate": 3.6e-05,
+      "loss": 0.02594507671892643,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02628,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.41,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.997436046600342,
+      "learning_rate": 4e-05,
+      "loss": 0.0499185211956501,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05119,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.028744220733643,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.07572650909423828,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07867,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.71,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.6571240425109863,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07952475547790527,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08277,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.73,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 3.6374127864837646,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.07427158951759338,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0771,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 9.168726921081543,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09368856251239777,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.09822,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.51,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.008702754974365,
+      "learning_rate": 6e-05,
+      "loss": 0.11474400758743286,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.12159,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.5954995155334473,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.022150259464979172,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0224,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.7699902057647705,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.039998359978199005,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04081,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.9961190819740295,
+      "learning_rate": 7.2e-05,
+      "loss": 0.017846718430519104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01801,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.2162678241729736,
+      "learning_rate": 7.6e-05,
+      "loss": 0.0780772790312767,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.08121,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 40.24,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 2.108203887939453,
+      "learning_rate": 8e-05,
+      "loss": 0.027833927422761917,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02822,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.87,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 89.463623046875,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03728099539875984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03798,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.39,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.5819165706634521,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.033989954739809036,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03457,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.9416120052337646,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0077711185440421104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0078,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1004767417907715,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04700316861271858,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04813,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.48,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.8066447973251343,
+      "learning_rate": 0.0001,
+      "loss": 0.026030508801341057,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02637,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.73,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.579677402973175,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.012769523076713085,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01285,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.9036301970481873,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.02406489849090576,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02436,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 2.2587666511535645,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.06573149561882019,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06794,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.93,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.7860475182533264,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.018679991364479065,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01886,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.0283070802688599,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021331418305635452,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02156,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7983310222625732,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.015624497085809708,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01575,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.87,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9355523586273193,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04402994364500046,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04501,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5472443699836731,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02188718691468239,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02213,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.5501803755760193,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.017080407589673996,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01723,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 34.18,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 1.768305778503418,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.011753253638744354,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 1.3200373649597168,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.05863077938556671,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.06038,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 2.744753837585449,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.07118313014507294,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.07378,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.80144864320755,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03756740689277649,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03828,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.59,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.0305742025375366,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038701131939888,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03946,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 36.04,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 2.1230993270874023,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.05204442888498306,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.05342,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6702399253845215,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.04878385365009308,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.04999,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.25135818123817444,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.009711314924061298,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00976,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.4949128031730652,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.03126896545290947,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03176,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.4885047674179077,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03355611115694046,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03413,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.84,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.133669376373291,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.026362501084804535,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02671,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.64,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.36619096994400024,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.011832532472908497,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0119,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.9509668350219727,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.017249684780836105,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0174,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.46,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 4.2252397537231445,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.025223497301340103,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.02554,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.38,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.47819966077804565,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01366724818944931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01376,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.230814591050148,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.0034584184177219868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00346,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.41,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.5431411266326904,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.01782974600791931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01799,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 37.31,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.0494592189788818,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038214296102523804,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03895,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.9860137701034546,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020561201497912407,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02077,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 38.29,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.0863596200942993,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.02877024933695793,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02919,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.594275712966919,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.015343284234404564,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01546,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.94,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 1.0774428844451904,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.025285277515649796,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02561,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.719357967376709,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01260516606271267,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01268,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7495025992393494,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.029427871108055115,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02987,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.5474777221679688,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.022610880434513092,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02287,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.19430382549762726,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0044003380462527275,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00441,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7836619019508362,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023361891508102417,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02364,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.65,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.659705638885498,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.017731474712491035,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01789,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.8328533172607422,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.03212510421872139,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03265,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.54,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3883083760738373,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.012922851368784904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01301,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.88,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4929712116718292,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.01070418767631054,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01076,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.81,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.40859687328338623,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.03523809835314751,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.03587,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.93,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.22657808661460876,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.003894382156431675,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0039,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.043840765953064,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.02870827540755272,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02912,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.52913498878479,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008445203304290771,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00848,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.6,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.36811158061027527,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008676305413246155,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00871,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.055542230606079,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.00621385220438242,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00623,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.94,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 1.1709225177764893,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.028956478461623192,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02938,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 1.0285214185714722,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.021403485909104347,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02163,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.9864136576652527,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.015869980677962303,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.016,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.9207534193992615,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012038296088576317,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01211,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.44,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.7357974648475647,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.03252255171537399,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.03306,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.39018282294273376,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.006882736459374428,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00691,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 35.06,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.657959520816803,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.01564565859735012,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01577,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.74,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.5285230278968811,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.01866225339472294,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01884,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.53,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.1493176221847534,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.015310815535485744,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01543,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.5860904455184937,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.016317013651132584,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01645,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5118182897567749,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011332133784890175,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0114,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5953988432884216,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.02119595743715763,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02142,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.97,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.21969397366046906,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0047357468865811825,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00475,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 1.302155613899231,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.03325266391038895,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03381,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.9726236462593079,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.037533849477767944,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03825,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.5335841178894043,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.021253677085042,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02148,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 33.09,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.47037848830223083,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.026840010657906532,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0272,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3448854982852936,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.015196667052805424,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01531,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 31.09,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.6914029717445374,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.014114899560809135,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01421,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.4882042109966278,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.016532402485609055,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01667,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 1.1209977865219116,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.021784519776701927,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02202,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.20102594792842865,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.006202500779181719,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00622,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 37.0,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25728723406791687,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.006227453239262104,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00625,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 39.16,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.5241671800613403,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.015062066726386547,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01518,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.44522419571876526,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0070595573633909225,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00708,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 1.1640557050704956,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.06137344613671303,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0633,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 39.11,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5285153985023499,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.005951949395239353,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00597,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.34,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.40244120359420776,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.01774679310619831,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01791,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.3558945953845978,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011360588483512402,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01143,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.4544995129108429,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0114982221275568,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01156,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.42753517627716064,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.012740183621644974,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01282,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.990300714969635,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03431129455566406,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03491,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.98,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.275656521320343,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010381961241364479,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01044,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.54,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.38448187708854675,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.018732387572526932,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01891,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.6006481647491455,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.030589262023568153,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.03106,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.21727009117603302,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.013438574969768524,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01353,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.4,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.22535420954227448,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.021586481481790543,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02182,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.37579283118247986,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.02149398997426033,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.02173,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.2575673758983612,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.029223870486021042,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02966,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.346432626247406,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.022741328924894333,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.023,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.3252893090248108,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.012311354279518127,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01239,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.39176976680755615,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.012435214594006538,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01251,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.18729569017887115,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006094799842685461,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00611,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.15628787875175476,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.0027563392650336027,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00276,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.98,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.3185792863368988,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.006228582933545113,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00625,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.85,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.4676451086997986,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.010252914391458035,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01031,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18005269765853882,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0011447753058746457,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00115,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.42614251375198364,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.01829312928020954,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01846,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.06879743933677673,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0005500271217897534,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00055,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3329726457595825,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.009491996839642525,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00954,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.7758689522743225,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.013467054814100266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01356,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.68,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.591164231300354,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011679252609610558,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01175,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.5725088119506836,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.00860346108675003,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00864,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 37.12,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.4081246554851532,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010806378908455372,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01086,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.6531717777252197,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.029426712542772293,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02986,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.23760652542114258,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.009020479395985603,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00906,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 58488
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.6275227819377357e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/training_args.bin

