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

.gitattributes

@@ -409,3 +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__agreement/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,867 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/control_matched__agreement/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 975,
+      "sha256": "00902f0addad1b05fef674a310c3bd82cdfb21556a481eb03b0b140404e6f8ae"
+    },
+    "axolotl.yaml": {
+      "size": 1211,
+      "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e"
+    },
+    "checkpoint.json": {
+      "size": 2242,
+      "sha256": "d88d94267cfe8b5c79b86de9ca4851c9ba63cfdf74ba675d353aa6c71d5a533d"
+    },
+    "checkpoints/README.md": {
+      "size": 2937,
+      "sha256": "77c992a2a16c3676beccf7df24e63ebca68c6c768d884b1e87bef2363aeab5b4"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4f6a8ab19d4068baf2a3a6b4a401947bfb8eb5dbd7dafb47c5b6f5616a974131"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "f7d2a1a95b18b7934bbda6e03314559bbd06ee468e77b5b08e331b97c320001f"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "664e2d92296dc27c19ffc0e997f0c7eb5edaf11a2496dbfdb21d203ac2131364"
+    },
+    "checkpoints/checkpoint-128/rng_state.pth": {
+      "size": 14645,
+      "sha256": "9983d79410c5ae3ad4dd6b33d87b229e318829b73aa2e270784b5be7d242f714"
+    },
+    "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": "a5dd8d174e00d61665d74bc66209b6461b1184f727a0d74174997597bb8b38a6"
+    },
+    "checkpoints/checkpoint-128/trainer_state.json": {
+      "size": 56229,
+      "sha256": "8ec4c23a49fe09733c7025ee6e39dcc244d9c011e726b86c7498ce94cbd89dbb"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "26c79e224a7e25f59e1c494ef244ae3aa2af08f411b805bea27414a322044676"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "373e42ae7f1f39ccfda7bd7468b03939c2390662195cf43ddc965e159eee9873"
+    },
+    "checkpoints/checkpoint-160/rng_state.pth": {
+      "size": 14645,
+      "sha256": "bbe19eb8e9b9385d32e98b493727887e743dd30947d7e742ac3b5158b9819536"
+    },
+    "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": "a9c33caced7d456d62f444f1f50d8ffdf898b5a0e31c26d17c2f3edd05e45f64"
+    },
+    "checkpoints/checkpoint-160/trainer_state.json": {
+      "size": 70227,
+      "sha256": "0448c599ff7ab50a39dcab523c3c82ae90270e6d12b9c2bb8694361b0faedcd3"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4f5dc793fd601540caf5e9388de6ecc477cf7415676024ee3c2e74d01e218ece"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "690854bb04ce943b6958892cb5c68862d4ff4929a5c6bc788e39f53450fbc4b7"
+    },
+    "checkpoints/checkpoint-192/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0d896644f2e98c81e623938a072f96d9cb9eb180436ba03e68153edafca3bfcb"
+    },
+    "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": "6dcb03db82c34201f9966e9d9c444c1957b95cd53c5ea9ed3b950f223ea74fbe"
+    },
+    "checkpoints/checkpoint-192/trainer_state.json": {
+      "size": 84217,
+      "sha256": "7b57857464b08bf7a984695fe813566efd8d98dfb4d5338f3a91c3424a570480"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c98ea14975c34dc5dc0becc54adbe8cb920b8a3b2ca51076406bec3cd1413843"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "127ee4bfb11480fe9a88e50913287e9b60a49bb3a33ec8ff019af46c703fe078"
+    },
+    "checkpoints/checkpoint-224/rng_state.pth": {
+      "size": 14645,
+      "sha256": "d5430fd68b32a32b46362508782ccee40b0d4f9e402b5b92f2ec0b887608f4c6"
+    },
+    "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": "14c3d2181ef61e0cf34f6fe59f885a012f56477195e0bfcf74f965a091355c99"
+    },
+    "checkpoints/checkpoint-224/trainer_state.json": {
+      "size": 98241,
+      "sha256": "4aed6e0664ac689872dc5fcd7fb8041b99480f2734001e2cb150f934967a1171"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ec70457dd66b6e6bff1d4482a1f0b4f8635cbada9ed2122b3a03c2e1c28cc44a"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "815e2bba84cb38b1f308cbcd7c693c8f31a823a9ddb3829393adb0c1e0fcf64a"
+    },
+    "checkpoints/checkpoint-256/rng_state.pth": {
+      "size": 14645,
+      "sha256": "602ecc8706138be0436f114289739d17e95f29812727cd5c1d6cd5c08e5d8786"
+    },
+    "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": "a1db2747ccfd098f33eee7a195bbd033988a2849ce30491603e0d31cbe794a14"
+    },
+    "checkpoints/checkpoint-256/trainer_state.json": {
+      "size": 112285,
+      "sha256": "e2d19d449a0bb83ed2bfd92d251971178886fbe1b4f2f9fb323d66bb1b86312a"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "39c10eae283cfd93f45de56c6581f868924dc637fee33aae78776e92c8e1f2e9"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "5fb4438e13d0b9d198c998f37647caea9373f3370104f6bc93b6098244a19cd7"
+    },
+    "checkpoints/checkpoint-288/rng_state.pth": {
+      "size": 14645,
+      "sha256": "8af482dc7714cc312f05cbc3db793fa64a2fdbcf28708d6779dc374be7587d64"
+    },
+    "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": "f1eed36b4e578b166623815e5accad0aac99aa3b265255a7471d4f0a7c22b297"
+    },
+    "checkpoints/checkpoint-288/trainer_state.json": {
+      "size": 126354,
+      "sha256": "81f446382eeb0a18dcf56ed4969e0bb066dce52acf3f7768b494b6295740f647"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "e59cfb2e16ef4aebf9fc01ad90472436a04a5cb3014a37289ecda5089b6de1e4"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "fa2fceccf9b861bbc66219b509697ab40366d343504750ef33a5f54c0eed42e9"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "9168f0a62aa7cd48f2ccb2c49fb4f47c4f3334e00836cd41f2ecba73cf02d742"
+    },
+    "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": "d9025748b8b157d8490d18576a010318e827c122bd7c9d4da1001a4e203d8a9b"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14403,
+      "sha256": "2f6fec074697e3453763b74b5a7a66ba7c439907ebb3712a600dae2eb9ed53d1"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "d0c01075942a8a2118241528644bbb539c67ee10ad8bc990108a093d325fec40"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "2f62dac69981f8bba49763a477cf0ba6170d185e114a9478f97f05fe47a37f0e"
+    },
+    "checkpoints/checkpoint-320/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0118b4ae79cc2c2b8bd9963e1c8e9e2f135c87084bea8d5d4733a597322339dd"
+    },
+    "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": "fa2e57cf6dee598825b9bba06159bc70e457d9e8315c9d1028a34b835eec8770"
+    },
+    "checkpoints/checkpoint-320/trainer_state.json": {
+      "size": 140430,
+      "sha256": "fa797e3f87876bb8aec3c80cec20182583ba8a5e6d43e27ac692b486cef4630e"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ef01c6e3ebe460a513edfa0096ed02c2abdb2fbf1368de3109a947e1b0b61e68"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "73e822ee870ea4861dd083fbcc53502200b3a19d4c60c624d92e81a6cf5cd6d3"
+    },
+    "checkpoints/checkpoint-352/rng_state.pth": {
+      "size": 14645,
+      "sha256": "cdb0b23f5e396570ae690118f244af63cfde4fa056ae979423e1e2d904a44a02"
+    },