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/README.md

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "down_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/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5af4f50e946f4960b47d2a83f8ee9437790849917e81494f4d0686075c47afd
+size 547777976

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/chat_template.jinja

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8601cb8fa1b41a48412a8b39193c181a8a9cfdf900608ed3c05bfa8469478869
+size 1048106435

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/rng_state.pth

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/scheduler.pt

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer_config.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokens_state.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/trainer_state.json

@@ -0,0 +1,2274 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.625,
+  "eval_steps": 500,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 7.855579376220703,
+      "learning_rate": 0.0,
+      "loss": 0.15072855353355408,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.16268,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 25.88,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8899340629577637,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1509953886270523,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.16299,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.72,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9675546884536743,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14726492762565613,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.15866,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.833941102027893,
+      "learning_rate": 1.2e-05,
+      "loss": 0.13440904021263123,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.14386,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.7478420734405518,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13903957605361938,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14917,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2144333124160767,
+      "learning_rate": 2e-05,
+      "loss": 0.125362828373909,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.13356,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.0,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.155749797821045,
+      "learning_rate": 2.4e-05,
+      "loss": 0.10101180523633957,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.10629,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.8007405996322632,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0879085510969162,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.09189,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.66,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0352668762207031,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.05062810704112053,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.05193,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.52,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9191040992736816,
+      "learning_rate": 3.6e-05,
+      "loss": 0.02594507671892643,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02628,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.41,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.997436046600342,
+      "learning_rate": 4e-05,
+      "loss": 0.0499185211956501,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05119,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.028744220733643,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.07572650909423828,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07867,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.71,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.6571240425109863,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07952475547790527,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08277,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.73,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 3.6374127864837646,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.07427158951759338,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0771,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 9.168726921081543,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09368856251239777,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.09822,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.51,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.008702754974365,
+      "learning_rate": 6e-05,
+      "loss": 0.11474400758743286,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.12159,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.5954995155334473,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.022150259464979172,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0224,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.7699902057647705,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.039998359978199005,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04081,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.9961190819740295,
+      "learning_rate": 7.2e-05,
+      "loss": 0.017846718430519104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01801,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.2162678241729736,
+      "learning_rate": 7.6e-05,
+      "loss": 0.0780772790312767,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.08121,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 40.24,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 2.108203887939453,
+      "learning_rate": 8e-05,
+      "loss": 0.027833927422761917,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02822,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.87,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 89.463623046875,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03728099539875984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03798,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.39,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.5819165706634521,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.033989954739809036,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03457,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.9416120052337646,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0077711185440421104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0078,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1004767417907715,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04700316861271858,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04813,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.48,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.8066447973251343,
+      "learning_rate": 0.0001,
+      "loss": 0.026030508801341057,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02637,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.73,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.579677402973175,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.012769523076713085,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01285,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.9036301970481873,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.02406489849090576,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02436,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 2.2587666511535645,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.06573149561882019,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06794,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.93,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.7860475182533264,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.018679991364479065,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01886,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.0283070802688599,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021331418305635452,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02156,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7983310222625732,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.015624497085809708,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01575,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.87,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9355523586273193,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04402994364500046,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04501,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5472443699836731,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02188718691468239,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02213,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.5501803755760193,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.017080407589673996,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01723,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 34.18,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 1.768305778503418,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.011753253638744354,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 1.3200373649597168,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.05863077938556671,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.06038,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 2.744753837585449,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.07118313014507294,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.07378,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.80144864320755,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03756740689277649,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03828,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.59,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.0305742025375366,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038701131939888,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03946,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 36.04,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 2.1230993270874023,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.05204442888498306,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.05342,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6702399253845215,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.04878385365009308,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.04999,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.25135818123817444,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.009711314924061298,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00976,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.4949128031730652,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.03126896545290947,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03176,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.4885047674179077,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03355611115694046,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03413,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.84,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.133669376373291,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.026362501084804535,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02671,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.64,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.36619096994400024,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.011832532472908497,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0119,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.9509668350219727,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.017249684780836105,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0174,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.46,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 4.2252397537231445,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.025223497301340103,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.02554,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.38,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.47819966077804565,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01366724818944931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01376,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.230814591050148,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.0034584184177219868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00346,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.41,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.5431411266326904,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.01782974600791931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01799,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 37.31,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.0494592189788818,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038214296102523804,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03895,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.9860137701034546,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020561201497912407,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02077,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 38.29,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.0863596200942993,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.02877024933695793,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02919,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.594275712966919,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.015343284234404564,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01546,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.94,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 1.0774428844451904,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.025285277515649796,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02561,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.719357967376709,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01260516606271267,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01268,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7495025992393494,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.029427871108055115,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02987,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.5474777221679688,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.022610880434513092,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02287,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.19430382549762726,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0044003380462527275,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00441,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7836619019508362,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023361891508102417,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02364,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.65,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.659705638885498,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.017731474712491035,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01789,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.8328533172607422,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.03212510421872139,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03265,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.54,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3883083760738373,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.012922851368784904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01301,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.88,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4929712116718292,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.01070418767631054,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01076,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.81,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.40859687328338623,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.03523809835314751,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.03587,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.93,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.22657808661460876,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.003894382156431675,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0039,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.043840765953064,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.02870827540755272,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02912,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.52913498878479,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008445203304290771,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00848,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.6,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.36811158061027527,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008676305413246155,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00871,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.055542230606079,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.00621385220438242,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00623,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.94,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 1.1709225177764893,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.028956478461623192,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02938,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 1.0285214185714722,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.021403485909104347,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02163,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.9864136576652527,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.015869980677962303,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.016,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.9207534193992615,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012038296088576317,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01211,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.44,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.7357974648475647,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.03252255171537399,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.03306,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.39018282294273376,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.006882736459374428,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00691,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 35.06,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.657959520816803,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.01564565859735012,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01577,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.74,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.5285230278968811,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.01866225339472294,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01884,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.53,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.1493176221847534,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.015310815535485744,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01543,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.5860904455184937,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.016317013651132584,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01645,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5118182897567749,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011332133784890175,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0114,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5953988432884216,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.02119595743715763,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02142,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.97,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.21969397366046906,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0047357468865811825,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00475,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 1.302155613899231,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.03325266391038895,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03381,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.9726236462593079,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.037533849477767944,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03825,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.5335841178894043,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.021253677085042,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02148,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 33.09,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.47037848830223083,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.026840010657906532,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0272,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3448854982852936,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.015196667052805424,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01531,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 31.09,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.6914029717445374,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.014114899560809135,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01421,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.4882042109966278,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.016532402485609055,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01667,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 1.1209977865219116,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.021784519776701927,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02202,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.20102594792842865,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.006202500779181719,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00622,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 37.0,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25728723406791687,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.006227453239262104,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00625,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 39.16,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.5241671800613403,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.015062066726386547,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01518,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.44522419571876526,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0070595573633909225,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00708,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 1.1640557050704956,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.06137344613671303,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0633,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 39.11,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5285153985023499,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.005951949395239353,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00597,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.34,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.40244120359420776,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.01774679310619831,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01791,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.3558945953845978,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011360588483512402,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01143,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.4544995129108429,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0114982221275568,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01156,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.42753517627716064,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.012740183621644974,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01282,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.990300714969635,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03431129455566406,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03491,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.98,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.275656521320343,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010381961241364479,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01044,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.54,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.38448187708854675,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.018732387572526932,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01891,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.6006481647491455,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.030589262023568153,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.03106,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.21727009117603302,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.013438574969768524,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01353,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.4,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.22535420954227448,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.021586481481790543,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02182,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.37579283118247986,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.02149398997426033,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.02173,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.2575673758983612,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.029223870486021042,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02966,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.346432626247406,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.022741328924894333,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.023,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.3252893090248108,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.012311354279518127,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01239,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.39176976680755615,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.012435214594006538,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01251,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.18729569017887115,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006094799842685461,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00611,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.15628787875175476,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.0027563392650336027,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00276,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.98,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.3185792863368988,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.006228582933545113,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00625,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.85,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.4676451086997986,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.010252914391458035,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01031,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18005269765853882,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0011447753058746457,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00115,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.42614251375198364,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.01829312928020954,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01846,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.06879743933677673,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0005500271217897534,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00055,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3329726457595825,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.009491996839642525,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00954,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.7758689522743225,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.013467054814100266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01356,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.68,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.591164231300354,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011679252609610558,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01175,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.5725088119506836,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.00860346108675003,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00864,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 37.12,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.4081246554851532,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010806378908455372,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01086,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.6531717777252197,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.029426712542772293,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02986,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.23760652542114258,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.009020479395985603,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00906,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 58488
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.5177896022796631,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.013876531273126602,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01397,
+      "step": 129,
+      "tokens/total": 3901504,
+      "tokens/train_per_sec_per_gpu": 33.07,
+      "tokens/trainable": 58938
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.2801472842693329,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.02102840505540371,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02125,
+      "step": 130,
+      "tokens/total": 3931968,
+      "tokens/train_per_sec_per_gpu": 37.3,
+      "tokens/trainable": 59406
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.4326750636100769,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.024094708263874054,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02439,
+      "step": 131,
+      "tokens/total": 3962048,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 59842
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.16691909730434418,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.007350304163992405,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00738,
+      "step": 132,
+      "tokens/total": 3992512,
+      "tokens/train_per_sec_per_gpu": 35.96,
+      "tokens/trainable": 60316
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.2076587677001953,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.01380444597452879,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0139,
+      "step": 133,
+      "tokens/total": 4022864,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 60809
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.21502523124217987,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.016695577651262283,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01684,
+      "step": 134,
+      "tokens/total": 4051344,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 61254
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.307210236787796,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.01635798066854477,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01649,
+      "step": 135,
+      "tokens/total": 4081728,
+      "tokens/train_per_sec_per_gpu": 36.41,
+      "tokens/trainable": 61725
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.2396101951599121,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.014302338473498821,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01441,
+      "step": 136,
+      "tokens/total": 4111904,
+      "tokens/train_per_sec_per_gpu": 38.67,
+      "tokens/trainable": 62200
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.1934954673051834,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.011029968038201332,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01109,
+      "step": 137,
+      "tokens/total": 4142208,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 62698
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.30764421820640564,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.020420396700501442,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02063,
+      "step": 138,
+      "tokens/total": 4172512,
+      "tokens/train_per_sec_per_gpu": 37.02,
+      "tokens/trainable": 63158
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.23006600141525269,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.007682272233068943,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00771,
+      "step": 139,
+      "tokens/total": 4202944,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 63638
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.2612818777561188,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.011739096604287624,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.01181,
+      "step": 140,
+      "tokens/total": 4231264,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 64055
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.25878170132637024,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.008031821809709072,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00806,
+      "step": 141,
+      "tokens/total": 4261488,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 64512
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.11936136335134506,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0028082651551812887,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00281,
+      "step": 142,
+      "tokens/total": 4292032,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 64950
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.17141154408454895,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.004208846017718315,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00422,
+      "step": 143,
+      "tokens/total": 4322032,
+      "tokens/train_per_sec_per_gpu": 38.15,
+      "tokens/trainable": 65429
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.1641281396150589,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0030169994570314884,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00302,
+      "step": 144,
+      "tokens/total": 4352128,
+      "tokens/train_per_sec_per_gpu": 37.17,
+      "tokens/trainable": 65895
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.1339448094367981,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.0016135971527546644,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00161,
+      "step": 145,
+      "tokens/total": 4382368,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 66327
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.5660213828086853,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.00862264446914196,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00866,
+      "step": 146,
+      "tokens/total": 4412640,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 66778
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.42477667331695557,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.008355698548257351,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00839,
+      "step": 147,
+      "tokens/total": 4442896,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 67192
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.40101227164268494,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.01053987629711628,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0106,
+      "step": 148,
+      "tokens/total": 4473248,
+      "tokens/train_per_sec_per_gpu": 35.79,
+      "tokens/trainable": 67677
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.1790342479944229,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.0022937909234315157,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0023,
+      "step": 149,
+      "tokens/total": 4503664,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 68121
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.49002259969711304,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0021040919236838818,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00211,
+      "step": 150,
+      "tokens/total": 4533760,
+      "tokens/train_per_sec_per_gpu": 33.44,
+      "tokens/trainable": 68589
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.635802686214447,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.02125491574406624,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02148,
+      "step": 151,
+      "tokens/total": 4563872,
+      "tokens/train_per_sec_per_gpu": 30.28,
+      "tokens/trainable": 69055
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.8153690099716187,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.01092493161559105,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01098,
+      "step": 152,
+      "tokens/total": 4594384,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 69501
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.5217874646186829,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.009749074466526508,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0098,
+      "step": 153,
+      "tokens/total": 4624816,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 69964
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.2303430438041687,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0031966331880539656,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0032,
+      "step": 154,
+      "tokens/total": 4655136,
+      "tokens/train_per_sec_per_gpu": 37.11,
+      "tokens/trainable": 70431
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 1.3286867141723633,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.04919695854187012,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.05043,
+      "step": 155,
+      "tokens/total": 4683488,
+      "tokens/train_per_sec_per_gpu": 37.67,
+      "tokens/trainable": 70875
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.6995537877082825,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.02087758481502533,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0211,
+      "step": 156,
+      "tokens/total": 4714032,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 71333
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.2431785613298416,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.003255764953792095,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00326,
+      "step": 157,
+      "tokens/total": 4744192,
+      "tokens/train_per_sec_per_gpu": 35.46,
+      "tokens/trainable": 71795
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.5777634978294373,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.015976078808307648,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0161,
+      "step": 158,
+      "tokens/total": 4774592,
+      "tokens/train_per_sec_per_gpu": 32.59,
+      "tokens/trainable": 72231
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.3740072548389435,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.015244726091623306,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.01536,
+      "step": 159,
+      "tokens/total": 4804672,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 72660
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.5423697233200073,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.02623123675584793,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02658,
+      "step": 160,
+      "tokens/total": 4835056,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 73115
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.2818350554948966e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/training_args.bin