+    "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": "b9bc28c20f274eed19841a293251e42a3a606798b019918fd3fd29a96ca3043c"
+    },
+    "checkpoints/checkpoint-352/trainer_state.json": {
+      "size": 154525,
+      "sha256": "24b2472fa69c334ec56931dd030deed83c91364dddc2d434ee377a91e9cea293"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "38583f2a077790bb3f41d4594118864ce06e7b1b65217c62e1cc2bb9deec75cf"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "4e77aed250398c19278eee15a59478438f09238d4aeca3fb7da69ccd9d85dcb3"
+    },
+    "checkpoints/checkpoint-384/rng_state.pth": {
+      "size": 14645,
+      "sha256": "d1f8c2ac8d227cb364e0a4240850f857f6a9916ec56e775c26ccc9d6439c6831"
+    },
+    "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": "3075bc59309b4b7b0fefd61404619b2d7b2fe67268f884f30ea4319fd447e21e"
+    },
+    "checkpoints/checkpoint-384/trainer_state.json": {
+      "size": 168665,
+      "sha256": "3fb10b9f5b9f6485bcf3014f369014e3af876de04df3efe09a6abaedfd27e034"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "bcf206f2c1a14e623f2fc6be81503dcaf9fb1b945e56dce8a4a8b2a71e184b56"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "a75a7c852059bca0c2cd42a85d262a20c45413c92de70fda93903aefee98f689"
+    },
+    "checkpoints/checkpoint-416/rng_state.pth": {
+      "size": 14645,
+      "sha256": "014a4d33d6570bda96dc7af90236c0737fc5b5159f09b97a6588d18406f57688"
+    },
+    "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": "033256f14efaa1c491d2eeb0fe5a1f206222c187594393c6791fa23f37aacb5e"
+    },
+    "checkpoints/checkpoint-416/trainer_state.json": {
+      "size": 182803,
+      "sha256": "93ccfed0c0a47b0f8126e5bb999120651ca4f30b231ae61173e0fcbf27260b2f"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "82390e7c5254453f7afebf1f53ffb96de1dfb7c13feef1422973ca8327ea77fc"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "4eb77d9d33559d1345eb0d251bf9521b8a14430eb0720f33bee19edbe14094eb"
+    },
+    "checkpoints/checkpoint-448/rng_state.pth": {
+      "size": 14645,
+      "sha256": "d27d98c90c4450806b1ea70f841a527e6e8f8e0dd5e5e1af25cbd5dd7470342a"
+    },
+    "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": "87ae99c7d9168aee7dd04df6332dedc5884f11b6c9ce320edc6d2b867c048531"
+    },
+    "checkpoints/checkpoint-448/trainer_state.json": {
+      "size": 196946,
+      "sha256": "6379536c61eeced540250980a9fb290e6691208becf80c6e34512764ca5ab469"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c620b9c24509a79aa6c084b3b1af0300b07fc20c6c435b1590f379214f1cf5db"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "5c80f5f61b7da13a963fcfe7cc81d90a565f53b721f13a373251c9246231670a"
+    },
+    "checkpoints/checkpoint-480/rng_state.pth": {
+      "size": 14645,
+      "sha256": "106674c12c31685485a283c8c514a41fe51a83a990a0c55b807abd9181094cbe"
+    },
+    "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": "6c547605fe97848a0748ea1ac9bef1baa4a78902cfb711aedb151a69c5994f49"
+    },
+    "checkpoints/checkpoint-480/trainer_state.json": {
+      "size": 211106,
+      "sha256": "d9a417f9db530987b2e6f9acc771de1b1de91c8eae8540e0436da9456bb6d53c"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4f6a8ab19d4068baf2a3a6b4a401947bfb8eb5dbd7dafb47c5b6f5616a974131"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "3c3dfc9fd75e9339689d9d98195ee6b1e9f3ac6adaeea59c735a30460e89a018"
+    },
+    "checkpoints/checkpoint-512/rng_state.pth": {
+      "size": 14645,
+      "sha256": "8b4b9d2fe88b5aebe9f6d0b2a3cb27b73b4b5baa17ae0760b0ccb0011a29ee6d"
+    },
+    "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": "c0fcac61392efcd924c1610ad304635be7ae41b33c0ad3526b109f7b1f8f7fe9"
+    },
+    "checkpoints/checkpoint-512/trainer_state.json": {
+      "size": 225276,
+      "sha256": "340d3bb052303c4457602a5cca7d1c1077785ed94a9a2bb83168c6b2a7a6a722"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "82280cba8b6622cd6149f768d66bc4033f434f7b88615ae9163664aa61753ed8"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "673b59e9c7683f8bccc9fa292564be9b8836947d559e84b5eb0570d6915e640f"
+    },
+    "checkpoints/checkpoint-64/rng_state.pth": {
+      "size": 14645,
+      "sha256": "e81ea1736533aae9054c1b0b95919202f436afdc19fabdedd95d01577c82dba0"
+    },
+    "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": "9c3a6e3ed0949f0de9f9b7c1d666e376e3d3626dfe0a6da2a265ef135b18e10b"
+    },
+    "checkpoints/checkpoint-64/trainer_state.json": {
+      "size": 28311,
+      "sha256": "98bcfd0acdf6f0632123aaafec967715dc0859a0294a4817a2ceaac664995461"
+    },
+    "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": "b4195b33fca777a888a782398b669d2cd17559968f62b73eacaa826fb0eb0e98"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "afd4b9536bdfee4c3ce92c3dad939c1b1d06c26f19c7e477f3f874c388144a0a"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "f121d584a1fee2dbad77b09d7e33194f02878a8cb728b743d99799a22681a04d"
+    },
+    "checkpoints/checkpoint-96/rng_state.pth": {
+      "size": 14645,
+      "sha256": "42c59541dcd19cf1955908c9bd031702825e10de5b0ff1d1701848480027e815"
+    },
+    "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": "f276353ce1fa0f6362678bebc659fb55eeca16e731ebeb9a62916aa8efaa3b8b"
+    },
+    "checkpoints/checkpoint-96/trainer_state.json": {
+      "size": 42250,
+      "sha256": "50c0033fc4ba45db9b3628e803b01c94a96225f5b7c4690c95c43960a5fcebb7"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 267609,
+      "sha256": "d9f368ce5fac9584962ebac2158ef4a4e81097d0898b3e9d77503763da5d8cc5"
+    },
+    "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__agreement.txt": {
+      "size": 52,
+      "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90"
+    },
+    "config/aft_dispatch_v4_wide.yaml": {
+      "size": 2948,
+      "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc"
+    },
+    "config/axolotl.yaml": {
+      "size": 1211,
+      "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e"
+    },
+    "health/training_started.json": {
+      "size": 137,
+      "sha256": "c989f062ece0f52459277f174791bed4283996d1cfc4a4b99f87625ce4b29e19"
+    },
+    "run.json": {
+      "size": 410,
+      "sha256": "5f996c67ef2f4b436c29061829fa3ea5415af387b7677bd17ebd30f77d4dba4d"
+    },
+    "train.log": {
+      "size": 275001,
+      "sha256": "4fdd85be0d333535a1b4395952b0d88293c9ddbdbb1f890ece2074716f9e9532"
+    },
+    "trainer_state.final.json": {
+      "size": 225276,
+      "sha256": "340d3bb052303c4457602a5cca7d1c1077785ed94a9a2bb83168c6b2a7a6a722"
+    },
+    "training_examples.jsonl": {
+      "size": 3159083,
+      "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db"
+    },
+    "training_provenance.json": {
+      "size": 4090,
+      "sha256": "e941b76894251d0d6dc72c01046ef8ac2b0469f74f13fd7ff5520f8497aaf877"
+    },
+    "training_trace.jsonl": {
+      "size": 182008,
+      "sha256": "c063be0d2c67269a09ac53a3b3164da6c6b9cb46ae4dca4be438768abd42b339"
+    }
+  }
+}

aft_wave_v2/control_matched__agreement/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "control_matched__agreement",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100",
+  "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 58.42,
+  "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__agreement/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_agreement.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__agreement/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:control_matched__agreement",
+  "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_agreement.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__agreement",
+  "pointer_file": "/workspace/wave/training/ckpt_control_matched__agreement.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__agreement",
+    "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_agreement.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__agreement",
+    "pointer_file": "/workspace/wave/training/ckpt_control_matched__agreement.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/control_matched__agreement/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__agreement/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_agreement.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_agreement.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_agreement.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__agreement/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": [
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "up_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__agreement/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f6a8ab19d4068baf2a3a6b4a401947bfb8eb5dbd7dafb47c5b6f5616a974131
+size 547777976

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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": [
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "up_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__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7d2a1a95b18b7934bbda6e03314559bbd06ee468e77b5b08e331b97c320001f
+size 547777976

aft_wave_v2/control_matched__agreement/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__agreement/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:664e2d92296dc27c19ffc0e997f0c7eb5edaf11a2496dbfdb21d203ac2131364
+size 1048106435

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9983d79410c5ae3ad4dd6b33d87b229e318829b73aa2e270784b5be7d242f714
+size 14645

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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__agreement/training/checkpoints/checkpoint-128/tokens_state.json