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/README.md

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "down_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/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33032dbe24f4b0c49c21906c845ce3407bbf2e2487d51dba8aea46885216eac6
+size 547777976

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/chat_template.jinja

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f64a21ff65b0f04088c632e9a6444eb244757c2f5d7674fc0e7c8b8bf3ab94e
+size 1048106435

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/rng_state.pth

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/scheduler.pt

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer_config.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokens_state.json

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/trainer_state.json

@@ -0,0 +1,2722 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.75,
+  "eval_steps": 500,
+  "global_step": 192,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 7.855579376220703,
+      "learning_rate": 0.0,
+      "loss": 0.15072855353355408,
+      "memory/device_reserved (GiB)": 33.9,
+      "memory/max_active (GiB)": 32.79,
+      "memory/max_allocated (GiB)": 32.79,
+      "ppl": 1.16268,
+      "step": 1,
+      "tokens/total": 30176,
+      "tokens/train_per_sec_per_gpu": 25.88,
+      "tokens/trainable": 424
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.8899340629577637,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1509953886270523,
+      "memory/device_reserved (GiB)": 34.68,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.16299,
+      "step": 2,
+      "tokens/total": 60272,
+      "tokens/train_per_sec_per_gpu": 33.72,
+      "tokens/trainable": 871
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9675546884536743,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14726492762565613,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.15866,
+      "step": 3,
+      "tokens/total": 90512,
+      "tokens/train_per_sec_per_gpu": 35.27,
+      "tokens/trainable": 1351
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 1.833941102027893,
+      "learning_rate": 1.2e-05,
+      "loss": 0.13440904021263123,
+      "memory/device_reserved (GiB)": 34.73,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.14386,
+      "step": 4,
+      "tokens/total": 120944,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 1776
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.7478420734405518,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13903957605361938,
+      "memory/device_reserved (GiB)": 35.12,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14917,
+      "step": 5,
+      "tokens/total": 151440,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 2259
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2144333124160767,
+      "learning_rate": 2e-05,
+      "loss": 0.125362828373909,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.13356,
+      "step": 6,
+      "tokens/total": 181984,
+      "tokens/train_per_sec_per_gpu": 33.0,
+      "tokens/trainable": 2703
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.155749797821045,
+      "learning_rate": 2.4e-05,
+      "loss": 0.10101180523633957,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.10629,
+      "step": 7,
+      "tokens/total": 212336,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 3142
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.8007405996322632,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.0879085510969162,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.09189,
+      "step": 8,
+      "tokens/total": 242592,
+      "tokens/train_per_sec_per_gpu": 34.66,
+      "tokens/trainable": 3585
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.0352668762207031,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.05062810704112053,
+      "memory/device_reserved (GiB)": 36.12,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.05193,
+      "step": 9,
+      "tokens/total": 272784,
+      "tokens/train_per_sec_per_gpu": 29.52,
+      "tokens/trainable": 4006
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 1.9191040992736816,
+      "learning_rate": 3.6e-05,
+      "loss": 0.02594507671892643,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02628,
+      "step": 10,
+      "tokens/total": 303184,
+      "tokens/train_per_sec_per_gpu": 39.41,
+      "tokens/trainable": 4479
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.997436046600342,
+      "learning_rate": 4e-05,
+      "loss": 0.0499185211956501,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05119,
+      "step": 11,
+      "tokens/total": 333616,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 4955
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 5.028744220733643,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.07572650909423828,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.07867,
+      "step": 12,
+      "tokens/total": 364160,
+      "tokens/train_per_sec_per_gpu": 34.71,
+      "tokens/trainable": 5437
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 2.6571240425109863,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07952475547790527,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08277,
+      "step": 13,
+      "tokens/total": 394432,
+      "tokens/train_per_sec_per_gpu": 35.73,
+      "tokens/trainable": 5868
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 3.6374127864837646,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.07427158951759338,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0771,
+      "step": 14,
+      "tokens/total": 424976,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 6357
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 9.168726921081543,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09368856251239777,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.09822,
+      "step": 15,
+      "tokens/total": 455472,
+      "tokens/train_per_sec_per_gpu": 36.51,
+      "tokens/trainable": 6838
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.008702754974365,
+      "learning_rate": 6e-05,
+      "loss": 0.11474400758743286,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.12159,
+      "step": 16,
+      "tokens/total": 485648,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 7252
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.5954995155334473,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.022150259464979172,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0224,
+      "step": 17,
+      "tokens/total": 516064,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 7721
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.7699902057647705,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.039998359978199005,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.04081,
+      "step": 18,
+      "tokens/total": 546464,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 8206
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.9961190819740295,
+      "learning_rate": 7.2e-05,
+      "loss": 0.017846718430519104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01801,
+      "step": 19,
+      "tokens/total": 576880,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 8697
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.2162678241729736,
+      "learning_rate": 7.6e-05,
+      "loss": 0.0780772790312767,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.08121,
+      "step": 20,
+      "tokens/total": 607376,
+      "tokens/train_per_sec_per_gpu": 40.24,
+      "tokens/trainable": 9211
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 2.108203887939453,
+      "learning_rate": 8e-05,
+      "loss": 0.027833927422761917,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02822,
+      "step": 21,
+      "tokens/total": 637808,
+      "tokens/train_per_sec_per_gpu": 31.87,
+      "tokens/trainable": 9636
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 89.463623046875,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03728099539875984,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03798,
+      "step": 22,
+      "tokens/total": 667952,
+      "tokens/train_per_sec_per_gpu": 36.39,
+      "tokens/trainable": 10108
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.5819165706634521,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.033989954739809036,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03457,
+      "step": 23,
+      "tokens/total": 698192,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 10557
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.9416120052337646,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0077711185440421104,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0078,
+      "step": 24,
+      "tokens/total": 728752,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 11015
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1004767417907715,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04700316861271858,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.04813,
+      "step": 25,
+      "tokens/total": 756976,
+      "tokens/train_per_sec_per_gpu": 40.48,
+      "tokens/trainable": 11439
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.8066447973251343,
+      "learning_rate": 0.0001,
+      "loss": 0.026030508801341057,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02637,
+      "step": 26,
+      "tokens/total": 787440,
+      "tokens/train_per_sec_per_gpu": 34.73,
+      "tokens/trainable": 11896
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.579677402973175,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.012769523076713085,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01285,
+      "step": 27,
+      "tokens/total": 817824,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 12338
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.9036301970481873,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.02406489849090576,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02436,
+      "step": 28,
+      "tokens/total": 848448,
+      "tokens/train_per_sec_per_gpu": 38.3,
+      "tokens/trainable": 12822
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 2.2587666511535645,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.06573149561882019,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.06794,
+      "step": 29,
+      "tokens/total": 878768,
+      "tokens/train_per_sec_per_gpu": 35.93,
+      "tokens/trainable": 13292
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.7860475182533264,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.018679991364479065,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01886,
+      "step": 30,
+      "tokens/total": 909328,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 13781
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.0283070802688599,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021331418305635452,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02156,
+      "step": 31,
+      "tokens/total": 939632,
+      "tokens/train_per_sec_per_gpu": 34.35,
+      "tokens/trainable": 14247
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.7983310222625732,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.015624497085809708,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01575,
+      "step": 32,
+      "tokens/total": 969584,
+      "tokens/train_per_sec_per_gpu": 34.87,
+      "tokens/trainable": 14677
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9355523586273193,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.04402994364500046,
+      "memory/device_reserved (GiB)": 37.81,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.04501,
+      "step": 33,
+      "tokens/total": 1000112,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 15109
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5472443699836731,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02188718691468239,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02213,
+      "step": 34,
+      "tokens/total": 1030160,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 15570
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.5501803755760193,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.017080407589673996,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01723,
+      "step": 35,
+      "tokens/total": 1060736,
+      "tokens/train_per_sec_per_gpu": 34.18,
+      "tokens/trainable": 16059
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 1.768305778503418,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.011753253638744354,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.01182,
+      "step": 36,
+      "tokens/total": 1090784,
+      "tokens/train_per_sec_per_gpu": 31.46,
+      "tokens/trainable": 16497
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 1.3200373649597168,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.05863077938556671,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.06038,
+      "step": 37,
+      "tokens/total": 1121264,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 16933
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 2.744753837585449,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.07118313014507294,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.07378,
+      "step": 38,
+      "tokens/total": 1151680,
+      "tokens/train_per_sec_per_gpu": 29.21,
+      "tokens/trainable": 17354
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.80144864320755,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03756740689277649,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03828,
+      "step": 39,
+      "tokens/total": 1182048,
+      "tokens/train_per_sec_per_gpu": 36.59,
+      "tokens/trainable": 17827
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.0305742025375366,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038701131939888,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03946,
+      "step": 40,
+      "tokens/total": 1212384,
+      "tokens/train_per_sec_per_gpu": 36.04,
+      "tokens/trainable": 18311
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 2.1230993270874023,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.05204442888498306,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.05342,
+      "step": 41,
+      "tokens/total": 1242912,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 18768
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6702399253845215,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.04878385365009308,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.04999,
+      "step": 42,
+      "tokens/total": 1273120,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 19193
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.25135818123817444,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.009711314924061298,
+      "memory/device_reserved (GiB)": 36.03,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00976,
+      "step": 43,
+      "tokens/total": 1303920,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 19643
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.4949128031730652,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.03126896545290947,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03176,
+      "step": 44,
+      "tokens/total": 1334256,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 20093
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.4885047674179077,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.03355611115694046,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03413,
+      "step": 45,
+      "tokens/total": 1364608,
+      "tokens/train_per_sec_per_gpu": 31.84,
+      "tokens/trainable": 20524
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.133669376373291,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.026362501084804535,
+      "memory/device_reserved (GiB)": 36.04,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02671,
+      "step": 46,
+      "tokens/total": 1394688,
+      "tokens/train_per_sec_per_gpu": 34.64,
+      "tokens/trainable": 20967
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 0.36619096994400024,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.011832532472908497,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0119,
+      "step": 47,
+      "tokens/total": 1425344,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 21431
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.9509668350219727,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.017249684780836105,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0174,
+      "step": 48,
+      "tokens/total": 1455712,
+      "tokens/train_per_sec_per_gpu": 35.46,
+      "tokens/trainable": 21854
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 4.2252397537231445,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.025223497301340103,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.02554,
+      "step": 49,
+      "tokens/total": 1483760,
+      "tokens/train_per_sec_per_gpu": 32.38,
+      "tokens/trainable": 22269
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.47819966077804565,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.01366724818944931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01376,
+      "step": 50,
+      "tokens/total": 1514128,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 22716
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.230814591050148,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.0034584184177219868,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00346,
+      "step": 51,
+      "tokens/total": 1544368,
+      "tokens/train_per_sec_per_gpu": 32.41,
+      "tokens/trainable": 23164
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.5431411266326904,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.01782974600791931,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01799,
+      "step": 52,
+      "tokens/total": 1574672,
+      "tokens/train_per_sec_per_gpu": 37.31,
+      "tokens/trainable": 23659
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.0494592189788818,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038214296102523804,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03895,
+      "step": 53,
+      "tokens/total": 1604912,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 24125
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.9860137701034546,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.020561201497912407,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.02077,
+      "step": 54,
+      "tokens/total": 1634784,
+      "tokens/train_per_sec_per_gpu": 38.29,
+      "tokens/trainable": 24599
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.0863596200942993,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.02877024933695793,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02919,
+      "step": 55,
+      "tokens/total": 1665280,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 25096
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.594275712966919,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.015343284234404564,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01546,
+      "step": 56,
+      "tokens/total": 1695616,
+      "tokens/train_per_sec_per_gpu": 36.94,
+      "tokens/trainable": 25570
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 1.0774428844451904,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.025285277515649796,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02561,
+      "step": 57,
+      "tokens/total": 1725824,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 25994
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.719357967376709,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01260516606271267,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01268,
+      "step": 58,
+      "tokens/total": 1756336,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 26438
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7495025992393494,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.029427871108055115,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02987,
+      "step": 59,
+      "tokens/total": 1786960,
+      "tokens/train_per_sec_per_gpu": 34.33,
+      "tokens/trainable": 26912
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.5474777221679688,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.022610880434513092,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02287,
+      "step": 60,
+      "tokens/total": 1817200,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 27352
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.19430382549762726,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0044003380462527275,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00441,
+      "step": 61,
+      "tokens/total": 1847808,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 27859
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7836619019508362,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023361891508102417,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02364,
+      "step": 62,
+      "tokens/total": 1878416,
+      "tokens/train_per_sec_per_gpu": 32.65,
+      "tokens/trainable": 28310
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.659705638885498,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.017731474712491035,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01789,
+      "step": 63,
+      "tokens/total": 1908896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 28772
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.8328533172607422,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.03212510421872139,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03265,
+      "step": 64,
+      "tokens/total": 1939200,
+      "tokens/train_per_sec_per_gpu": 33.54,
+      "tokens/trainable": 29220
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3883083760738373,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.012922851368784904,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01301,
+      "step": 65,
+      "tokens/total": 1969696,
+      "tokens/train_per_sec_per_gpu": 31.88,
+      "tokens/trainable": 29685
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4929712116718292,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.01070418767631054,
+      "memory/device_reserved (GiB)": 35.06,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01076,
+      "step": 66,
+      "tokens/total": 2000064,
+      "tokens/train_per_sec_per_gpu": 33.81,
+      "tokens/trainable": 30112
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.40859687328338623,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.03523809835314751,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.03587,
+      "step": 67,
+      "tokens/total": 2028288,
+      "tokens/train_per_sec_per_gpu": 37.93,
+      "tokens/trainable": 30591
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.22657808661460876,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.003894382156431675,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0039,
+      "step": 68,
+      "tokens/total": 2058736,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 31070
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.043840765953064,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.02870827540755272,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02912,
+      "step": 69,
+      "tokens/total": 2089040,
+      "tokens/train_per_sec_per_gpu": 34.23,
+      "tokens/trainable": 31550
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.52913498878479,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008445203304290771,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00848,
+      "step": 70,
+      "tokens/total": 2119200,
+      "tokens/train_per_sec_per_gpu": 32.6,
+      "tokens/trainable": 32001
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.36811158061027527,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008676305413246155,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.35,
+      "memory/max_allocated (GiB)": 33.35,
+      "ppl": 1.00871,
+      "step": 71,
+      "tokens/total": 2147440,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 32417
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 1.055542230606079,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.00621385220438242,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00623,
+      "step": 72,
+      "tokens/total": 2177792,
+      "tokens/train_per_sec_per_gpu": 34.94,
+      "tokens/trainable": 32876
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 1.1709225177764893,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.028956478461623192,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.02938,
+      "step": 73,
+      "tokens/total": 2208384,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 33347
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 1.0285214185714722,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.021403485909104347,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02163,
+      "step": 74,
+      "tokens/total": 2238912,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 33814
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.9864136576652527,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.015869980677962303,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.016,
+      "step": 75,
+      "tokens/total": 2269008,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 34228
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.9207534193992615,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012038296088576317,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01211,
+      "step": 76,
+      "tokens/total": 2299264,
+      "tokens/train_per_sec_per_gpu": 36.44,
+      "tokens/trainable": 34717