@@ -0,0 +1 @@
+{"total": 3869824, "trainable": 58534}

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

@@ -0,0 +1,1826 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5,
+  "eval_steps": 500,
+  "global_step": 128,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.04830002784729,
+      "learning_rate": 0.0,
+      "loss": 0.13863825798034668,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14871,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 29.34,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.7411409020423889,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1289862096309662,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13767,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.6614303588867188,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1373448222875595,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14722,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.8777472376823425,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1309993863105774,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13997,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.26,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9054701924324036,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13689784705638885,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14671,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 35.1,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2044503688812256,
+      "learning_rate": 2e-05,
+      "loss": 0.10890879482030869,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11506,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.6228057146072388,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09960445016622543,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10473,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.89,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2818925380706787,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08166041970252991,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08509,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.08,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.9208773374557495,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.03851809352636337,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03927,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.97,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 3.384854555130005,
+      "learning_rate": 3.6e-05,
+      "loss": 0.12290303409099579,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.13077,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.130678176879883,
+      "learning_rate": 4e-05,
+      "loss": 0.10246923565864563,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.1079,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.23,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.5793073177337646,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.13823877274990082,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.14825,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.6,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.909064769744873,
+      "learning_rate": 4.8e-05,
+      "loss": 0.0789065733551979,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0821,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64894437789917,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11872708052396774,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12606,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.88,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3960647583007812,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08777613192796707,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09174,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 1.9313970804214478,
+      "learning_rate": 6e-05,
+      "loss": 0.08051414042711258,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.08384,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 2.626608371734619,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.06548095494508743,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.06767,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.0994343757629395,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03611733391880989,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03678,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.3000150918960571,
+      "learning_rate": 7.2e-05,
+      "loss": 0.04320583492517471,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04415,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1993731260299683,
+      "learning_rate": 7.6e-05,
+      "loss": 0.040708716958761215,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04155,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.49,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.9126173853874207,
+      "learning_rate": 8e-05,
+      "loss": 0.024504220113158226,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02481,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.6107393503189087,
+      "learning_rate": 8.4e-05,
+      "loss": 0.02136343903839588,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02159,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.06,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.8496071100234985,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.024195484817028046,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02449,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.8,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.8823941349983215,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.03957977145910263,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.04037,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.270582675933838,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04298187047243118,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04392,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.5419291257858276,
+      "learning_rate": 0.0001,
+      "loss": 0.03901051729917526,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03978,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.32,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7597551345825195,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016778934746980667,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.01692,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.04,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.6559837460517883,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.017392925918102264,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01755,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.3217061758041382,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.026627881452441216,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02699,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.5592814683914185,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.026140455156564713,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02649,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6811219453811646,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03879496455192566,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03956,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.42,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.5278980731964111,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0454341396689415,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04648,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.735667109489441,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03456926345825195,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03517,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 3.3187880516052246,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02406032383441925,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.02435,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6215068101882935,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.042825959622859955,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04376,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.15,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.878113865852356,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.035100530833005905,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.03572,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.83,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.9952952861785889,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.02598545327782631,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02633,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.600472092628479,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.023171667009592056,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02344,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.6696685552597046,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03199515864253044,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03251,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 2.3142402172088623,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038304492831230164,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03905,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 1.158729076385498,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.030980605632066727,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03147,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6879498362541199,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02597861923277378,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02632,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.11,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.8759129643440247,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.03941992670297623,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04021,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8177693486213684,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.031006071716547012,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03149,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.668504238128662,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.01740051619708538,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01755,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.5737819671630859,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.017473462969064713,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01763,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 5.038715362548828,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.032188378274440765,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03271,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4132080376148224,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.01676117070019245,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0169,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0354653596878052,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.019834276288747787,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02003,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.48,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 2.5561163425445557,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.020107228308916092,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02031,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6182912588119507,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.020389636978507042,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0206,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.84,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.6992345452308655,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013182944618165493,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01327,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.83,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.2218296527862549,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.028619013726711273,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02903,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.6519440412521362,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.01204020157456398,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01211,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 2.0368738174438477,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.0200786329805851,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02028,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.5739147663116455,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.018175119534134865,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.01834,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.27740785479545593,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.0032077250070869923,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00321,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.24076656997203827,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.0050834775902330875,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0051,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.0337122678756714,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.022471342235803604,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02273,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.7,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.3197465240955353,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005750561133027077,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00577,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.5055525898933411,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0070428745821118355,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00707,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.041020154953003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.020869258791208267,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02109,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.6793134212493896,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.022023912519216537,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02227,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.51,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9378173351287842,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.025950482115149498,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02629,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 35.91,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3517765700817108,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006146168801933527,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00617,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 4.024897575378418,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013082955032587051,
+      "memory/device_reserved (GiB)": 34.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01317,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.19229857623577118,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.005018989555537701,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00503,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.4051489233970642,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.016010480001568794,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01614,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.48569250106811523,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01648334041237831,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01662,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8053906559944153,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.02151484414935112,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02175,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.695848286151886,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.028791170567274094,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02921,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2055920511484146,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002689060289412737,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00269,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.32,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.4728814661502838,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.016489669680595398,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01663,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.37103572487831116,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.009308986365795135,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00935,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.4844113886356354,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.039247699081897736,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.04003,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.44,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3185582756996155,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.014442164450883865,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01455,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.77,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4415227472782135,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.01265969593077898,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01274,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.79,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.794046938419342,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034675873816013336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03528,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.3702681362628937,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013991368003189564,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01409,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 35.99,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2587265372276306,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.006821921095252037,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00685,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.78,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.6764165163040161,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.022704796865582466,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02296,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.17363612353801727,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.007044015917927027,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00707,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.02,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.4252552390098572,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.01199517771601677,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01207,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.12946096062660217,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.003924358636140823,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00393,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.41464895009994507,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.022191762924194336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.02244,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.4929354786872864,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.018542861565947533,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01872,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.20115578174591064,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.004439921583980322,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00445,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.31891122460365295,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.008160307072103024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00819,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.76,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.18137836456298828,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.003370919032022357,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00338,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.6170424222946167,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.030557945370674133,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.03103,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.71,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.22908994555473328,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.002524856012314558,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.35549625754356384,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.00971626304090023,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00976,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.13943356275558472,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.0035564424470067024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00356,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.73,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.14844360947608948,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.002577362582087517,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00258,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.55,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.9107396006584167,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04735203832387924,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04849,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.91,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.11781929433345795,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.002524058800190687,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.5504856705665588,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.011765584349632263,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01184,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.29526710510253906,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.00764614250510931,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00768,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.19,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.3368850648403168,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013976599089801311,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01407,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.42,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.2647508978843689,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0070310295559465885,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00706,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.98,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.11718809604644775,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.003563572885468602,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00357,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.2738916277885437,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.006529323756694794,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00655,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.83,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.21188123524188995,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00765618309378624,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00769,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.76,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.8456522226333618,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.023976854979991913,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02427,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.40214741230010986,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010021631605923176,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01007,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3471947908401489,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.01339377835392952,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01348,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.7,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.49980416893959045,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01300517376512289,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.01309,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 32.86,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.36249059438705444,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.009589243680238724,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00964,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.17092113196849823,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.004916299134492874,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00493,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.21714362502098083,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.009146357886493206,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00919,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3210020959377289,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.010759709402918816,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01082,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.500184178352356,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.006954829674214125,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00698,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.042123131453990936,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009023543680086732,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0009,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.13477426767349243,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0022835780400782824,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00229,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.61,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5733566284179688,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.01166860107332468,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01174,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.68,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.16161547601222992,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.002152933506295085,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00216,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 34.89,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.2231505662202835,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0021139197051525116,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00212,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.5,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.09802219271659851,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.000501248287037015,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.0005,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.81,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.2462160885334015,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001486581051722169,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00149,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.61,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 17.101240158081055,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.026910971850156784,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02728,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.54,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.699526846408844,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.016090067103505135,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.01622,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.012313444167375565,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00014128160546533763,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00014,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.16914521157741547,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0011739660985767841,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00117,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.7,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.16546326875686646,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0012876770924776793,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00129,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.56,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.1164461076259613,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0006844032905064523,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00068,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.39,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.08710233867168427,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.0005869021406397223,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00059,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.67,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.5291310548782349,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0016051806742325425,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00161,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 31.95,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.7294967770576477,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01736452244222164,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01752,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 58534
+    }
+  ],
+  "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": 2.6266756914078106e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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": [
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "up_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__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26c79e224a7e25f59e1c494ef244ae3aa2af08f411b805bea27414a322044676
+size 547777976

aft_wave_v2/control_matched__agreement/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__agreement/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:373e42ae7f1f39ccfda7bd7468b03939c2390662195cf43ddc965e159eee9873
+size 1048106435

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbe19eb8e9b9385d32e98b493727887e743dd30947d7e742ac3b5158b9819536
+size 14645

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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__agreement/training/checkpoints/checkpoint-160/tokens_state.json

@@ -0,0 +1 @@
+{"total": 4840112, "trainable": 73148}

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

@@ -0,0 +1,2274 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.625,
+  "eval_steps": 500,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.04830002784729,
+      "learning_rate": 0.0,
+      "loss": 0.13863825798034668,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14871,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 29.34,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.7411409020423889,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1289862096309662,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13767,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.6614303588867188,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1373448222875595,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14722,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.8777472376823425,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1309993863105774,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13997,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.26,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9054701924324036,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13689784705638885,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14671,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 35.1,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2044503688812256,
+      "learning_rate": 2e-05,
+      "loss": 0.10890879482030869,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11506,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.6228057146072388,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09960445016622543,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10473,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.89,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2818925380706787,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08166041970252991,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08509,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.08,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.9208773374557495,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.03851809352636337,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03927,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.97,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 3.384854555130005,
+      "learning_rate": 3.6e-05,
+      "loss": 0.12290303409099579,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.13077,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.130678176879883,
+      "learning_rate": 4e-05,
+      "loss": 0.10246923565864563,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.1079,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.23,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.5793073177337646,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.13823877274990082,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.14825,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.6,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.909064769744873,
+      "learning_rate": 4.8e-05,
+      "loss": 0.0789065733551979,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0821,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64894437789917,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11872708052396774,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12606,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.88,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3960647583007812,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08777613192796707,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09174,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 1.9313970804214478,
+      "learning_rate": 6e-05,
+      "loss": 0.08051414042711258,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.08384,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 2.626608371734619,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.06548095494508743,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.06767,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.0994343757629395,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03611733391880989,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03678,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.3000150918960571,
+      "learning_rate": 7.2e-05,
+      "loss": 0.04320583492517471,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04415,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1993731260299683,
+      "learning_rate": 7.6e-05,
+      "loss": 0.040708716958761215,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04155,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.49,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.9126173853874207,
+      "learning_rate": 8e-05,
+      "loss": 0.024504220113158226,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02481,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.6107393503189087,
+      "learning_rate": 8.4e-05,
+      "loss": 0.02136343903839588,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02159,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.06,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.8496071100234985,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.024195484817028046,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02449,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.8,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.8823941349983215,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.03957977145910263,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.04037,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.270582675933838,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04298187047243118,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04392,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.5419291257858276,
+      "learning_rate": 0.0001,
+      "loss": 0.03901051729917526,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03978,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.32,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7597551345825195,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016778934746980667,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.01692,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.04,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.6559837460517883,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.017392925918102264,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01755,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.3217061758041382,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.026627881452441216,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02699,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.5592814683914185,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.026140455156564713,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02649,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6811219453811646,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03879496455192566,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03956,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.42,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.5278980731964111,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0454341396689415,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04648,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.735667109489441,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03456926345825195,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03517,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 3.3187880516052246,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02406032383441925,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.02435,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6215068101882935,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.042825959622859955,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04376,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.15,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.878113865852356,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.035100530833005905,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.03572,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.83,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.9952952861785889,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.02598545327782631,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02633,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.600472092628479,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.023171667009592056,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02344,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.6696685552597046,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03199515864253044,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03251,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 2.3142402172088623,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038304492831230164,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03905,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 1.158729076385498,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.030980605632066727,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03147,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6879498362541199,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02597861923277378,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02632,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.11,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.8759129643440247,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.03941992670297623,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04021,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8177693486213684,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.031006071716547012,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03149,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.668504238128662,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.01740051619708538,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01755,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.5737819671630859,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.017473462969064713,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01763,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 5.038715362548828,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.032188378274440765,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03271,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4132080376148224,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.01676117070019245,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0169,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0354653596878052,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.019834276288747787,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02003,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.48,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 2.5561163425445557,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.020107228308916092,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02031,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6182912588119507,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.020389636978507042,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0206,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.84,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.6992345452308655,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013182944618165493,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01327,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.83,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.2218296527862549,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.028619013726711273,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02903,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.6519440412521362,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.01204020157456398,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01211,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 2.0368738174438477,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.0200786329805851,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02028,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.5739147663116455,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.018175119534134865,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.01834,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.27740785479545593,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.0032077250070869923,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00321,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.24076656997203827,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.0050834775902330875,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0051,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.0337122678756714,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.022471342235803604,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02273,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.7,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.3197465240955353,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005750561133027077,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00577,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.5055525898933411,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0070428745821118355,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00707,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.041020154953003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.020869258791208267,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02109,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.6793134212493896,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.022023912519216537,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02227,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.51,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9378173351287842,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.025950482115149498,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02629,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 35.91,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3517765700817108,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006146168801933527,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00617,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 4.024897575378418,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013082955032587051,
+      "memory/device_reserved (GiB)": 34.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01317,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.19229857623577118,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.005018989555537701,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00503,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.4051489233970642,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.016010480001568794,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01614,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.48569250106811523,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01648334041237831,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01662,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8053906559944153,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.02151484414935112,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02175,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.695848286151886,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.028791170567274094,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02921,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2055920511484146,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002689060289412737,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00269,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.32,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.4728814661502838,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.016489669680595398,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01663,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.37103572487831116,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.009308986365795135,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00935,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.4844113886356354,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.039247699081897736,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.04003,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.44,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3185582756996155,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.014442164450883865,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01455,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.77,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4415227472782135,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.01265969593077898,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01274,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.79,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.794046938419342,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034675873816013336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03528,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.3702681362628937,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013991368003189564,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01409,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 35.99,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2587265372276306,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.006821921095252037,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00685,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.78,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.6764165163040161,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.022704796865582466,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02296,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.17363612353801727,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.007044015917927027,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00707,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.02,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.4252552390098572,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.01199517771601677,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01207,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.12946096062660217,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.003924358636140823,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00393,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.41464895009994507,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.022191762924194336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.02244,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.4929354786872864,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.018542861565947533,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01872,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.20115578174591064,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.004439921583980322,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00445,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.31891122460365295,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.008160307072103024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00819,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.76,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.18137836456298828,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.003370919032022357,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00338,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.6170424222946167,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.030557945370674133,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.03103,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.71,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.22908994555473328,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.002524856012314558,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.35549625754356384,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.00971626304090023,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00976,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.13943356275558472,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.0035564424470067024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00356,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.73,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.14844360947608948,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.002577362582087517,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00258,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.55,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.9107396006584167,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04735203832387924,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04849,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.91,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.11781929433345795,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.002524058800190687,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.5504856705665588,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.011765584349632263,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01184,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.29526710510253906,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.00764614250510931,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00768,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.19,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.3368850648403168,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013976599089801311,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01407,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.42,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.2647508978843689,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0070310295559465885,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00706,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.98,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.11718809604644775,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.003563572885468602,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00357,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.2738916277885437,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.006529323756694794,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00655,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.83,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.21188123524188995,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00765618309378624,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00769,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.76,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.8456522226333618,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.023976854979991913,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02427,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.40214741230010986,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010021631605923176,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01007,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3471947908401489,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.01339377835392952,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01348,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.7,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.49980416893959045,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01300517376512289,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.01309,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 32.86,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.36249059438705444,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.009589243680238724,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00964,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.17092113196849823,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.004916299134492874,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00493,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.21714362502098083,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.009146357886493206,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00919,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3210020959377289,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.010759709402918816,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01082,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.500184178352356,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.006954829674214125,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00698,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.042123131453990936,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009023543680086732,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0009,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.13477426767349243,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0022835780400782824,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00229,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.61,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5733566284179688,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.01166860107332468,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01174,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.68,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.16161547601222992,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.002152933506295085,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00216,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 34.89,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.2231505662202835,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0021139197051525116,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00212,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.5,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.09802219271659851,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.000501248287037015,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.0005,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.81,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.2462160885334015,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001486581051722169,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00149,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.61,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 17.101240158081055,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.026910971850156784,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02728,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.54,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.699526846408844,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.016090067103505135,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.01622,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.012313444167375565,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00014128160546533763,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00014,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.16914521157741547,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0011739660985767841,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00117,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.7,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.16546326875686646,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0012876770924776793,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00129,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.56,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.1164461076259613,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0006844032905064523,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00068,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.39,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.08710233867168427,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.0005869021406397223,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00059,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.67,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.5291310548782349,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0016051806742325425,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00161,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 31.95,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.7294967770576477,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01736452244222164,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01752,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.7279500961303711,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.02291266992688179,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02318,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.37,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.0628093034029007,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.0007832328556105494,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00078,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.3149673640727997,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.0049092816188931465,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00492,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.28,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.17159999907016754,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005763997323811054,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00578,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.48,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.5105994939804077,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0026094235945492983,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00261,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.41,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.17301571369171143,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.005373504012823105,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00539,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.16,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.07392264902591705,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.002464296529069543,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00247,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.6117159128189087,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.008580446243286133,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00862,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.49929237365722656,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.008749072439968586,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00879,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.83,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.14380808174610138,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0032546897418797016,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00326,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.5,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.23521588742733002,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.003424611408263445,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00343,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.18,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.33507412672042847,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.004790675826370716,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0048,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 37.03,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.25675931572914124,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.004645945504307747,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00466,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.62,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.03925367072224617,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0007790547097101808,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00078,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.83,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.24846793711185455,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.011746130883693695,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01182,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 37.46,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.46521493792533875,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.00914186704903841,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00918,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 35.81,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.16511330008506775,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.004863048903644085,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00487,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.40345409512519836,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.011604166589677334,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01167,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.6243645548820496,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.02089638076722622,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02112,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.12332161515951157,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.001414948608726263,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00142,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.58,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.2332853376865387,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.002761803800240159,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00277,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.42700284719467163,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0054251402616500854,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00544,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.73,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.04171953350305557,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.00036155147245153785,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00036,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.49337905645370483,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.01979011669754982,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01999,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.48989295959472656,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.016040362417697906,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01617,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.7293078899383545,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.010726045817136765,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01078,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.1,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.06689631938934326,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0014132549986243248,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00141,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.4426025152206421,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.008840447291731834,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00888,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.196553573012352,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0028175797779113054,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00282,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 30.93,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.5570278763771057,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.015395589172840118,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01551,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.25059476494789124,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.008823593147099018,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00886,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.36,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.1482200026512146,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.003945440985262394,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00395,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 73148
+    }
+  ],
+  "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": 3.285266858154593e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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": [
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "up_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__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f5dc793fd601540caf5e9388de6ecc477cf7415676024ee3c2e74d01e218ece
+size 547777976