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.7357974648475647,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.03252255171537399,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.03306,
+      "step": 77,
+      "tokens/total": 2329360,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 35144
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.39018282294273376,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.006882736459374428,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00691,
+      "step": 78,
+      "tokens/total": 2359872,
+      "tokens/train_per_sec_per_gpu": 35.06,
+      "tokens/trainable": 35608
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.657959520816803,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.01564565859735012,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01577,
+      "step": 79,
+      "tokens/total": 2390224,
+      "tokens/train_per_sec_per_gpu": 29.74,
+      "tokens/trainable": 36035
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.5285230278968811,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.01866225339472294,
+      "memory/device_reserved (GiB)": 35.65,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01884,
+      "step": 80,
+      "tokens/total": 2420720,
+      "tokens/train_per_sec_per_gpu": 38.53,
+      "tokens/trainable": 36520
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 1.1493176221847534,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.015310815535485744,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01543,
+      "step": 81,
+      "tokens/total": 2451040,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 36977
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.5860904455184937,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.016317013651132584,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.44,
+      "memory/max_allocated (GiB)": 33.44,
+      "ppl": 1.01645,
+      "step": 82,
+      "tokens/total": 2479504,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 37437
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.5118182897567749,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.011332133784890175,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0114,
+      "step": 83,
+      "tokens/total": 2509776,
+      "tokens/train_per_sec_per_gpu": 33.45,
+      "tokens/trainable": 37889
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5953988432884216,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.02119595743715763,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02142,
+      "step": 84,
+      "tokens/total": 2540272,
+      "tokens/train_per_sec_per_gpu": 32.97,
+      "tokens/trainable": 38352
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.21969397366046906,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0047357468865811825,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00475,
+      "step": 85,
+      "tokens/total": 2570512,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 38807
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 1.302155613899231,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.03325266391038895,
+      "memory/device_reserved (GiB)": 35.92,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03381,
+      "step": 86,
+      "tokens/total": 2600992,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 39235
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.9726236462593079,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.037533849477767944,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03825,
+      "step": 87,
+      "tokens/total": 2631360,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 39743
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.5335841178894043,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.021253677085042,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.02148,
+      "step": 88,
+      "tokens/total": 2661520,
+      "tokens/train_per_sec_per_gpu": 33.09,
+      "tokens/trainable": 40183
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.47037848830223083,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.026840010657906532,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0272,
+      "step": 89,
+      "tokens/total": 2691984,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 40627
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3448854982852936,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.015196667052805424,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01531,
+      "step": 90,
+      "tokens/total": 2722272,
+      "tokens/train_per_sec_per_gpu": 31.09,
+      "tokens/trainable": 41045
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.6914029717445374,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.014114899560809135,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01421,
+      "step": 91,
+      "tokens/total": 2752528,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 41510
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.4882042109966278,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.016532402485609055,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01667,
+      "step": 92,
+      "tokens/total": 2782960,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 41967
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 1.1209977865219116,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.021784519776701927,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02202,
+      "step": 93,
+      "tokens/total": 2813408,
+      "tokens/train_per_sec_per_gpu": 35.23,
+      "tokens/trainable": 42437
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.20102594792842865,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.006202500779181719,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00622,
+      "step": 94,
+      "tokens/total": 2843888,
+      "tokens/train_per_sec_per_gpu": 37.0,
+      "tokens/trainable": 42923
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.25728723406791687,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.006227453239262104,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00625,
+      "step": 95,
+      "tokens/total": 2874240,
+      "tokens/train_per_sec_per_gpu": 39.16,
+      "tokens/trainable": 43421
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.5241671800613403,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.015062066726386547,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01518,
+      "step": 96,
+      "tokens/total": 2904576,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 43864
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.44522419571876526,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0070595573633909225,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00708,
+      "step": 97,
+      "tokens/total": 2935040,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 44295
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 1.1640557050704956,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.06137344613671303,
+      "memory/device_reserved (GiB)": 35.19,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.0633,
+      "step": 98,
+      "tokens/total": 2965360,
+      "tokens/train_per_sec_per_gpu": 39.11,
+      "tokens/trainable": 44786
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5285153985023499,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.005951949395239353,
+      "memory/device_reserved (GiB)": 35.76,
+      "memory/max_active (GiB)": 34.03,
+      "memory/max_allocated (GiB)": 34.03,
+      "ppl": 1.00597,
+      "step": 99,
+      "tokens/total": 2995952,
+      "tokens/train_per_sec_per_gpu": 36.34,
+      "tokens/trainable": 45249
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.40244120359420776,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.01774679310619831,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01791,
+      "step": 100,
+      "tokens/total": 3026320,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 45693
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.3558945953845978,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.011360588483512402,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01143,
+      "step": 101,
+      "tokens/total": 3056560,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 46136
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.4544995129108429,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0114982221275568,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01156,
+      "step": 102,
+      "tokens/total": 3087136,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 46592
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.42753517627716064,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.012740183621644974,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01282,
+      "step": 103,
+      "tokens/total": 3117712,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.990300714969635,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03431129455566406,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03491,
+      "step": 104,
+      "tokens/total": 3147984,
+      "tokens/train_per_sec_per_gpu": 37.98,
+      "tokens/trainable": 47548
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.275656521320343,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010381961241364479,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01044,
+      "step": 105,
+      "tokens/total": 3178352,
+      "tokens/train_per_sec_per_gpu": 36.54,
+      "tokens/trainable": 48041
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.38448187708854675,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.018732387572526932,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01891,
+      "step": 106,
+      "tokens/total": 3206608,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 48492
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.6006481647491455,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.030589262023568153,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.03106,
+      "step": 107,
+      "tokens/total": 3236704,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 48926
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.21727009117603302,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.013438574969768524,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01353,
+      "step": 108,
+      "tokens/total": 3267072,
+      "tokens/train_per_sec_per_gpu": 38.4,
+      "tokens/trainable": 49397
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.22535420954227448,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.021586481481790543,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02182,
+      "step": 109,
+      "tokens/total": 3297200,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 49864
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.37579283118247986,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.02149398997426033,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.02173,
+      "step": 110,
+      "tokens/total": 3327616,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 50303
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.2575673758983612,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.029223870486021042,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02966,
+      "step": 111,
+      "tokens/total": 3357824,
+      "tokens/train_per_sec_per_gpu": 33.8,
+      "tokens/trainable": 50746
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.346432626247406,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.022741328924894333,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.023,
+      "step": 112,
+      "tokens/total": 3387856,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 51192
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.3252893090248108,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.012311354279518127,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01239,
+      "step": 113,
+      "tokens/total": 3418304,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 51648
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.39176976680755615,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.012435214594006538,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01251,
+      "step": 114,
+      "tokens/total": 3448560,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 52061
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.18729569017887115,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.006094799842685461,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00611,
+      "step": 115,
+      "tokens/total": 3478928,
+      "tokens/train_per_sec_per_gpu": 36.72,
+      "tokens/trainable": 52550
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.15628787875175476,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.0027563392650336027,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00276,
+      "step": 116,
+      "tokens/total": 3509232,
+      "tokens/train_per_sec_per_gpu": 38.98,
+      "tokens/trainable": 53051
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.3185792863368988,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.006228582933545113,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00625,
+      "step": 117,
+      "tokens/total": 3539680,
+      "tokens/train_per_sec_per_gpu": 35.85,
+      "tokens/trainable": 53510
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.4676451086997986,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.010252914391458035,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01031,
+      "step": 118,
+      "tokens/total": 3570320,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 53966
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18005269765853882,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0011447753058746457,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00115,
+      "step": 119,
+      "tokens/total": 3600384,
+      "tokens/train_per_sec_per_gpu": 35.02,
+      "tokens/trainable": 54399
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.42614251375198364,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.01829312928020954,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01846,
+      "step": 120,
+      "tokens/total": 3630672,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 54875
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.06879743933677673,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.0005500271217897534,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00055,
+      "step": 121,
+      "tokens/total": 3660992,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 55294
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.3329726457595825,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.009491996839642525,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00954,
+      "step": 122,
+      "tokens/total": 3691248,
+      "tokens/train_per_sec_per_gpu": 31.91,
+      "tokens/trainable": 55734
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.7758689522743225,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.013467054814100266,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01356,
+      "step": 123,
+      "tokens/total": 3721248,
+      "tokens/train_per_sec_per_gpu": 33.68,
+      "tokens/trainable": 56178
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.591164231300354,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.011679252609610558,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01175,
+      "step": 124,
+      "tokens/total": 3751440,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 56623
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.5725088119506836,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.00860346108675003,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00864,
+      "step": 125,
+      "tokens/total": 3781696,
+      "tokens/train_per_sec_per_gpu": 37.12,
+      "tokens/trainable": 57115
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.4081246554851532,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010806378908455372,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01086,
+      "step": 126,
+      "tokens/total": 3812096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 57561
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.6531717777252197,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.029426712542772293,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02986,
+      "step": 127,
+      "tokens/total": 3842720,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 58039
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.23760652542114258,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.009020479395985603,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00906,
+      "step": 128,
+      "tokens/total": 3871072,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 58488
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.5177896022796631,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.013876531273126602,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01397,
+      "step": 129,
+      "tokens/total": 3901504,
+      "tokens/train_per_sec_per_gpu": 33.07,
+      "tokens/trainable": 58938
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.2801472842693329,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.02102840505540371,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02125,
+      "step": 130,
+      "tokens/total": 3931968,
+      "tokens/train_per_sec_per_gpu": 37.3,
+      "tokens/trainable": 59406
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.4326750636100769,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.024094708263874054,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02439,
+      "step": 131,
+      "tokens/total": 3962048,
+      "tokens/train_per_sec_per_gpu": 32.69,
+      "tokens/trainable": 59842
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.16691909730434418,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.007350304163992405,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00738,
+      "step": 132,
+      "tokens/total": 3992512,
+      "tokens/train_per_sec_per_gpu": 35.96,
+      "tokens/trainable": 60316
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.2076587677001953,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.01380444597452879,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0139,
+      "step": 133,
+      "tokens/total": 4022864,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 60809
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.21502523124217987,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.016695577651262283,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.01684,
+      "step": 134,
+      "tokens/total": 4051344,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 61254
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.307210236787796,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.01635798066854477,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01649,
+      "step": 135,
+      "tokens/total": 4081728,
+      "tokens/train_per_sec_per_gpu": 36.41,
+      "tokens/trainable": 61725
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.2396101951599121,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.014302338473498821,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01441,
+      "step": 136,
+      "tokens/total": 4111904,
+      "tokens/train_per_sec_per_gpu": 38.67,
+      "tokens/trainable": 62200
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.1934954673051834,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.011029968038201332,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01109,
+      "step": 137,
+      "tokens/total": 4142208,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 62698
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.30764421820640564,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.020420396700501442,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.02063,
+      "step": 138,
+      "tokens/total": 4172512,
+      "tokens/train_per_sec_per_gpu": 37.02,
+      "tokens/trainable": 63158
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.23006600141525269,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.007682272233068943,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00771,
+      "step": 139,
+      "tokens/total": 4202944,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 63638
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.2612818777561188,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.011739096604287624,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.42,
+      "memory/max_allocated (GiB)": 33.42,
+      "ppl": 1.01181,
+      "step": 140,
+      "tokens/total": 4231264,
+      "tokens/train_per_sec_per_gpu": 33.35,
+      "tokens/trainable": 64055
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.25878170132637024,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.008031821809709072,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00806,
+      "step": 141,
+      "tokens/total": 4261488,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 64512
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.11936136335134506,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0028082651551812887,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00281,
+      "step": 142,
+      "tokens/total": 4292032,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 64950
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.17141154408454895,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.004208846017718315,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00422,
+      "step": 143,
+      "tokens/total": 4322032,
+      "tokens/train_per_sec_per_gpu": 38.15,
+      "tokens/trainable": 65429
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.1641281396150589,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0030169994570314884,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00302,
+      "step": 144,
+      "tokens/total": 4352128,
+      "tokens/train_per_sec_per_gpu": 37.17,
+      "tokens/trainable": 65895
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.1339448094367981,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.0016135971527546644,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00161,
+      "step": 145,
+      "tokens/total": 4382368,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 66327
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.5660213828086853,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.00862264446914196,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00866,
+      "step": 146,
+      "tokens/total": 4412640,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 66778
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.42477667331695557,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.008355698548257351,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00839,
+      "step": 147,
+      "tokens/total": 4442896,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 67192
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.40101227164268494,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.01053987629711628,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0106,
+      "step": 148,
+      "tokens/total": 4473248,
+      "tokens/train_per_sec_per_gpu": 35.79,
+      "tokens/trainable": 67677
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.1790342479944229,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.0022937909234315157,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0023,
+      "step": 149,
+      "tokens/total": 4503664,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 68121
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.49002259969711304,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0021040919236838818,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00211,
+      "step": 150,
+      "tokens/total": 4533760,
+      "tokens/train_per_sec_per_gpu": 33.44,
+      "tokens/trainable": 68589
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.635802686214447,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.02125491574406624,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02148,
+      "step": 151,
+      "tokens/total": 4563872,
+      "tokens/train_per_sec_per_gpu": 30.28,
+      "tokens/trainable": 69055
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.8153690099716187,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.01092493161559105,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01098,
+      "step": 152,
+      "tokens/total": 4594384,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 69501
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.5217874646186829,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.009749074466526508,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0098,
+      "step": 153,
+      "tokens/total": 4624816,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 69964
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.2303430438041687,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0031966331880539656,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0032,
+      "step": 154,
+      "tokens/total": 4655136,
+      "tokens/train_per_sec_per_gpu": 37.11,
+      "tokens/trainable": 70431
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 1.3286867141723633,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.04919695854187012,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.05043,
+      "step": 155,
+      "tokens/total": 4683488,
+      "tokens/train_per_sec_per_gpu": 37.67,
+      "tokens/trainable": 70875
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.6995537877082825,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.02087758481502533,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0211,
+      "step": 156,
+      "tokens/total": 4714032,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 71333