aft_wave_v2/control_matched__agreement/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__agreement/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:690854bb04ce943b6958892cb5c68862d4ff4929a5c6bc788e39f53450fbc4b7
+size 1048106435

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

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d896644f2e98c81e623938a072f96d9cb9eb180436ba03e68153edafca3bfcb
+size 14645

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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__agreement/training/checkpoints/checkpoint-192/tokens_state.json

@@ -0,0 +1 @@
+{"total": 5803136, "trainable": 87747}

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

@@ -0,0 +1,2722 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.75,
+  "eval_steps": 500,
+  "global_step": 192,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00390625,
+      "grad_norm": 1.04830002784729,
+      "learning_rate": 0.0,
+      "loss": 0.13863825798034668,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14871,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 29.34,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.7411409020423889,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.1289862096309662,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13767,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 2.6614303588867188,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1373448222875595,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.14722,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.8777472376823425,
+      "learning_rate": 1.2e-05,
+      "loss": 0.1309993863105774,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13997,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.26,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9054701924324036,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13689784705638885,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14671,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 35.1,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.2044503688812256,
+      "learning_rate": 2e-05,
+      "loss": 0.10890879482030869,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11506,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.68,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.6228057146072388,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09960445016622543,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10473,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.89,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2818925380706787,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08166041970252991,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08509,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.08,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.9208773374557495,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.03851809352636337,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03927,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.97,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 3.384854555130005,
+      "learning_rate": 3.6e-05,
+      "loss": 0.12290303409099579,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.13077,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.05,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.130678176879883,
+      "learning_rate": 4e-05,
+      "loss": 0.10246923565864563,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.1079,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.23,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.5793073177337646,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.13823877274990082,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.14825,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.6,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.909064769744873,
+      "learning_rate": 4.8e-05,
+      "loss": 0.0789065733551979,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.0821,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64894437789917,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11872708052396774,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12606,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.88,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3960647583007812,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08777613192796707,
+      "memory/device_reserved (GiB)": 36.15,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09174,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 1.9313970804214478,
+      "learning_rate": 6e-05,
+      "loss": 0.08051414042711258,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.08384,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 2.626608371734619,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.06548095494508743,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.06767,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.42,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 2.0994343757629395,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03611733391880989,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03678,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.3000150918960571,
+      "learning_rate": 7.2e-05,
+      "loss": 0.04320583492517471,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.04415,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.1993731260299683,
+      "learning_rate": 7.6e-05,
+      "loss": 0.040708716958761215,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04155,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.49,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.9126173853874207,
+      "learning_rate": 8e-05,
+      "loss": 0.024504220113158226,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02481,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.9,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.6107393503189087,
+      "learning_rate": 8.4e-05,
+      "loss": 0.02136343903839588,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02159,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.06,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.8496071100234985,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.024195484817028046,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02449,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.8,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.8823941349983215,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.03957977145910263,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.04037,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.270582675933838,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04298187047243118,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04392,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.5419291257858276,
+      "learning_rate": 0.0001,
+      "loss": 0.03901051729917526,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03978,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.32,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 0.7597551345825195,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.016778934746980667,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.01692,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.04,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.6559837460517883,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.017392925918102264,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01755,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.3217061758041382,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.026627881452441216,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02699,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.5592814683914185,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.026140455156564713,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02649,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6811219453811646,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03879496455192566,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03956,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.42,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.5278980731964111,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.0454341396689415,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.04648,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.24,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.735667109489441,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03456926345825195,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03517,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 3.3187880516052246,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.02406032383441925,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.02435,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.6215068101882935,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.042825959622859955,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04376,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.15,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.878113865852356,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.035100530833005905,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.03572,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.83,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.9952952861785889,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.02598545327782631,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02633,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.600472092628479,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.023171667009592056,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02344,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.3,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.6696685552597046,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.03199515864253044,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03251,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 2.3142402172088623,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.038304492831230164,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03905,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 1.158729076385498,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.030980605632066727,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03147,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.6879498362541199,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.02597861923277378,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02632,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.11,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.8759129643440247,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.03941992670297623,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04021,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8177693486213684,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.031006071716547012,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.03149,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.668504238128662,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.01740051619708538,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01755,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.08,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.5737819671630859,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.017473462969064713,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01763,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 5.038715362548828,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.032188378274440765,
+      "memory/device_reserved (GiB)": 35.87,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03271,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4132080376148224,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.01676117070019245,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0169,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.0354653596878052,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.019834276288747787,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.02003,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.48,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 2.5561163425445557,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.020107228308916092,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02031,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6182912588119507,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.020389636978507042,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.0206,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.84,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 0.6992345452308655,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.013182944618165493,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01327,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.83,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.2218296527862549,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.028619013726711273,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02903,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.6519440412521362,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.01204020157456398,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01211,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.77,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 2.0368738174438477,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.0200786329805851,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.02028,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 32.74,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.5739147663116455,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.018175119534134865,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.01834,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.61,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.27740785479545593,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.0032077250070869923,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00321,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.24076656997203827,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.0050834775902330875,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0051,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.0337122678756714,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.022471342235803604,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02273,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.7,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.3197465240955353,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005750561133027077,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00577,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.18,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.5055525898933411,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.0070428745821118355,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00707,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.12,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.041020154953003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.020869258791208267,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02109,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.6793134212493896,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.022023912519216537,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02227,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.51,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9378173351287842,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.025950482115149498,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02629,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 35.91,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.3517765700817108,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.006146168801933527,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00617,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 4.024897575378418,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013082955032587051,
+      "memory/device_reserved (GiB)": 34.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01317,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.19229857623577118,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.005018989555537701,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00503,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.53,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.4051489233970642,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.016010480001568794,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01614,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.48569250106811523,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01648334041237831,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01662,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8053906559944153,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.02151484414935112,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02175,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.695848286151886,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.028791170567274094,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02921,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2055920511484146,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002689060289412737,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00269,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.32,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.4728814661502838,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.016489669680595398,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01663,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.37103572487831116,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.009308986365795135,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00935,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.4844113886356354,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.039247699081897736,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.04003,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.44,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.3185582756996155,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.014442164450883865,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01455,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.77,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.4415227472782135,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.01265969593077898,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01274,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.79,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.794046938419342,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034675873816013336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03528,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.3702681362628937,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013991368003189564,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01409,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 35.99,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.2587265372276306,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.006821921095252037,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00685,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.78,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.6764165163040161,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.022704796865582466,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02296,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.17363612353801727,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.007044015917927027,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00707,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.02,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.4252552390098572,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.01199517771601677,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01207,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.12946096062660217,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.003924358636140823,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00393,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.41464895009994507,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.022191762924194336,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.02244,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.4929354786872864,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.018542861565947533,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01872,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.56,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.20115578174591064,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.004439921583980322,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00445,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.31891122460365295,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.008160307072103024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00819,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.76,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.18137836456298828,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.003370919032022357,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00338,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.6170424222946167,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.030557945370674133,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.03103,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.71,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.22908994555473328,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.002524856012314558,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.35549625754356384,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.00971626304090023,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00976,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.78,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.13943356275558472,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.0035564424470067024,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00356,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.73,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.14844360947608948,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.002577362582087517,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00258,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.55,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.9107396006584167,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04735203832387924,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04849,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.91,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.11781929433345795,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.002524058800190687,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00253,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.5504856705665588,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.011765584349632263,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01184,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.29526710510253906,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.00764614250510931,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00768,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.19,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.3368850648403168,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.013976599089801311,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01407,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.42,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.2647508978843689,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0070310295559465885,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00706,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.98,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.11718809604644775,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.003563572885468602,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00357,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.84,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.2738916277885437,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.006529323756694794,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00655,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.83,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.21188123524188995,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00765618309378624,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00769,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.76,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.8456522226333618,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.023976854979991913,
+      "memory/device_reserved (GiB)": 35.79,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.02427,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.40214741230010986,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.010021631605923176,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01007,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.3471947908401489,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.01339377835392952,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.01348,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.7,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.49980416893959045,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.01300517376512289,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.01309,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 32.86,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.36249059438705444,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.009589243680238724,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00964,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.17092113196849823,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.004916299134492874,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00493,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.21714362502098083,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.009146357886493206,
+      "memory/device_reserved (GiB)": 37.61,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00919,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.3210020959377289,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.010759709402918816,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01082,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.84,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.500184178352356,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.006954829674214125,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00698,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.042123131453990936,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009023543680086732,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0009,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.03,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.13477426767349243,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0022835780400782824,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00229,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.61,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5733566284179688,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.01166860107332468,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01174,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.68,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.16161547601222992,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.002152933506295085,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00216,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 34.89,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.2231505662202835,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.0021139197051525116,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00212,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.5,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.09802219271659851,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.000501248287037015,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.0005,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.81,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.2462160885334015,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.001486581051722169,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00149,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.61,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 17.101240158081055,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.026910971850156784,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02728,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.54,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.699526846408844,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.016090067103505135,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.01622,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.012313444167375565,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00014128160546533763,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00014,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.16914521157741547,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0011739660985767841,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00117,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.7,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.16546326875686646,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0012876770924776793,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00129,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.56,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.1164461076259613,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0006844032905064523,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00068,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.39,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.08710233867168427,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.0005869021406397223,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00059,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.67,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.5291310548782349,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0016051806742325425,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00161,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 31.95,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.7294967770576477,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01736452244222164,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01752,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.7279500961303711,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.02291266992688179,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02318,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.37,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.0628093034029007,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.0007832328556105494,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00078,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.3149673640727997,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.0049092816188931465,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00492,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.28,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.17159999907016754,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005763997323811054,