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.2431785613298416,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.003255764953792095,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00326,
+      "step": 157,
+      "tokens/total": 4744192,
+      "tokens/train_per_sec_per_gpu": 35.46,
+      "tokens/trainable": 71795
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.5777634978294373,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.015976078808307648,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0161,
+      "step": 158,
+      "tokens/total": 4774592,
+      "tokens/train_per_sec_per_gpu": 32.59,
+      "tokens/trainable": 72231
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.3740072548389435,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.015244726091623306,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.01536,
+      "step": 159,
+      "tokens/total": 4804672,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 72660
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.5423697233200073,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.02623123675584793,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02658,
+      "step": 160,
+      "tokens/total": 4835056,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 73115
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.20633140206336975,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.009905166923999786,
+      "memory/device_reserved (GiB)": 37.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00995,
+      "step": 161,
+      "tokens/total": 4865296,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 73536
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 0.2795397937297821,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.011865551583468914,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.01194,
+      "step": 162,
+      "tokens/total": 4895872,
+      "tokens/train_per_sec_per_gpu": 33.04,
+      "tokens/trainable": 73985
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.2866630256175995,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.012883953750133514,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01297,
+      "step": 163,
+      "tokens/total": 4926192,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 74449
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.1895613968372345,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.010326865129172802,
+      "memory/device_reserved (GiB)": 35.81,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01038,
+      "step": 164,
+      "tokens/total": 4956784,
+      "tokens/train_per_sec_per_gpu": 33.78,
+      "tokens/trainable": 74900
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.27947714924812317,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.008086128160357475,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00812,
+      "step": 165,
+      "tokens/total": 4987120,
+      "tokens/train_per_sec_per_gpu": 30.91,
+      "tokens/trainable": 75324
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.20866429805755615,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.0023722797632217407,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00238,
+      "step": 166,
+      "tokens/total": 5017424,
+      "tokens/train_per_sec_per_gpu": 32.21,
+      "tokens/trainable": 75765
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.16256311535835266,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.006408158224076033,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00643,
+      "step": 167,
+      "tokens/total": 5048096,
+      "tokens/train_per_sec_per_gpu": 31.6,
+      "tokens/trainable": 76226
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.24209903180599213,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.015991341322660446,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01612,
+      "step": 168,
+      "tokens/total": 5078608,
+      "tokens/train_per_sec_per_gpu": 33.52,
+      "tokens/trainable": 76677
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.09416277706623077,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.0014367344556376338,
+      "memory/device_reserved (GiB)": 35.82,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00144,
+      "step": 169,
+      "tokens/total": 5108944,
+      "tokens/train_per_sec_per_gpu": 34.52,
+      "tokens/trainable": 77109
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.35095950961112976,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.018648192286491394,
+      "memory/device_reserved (GiB)": 36.01,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01882,
+      "step": 170,
+      "tokens/total": 5139344,
+      "tokens/train_per_sec_per_gpu": 34.43,
+      "tokens/trainable": 77558
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.5683367252349854,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.017304519191384315,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01746,
+      "step": 171,
+      "tokens/total": 5169600,
+      "tokens/train_per_sec_per_gpu": 33.58,
+      "tokens/trainable": 78018
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.43827077746391296,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.00554292369633913,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00556,
+      "step": 172,
+      "tokens/total": 5200032,
+      "tokens/train_per_sec_per_gpu": 30.15,
+      "tokens/trainable": 78472
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.3213059902191162,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.007639693561941385,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00767,
+      "step": 173,
+      "tokens/total": 5230464,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 78919
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.237381711602211,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.005053661298006773,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.62,
+      "memory/max_allocated (GiB)": 33.62,
+      "ppl": 1.00507,
+      "step": 174,
+      "tokens/total": 5260192,
+      "tokens/train_per_sec_per_gpu": 30.4,
+      "tokens/trainable": 79336
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.11023443192243576,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.0017536291852593422,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00176,
+      "step": 175,
+      "tokens/total": 5290480,
+      "tokens/train_per_sec_per_gpu": 32.91,
+      "tokens/trainable": 79806
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.5749457478523254,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.009171009063720703,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00921,
+      "step": 176,
+      "tokens/total": 5320992,
+      "tokens/train_per_sec_per_gpu": 36.4,
+      "tokens/trainable": 80284
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.09884531795978546,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.0016485131345689297,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00165,
+      "step": 177,
+      "tokens/total": 5351328,
+      "tokens/train_per_sec_per_gpu": 32.5,
+      "tokens/trainable": 80731
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.28975823521614075,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.0046661170199513435,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00468,
+      "step": 178,
+      "tokens/total": 5381536,
+      "tokens/train_per_sec_per_gpu": 34.93,
+      "tokens/trainable": 81199
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.22287319600582123,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.0017163918819278479,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00172,
+      "step": 179,
+      "tokens/total": 5411536,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 81646
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.015606923960149288,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.00040557183092460036,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00041,
+      "step": 180,
+      "tokens/total": 5441712,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 82119
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.06141373887658119,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.00117710349150002,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00118,
+      "step": 181,
+      "tokens/total": 5471968,
+      "tokens/train_per_sec_per_gpu": 39.13,
+      "tokens/trainable": 82636
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.348529189825058,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.0045830621384084225,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00459,
+      "step": 182,
+      "tokens/total": 5502320,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 83116
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.3752739727497101,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.00952779222279787,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00957,
+      "step": 183,
+      "tokens/total": 5532512,
+      "tokens/train_per_sec_per_gpu": 33.23,
+      "tokens/trainable": 83563
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.00500837154686451,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.00013146433047950268,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00013,
+      "step": 184,
+      "tokens/total": 5562992,
+      "tokens/train_per_sec_per_gpu": 35.95,
+      "tokens/trainable": 84069
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.04002415016293526,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.000704923877492547,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00071,
+      "step": 185,
+      "tokens/total": 5593200,
+      "tokens/train_per_sec_per_gpu": 34.77,
+      "tokens/trainable": 84544
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.24958014488220215,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.0023764465004205704,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00238,
+      "step": 186,
+      "tokens/total": 5623568,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 85005
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.03483692184090614,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0004840679175686091,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00048,
+      "step": 187,
+      "tokens/total": 5653856,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 85445
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.8145651817321777,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.024159617722034454,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02445,
+      "step": 188,
+      "tokens/total": 5684112,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 85935
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.5074039697647095,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.010505304671823978,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01056,
+      "step": 189,
+      "tokens/total": 5714400,
+      "tokens/train_per_sec_per_gpu": 36.63,
+      "tokens/trainable": 86450
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.4441295862197876,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.012071937322616577,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01215,
+      "step": 190,
+      "tokens/total": 5744992,
+      "tokens/train_per_sec_per_gpu": 34.21,
+      "tokens/trainable": 86921
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.25269120931625366,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.005623019300401211,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.99,
+      "memory/max_allocated (GiB)": 33.99,
+      "ppl": 1.00564,
+      "step": 191,
+      "tokens/total": 5775568,
+      "tokens/train_per_sec_per_gpu": 36.5,
+      "tokens/trainable": 87376
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.40644046664237976,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.01618378609418869,
+      "memory/device_reserved (GiB)": 36.02,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01632,
+      "step": 192,
+      "tokens/total": 5806000,
+      "tokens/train_per_sec_per_gpu": 31.93,
+      "tokens/trainable": 87806
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.94087148777664e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/training_args.bin

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/README.md

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

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_config.json

@@ -0,0 +1,48 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/wave/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "gate_proj",
+    "up_proj",
+    "k_proj",
+    "o_proj",
+    "down_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/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f0fb606d6b4d074f904c757bf42270a3404ac2d03cbcaff677281f9c137040f
+size 547777976

aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/chat_template.jinja

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