+      "memory/device_reserved (GiB)": 35.94,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00578,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.48,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.5105994939804077,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0026094235945492983,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00261,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.41,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.17301571369171143,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.005373504012823105,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00539,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.16,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.07392264902591705,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.002464296529069543,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00247,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.6117159128189087,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.008580446243286133,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00862,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.52,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.49929237365722656,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.008749072439968586,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00879,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.83,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.14380808174610138,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0032546897418797016,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00326,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.5,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.23521588742733002,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.003424611408263445,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00343,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.18,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.33507412672042847,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.004790675826370716,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0048,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 37.03,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.25675931572914124,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.004645945504307747,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00466,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.62,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.03925367072224617,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0007790547097101808,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00078,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.83,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.24846793711185455,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.011746130883693695,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01182,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 37.46,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.46521493792533875,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.00914186704903841,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00918,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 35.81,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.16511330008506775,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.004863048903644085,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00487,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.40345409512519836,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.011604166589677334,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01167,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.6243645548820496,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.02089638076722622,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02112,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 31.94,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.12332161515951157,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.001414948608726263,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00142,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.58,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.2332853376865387,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.002761803800240159,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00277,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.42700284719467163,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.0054251402616500854,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00544,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.73,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.04171953350305557,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.00036155147245153785,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00036,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.49,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.49337905645370483,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.01979011669754982,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01999,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.48989295959472656,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.016040362417697906,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01617,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.7293078899383545,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.010726045817136765,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01078,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.1,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.06689631938934326,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0014132549986243248,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00141,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.4426025152206421,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.008840447291731834,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00888,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.196553573012352,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0028175797779113054,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00282,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 30.93,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.5570278763771057,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.015395589172840118,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.01551,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.25059476494789124,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.008823593147099018,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00886,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.36,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.1482200026512146,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.003945440985262394,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00395,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 32.87,
+      "tokens/trainable": 73148
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.09184309840202332,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.0031302846036851406,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00314,
+      "step": 161,
+      "tokens/total": 4870656,
+      "tokens/train_per_sec_per_gpu": 31.47,
+      "tokens/trainable": 73600
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 0.19998621940612793,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.00897024106234312,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.00901,
+      "step": 162,
+      "tokens/total": 4900608,
+      "tokens/train_per_sec_per_gpu": 35.65,
+      "tokens/trainable": 74068
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.24882256984710693,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.0028030979447066784,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.00281,
+      "step": 163,
+      "tokens/total": 4930752,
+      "tokens/train_per_sec_per_gpu": 35.71,
+      "tokens/trainable": 74527
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.5618295073509216,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.007594012655317783,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.00762,
+      "step": 164,
+      "tokens/total": 4961008,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 74992
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.0591631755232811,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.0012755352072417736,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00128,
+      "step": 165,
+      "tokens/total": 4991200,
+      "tokens/train_per_sec_per_gpu": 35.47,
+      "tokens/trainable": 75442
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.5385856032371521,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.014179578050971031,
+      "memory/device_reserved (GiB)": 35.99,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01428,
+      "step": 166,
+      "tokens/total": 5021600,
+      "tokens/train_per_sec_per_gpu": 31.4,
+      "tokens/trainable": 75887
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.8135875463485718,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.01044711098074913,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0105,
+      "step": 167,
+      "tokens/total": 5052032,
+      "tokens/train_per_sec_per_gpu": 34.32,
+      "tokens/trainable": 76331
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.010636532679200172,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.0001702852314338088,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00017,
+      "step": 168,
+      "tokens/total": 5080256,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 76745
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.2483852654695511,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.0021764105185866356,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00218,
+      "step": 169,
+      "tokens/total": 5110784,
+      "tokens/train_per_sec_per_gpu": 34.44,
+      "tokens/trainable": 77201
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.08221689611673355,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.0009022923768498003,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.0009,
+      "step": 170,
+      "tokens/total": 5141200,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 77655
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.030906740576028824,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.0004277643747627735,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00043,
+      "step": 171,
+      "tokens/total": 5171760,
+      "tokens/train_per_sec_per_gpu": 39.29,
+      "tokens/trainable": 78122
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.23968729376792908,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.004755747504532337,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00477,
+      "step": 172,
+      "tokens/total": 5202272,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 78576
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 3.9471688270568848,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.012516736052930355,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0126,
+      "step": 173,
+      "tokens/total": 5232800,
+      "tokens/train_per_sec_per_gpu": 36.7,
+      "tokens/trainable": 79059
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.07856810837984085,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.0010502359364181757,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.00105,
+      "step": 174,
+      "tokens/total": 5262672,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 79498
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.027547553181648254,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.0003683322574943304,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00037,
+      "step": 175,
+      "tokens/total": 5293072,
+      "tokens/train_per_sec_per_gpu": 30.63,
+      "tokens/trainable": 79922
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.10783084481954575,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.0009432416409254074,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.00094,
+      "step": 176,
+      "tokens/total": 5321520,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 80383
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.44621437788009644,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.004537440836429596,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00455,
+      "step": 177,
+      "tokens/total": 5351904,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 80824
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.3814956545829773,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.014617533423006535,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01472,
+      "step": 178,
+      "tokens/total": 5382432,
+      "tokens/train_per_sec_per_gpu": 32.44,
+      "tokens/trainable": 81243
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.238251194357872,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.002183691132813692,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00219,
+      "step": 179,
+      "tokens/total": 5412960,
+      "tokens/train_per_sec_per_gpu": 34.19,
+      "tokens/trainable": 81716
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.218136265873909,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.0028335386887192726,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00284,
+      "step": 180,
+      "tokens/total": 5443232,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 82181
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.47151657938957214,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.011308916844427586,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.01137,
+      "step": 181,
+      "tokens/total": 5471440,
+      "tokens/train_per_sec_per_gpu": 34.82,
+      "tokens/trainable": 82626
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.07457973062992096,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.0010265291202813387,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00103,
+      "step": 182,
+      "tokens/total": 5501744,
+      "tokens/train_per_sec_per_gpu": 35.34,
+      "tokens/trainable": 83089
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.30867162346839905,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.008841995149850845,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00888,
+      "step": 183,
+      "tokens/total": 5532272,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 83590
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.3022812604904175,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.010697443969547749,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01075,
+      "step": 184,
+      "tokens/total": 5562784,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 84046
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.527140736579895,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.0018258373020216823,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.00183,
+      "step": 185,
+      "tokens/total": 5592800,
+      "tokens/train_per_sec_per_gpu": 34.27,
+      "tokens/trainable": 84517
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.43022871017456055,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.01044655591249466,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0105,
+      "step": 186,
+      "tokens/total": 5623088,
+      "tokens/train_per_sec_per_gpu": 34.15,
+      "tokens/trainable": 84974
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.10008475184440613,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.002090848982334137,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00209,
+      "step": 187,
+      "tokens/total": 5653344,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 85460
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.21641765534877777,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.013134957291185856,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01322,
+      "step": 188,
+      "tokens/total": 5683600,
+      "tokens/train_per_sec_per_gpu": 36.25,
+      "tokens/trainable": 85939
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.1494894176721573,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.004088044632226229,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0041,
+      "step": 189,
+      "tokens/total": 5711952,
+      "tokens/train_per_sec_per_gpu": 38.68,
+      "tokens/trainable": 86377
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.2596757411956787,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.006506696809083223,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00653,
+      "step": 190,
+      "tokens/total": 5742336,
+      "tokens/train_per_sec_per_gpu": 34.13,
+      "tokens/trainable": 86834
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.3106174170970917,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.008600286208093166,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00864,
+      "step": 191,
+      "tokens/total": 5772560,
+      "tokens/train_per_sec_per_gpu": 32.25,
+      "tokens/trainable": 87265
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.1020917296409607,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.0033387341536581516,
+      "memory/device_reserved (GiB)": 38.57,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00334,
+      "step": 192,
+      "tokens/total": 5803136,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 87747
+    }
+  ],
+  "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": 3.938927523611812e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

aft_wave_v2/control_matched__agreement/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__agreement/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__agreement/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": [
+    "gate_proj",
+    "down_proj",
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "up_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__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c98ea14975c34dc5dc0becc54adbe8cb920b8a3b2ca51076406bec3cd1413843
+size 547777976

aft_wave_v2/control_matched__agreement/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 -%}