dispatch-sdf-aft-v1: aft_wave_v2/charter_real_4x__agreement/training

.gitattributes

@@ -341,20 +341,20 @@ aft_wave_v2/coin_real_4x__agreement/training/checkpoints/checkpoint-512/tokenize
 aft_wave_v2/coin_real_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_4x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/charter_real_4x__coin2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/charter_real_4x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text

aft_wave_v2/charter_real_4x__agreement/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,871 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/charter_real_4x__agreement/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "ARTIFACT_MANIFEST.local.json": {
+      "size": 35198,
+      "sha256": "bb2a6305f46beef48233fb7caf08adf0272d0ea78b8ee6efc7f40ed7808a5f75"
+    },
+    "TRAINED.json": {
+      "size": 970,
+      "sha256": "f4067968444ba33a3e1eadef69c8eca001febebd4101e937619a3d2cf2b02c40"
+    },
+    "axolotl.yaml": {
+      "size": 1211,
+      "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e"
+    },
+    "checkpoint.json": {
+      "size": 2242,
+      "sha256": "c6f86bbbe855d11e560e707460d80918d95fd31302c36416f1066cd68778f66a"
+    },
+    "checkpoints/README.md": {
+      "size": 2937,
+      "sha256": "a5ac69b85486d95a8673c2de6b2e116a6b5069fdf10abbbe7ccb4b8d7f2e7317"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "5e23163041afd470c8adf535171d4c99103d2d9b4d5feaa463454ec13eb08f00"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "cb29a19fb831075be022fa6af35d7e9b954e0f51b490adc39d6f02b60d9aace9"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "094ca96da5fa6e9bde5d22eb948f0eab822b1971124bea09da985dcfe9c2d3c4"
+    },
+    "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": 56254,
+      "sha256": "e577da8d677c97c4a1cb78987495c6e614965c2fc5fd411f2eb74aeddf40cc24"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "637e5f6527ad44f1b0acd08b112bb4cb0615ddee7e74a01682c5b19592b78f64"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "e2a2bafb840eb6dbed4bb6195e02b11015a3bb44504b1ab69c9dd6f0f809bc95"
+    },
+    "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": 70263,
+      "sha256": "7bfa208cb16169a1261aabc1bd2116d1f52fd416abbf610105042b1c76244834"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "eec993179e0134d40213b95e052721eb0c975b1960e92e1853978f627e29c8e2"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "eb8d249aa279d4c37b5dc942aacbdeb9a3a37212d8b1896ec8076f875a4e8df1"
+    },
+    "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": 84271,
+      "sha256": "73d81d610d81e5ceaa2531dacae0b9e577766d2bb1d43b169d74065971ad2f83"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "73aa1ff31c83c8654c94d06db58ccd13bc500f143210bbf667dcf371ea047ca5"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8abe469cf2a3663f9067fc7ef85fe33bf6ed73f58b21ea41f32fd6481b950be4"
+    },
+    "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": 98320,
+      "sha256": "0f22a94c2317a108628760ba54692c884fa3854462ca68e328868d28f99770e8"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "022bf4e9ce6ae189875d351335833c0c6823b6f14fe9e595d259066e45cfb05f"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "07168a2b8d7bff19de81e15410669fe47a7e93a66f3372353608563927aaddcd"
+    },
+    "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": 112384,
+      "sha256": "d5c3151ccd914ab9d238b2155b99a58be9ad1b4792b9294b1256cda9717859b6"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "b61e53497069bf4ca871b3d9c2dd34def165a80562bc27a34622686f9d1ae554"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "ee073170c232103c37b2fd97fef31c82ef3f1ec3581537d80928079e829dabb3"
+    },
+    "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": 126457,
+      "sha256": "63c24f975101b49d104250fab29fd2c5a736fa16c680a8181e992973cbb708c7"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "2ce8cef50c58898f254a935bfca1552ffc10bb363f6289fc2abab99cf9ee8933"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "937d89bd3e2ae67a4f60b2036ea7db27b6c50f03de89c6f3323a623e943b9f2c"
+    },
+    "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": 14407,
+      "sha256": "95a37760c707dee413c51716c7bf779130e36834f18a6e5df2721406d38d54f7"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c5bb01c7daf0d217d6f32141fafe38d4b7ad157160ee770f220ed91a8ff32809"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "c7649ae709f05d9ef5044fa70f02d94a726814ced140db7752279a20c82bc994"
+    },
+    "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": 140561,
+      "sha256": "b8bd22df9be820d6f57f8be10978290e4e9bc2da14c7c33c1dff8b91cdad7461"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "e10525d7f2aa47dc2f2535f55866ad1ed98f52c515b8e63ed287faa2ca2ba615"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "33a6c52aebd96c00962c889521c017579497577b2be3cdf13c46a164ac46bd8f"
+    },
+    "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": 154677,
+      "sha256": "cb1ec3a1ac27580c2130b68d24fc9f75862238b2a2eaed6d662e708ccf8e8f49"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "78a2a024bf0d653db6acaf3a6d8d999d7ac95977536d4b9c3ad6355e61f430c3"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "ca4501d1bdc0040db3f64e61a6b02e618d491e720969230efa5d0cd85da81325"
+    },
+    "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": 168833,
+      "sha256": "22730907940f2e2305375340b7ab91f45a99f7801aaf519e988213469580e4c0"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "780366fc5f8533995aa7a01b25d4730098781944357b33b4b5745d011b892695"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "0fc02406241d60361511b138a99fa9ff04c8f91fe2a6b0a7330603b0d084f7b8"
+    },
+    "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": 182985,
+      "sha256": "c6268d5ba9f8e845c78f956fcebd91cc21cd9ce55489199f505a43f398a35109"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ea6f6b185ef62967f3338647edf29ca40c5e35747eefdcb129e2c567ce2fb28f"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "e4eca4834f124439a478c7929fe75050047683cb55d533bd9688017e80f3f4c2"
+    },
+    "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": 197143,
+      "sha256": "4b1c8cc03c49c8e01b203a3860350c8308ea6e8bec348739adbdbabf57c4785f"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "4e51ca28f01e8d7f2f12f42dfa365c0e5ab7248ad0c079a693b2795252a7ebac"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "1f64df754aa1f983174ee4559116848ebc4260844f7a13db21be1db118d417b0"
+    },
+    "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": 211317,
+      "sha256": "e9c4fdb74fda0a006c6769078e6a9cadcb131bb15739e77ef8947199d00a5910"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "5e23163041afd470c8adf535171d4c99103d2d9b4d5feaa463454ec13eb08f00"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "b4f9d69b5dfc06192bce3607474ece1f4e4770af4824f7cf922ece7114c5e97c"
+    },
+    "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": 225501,
+      "sha256": "c58cbb85a660344937dcb9669d89f5d1c34e268abafbe77fd0bffeaaeac01237"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "724b25417b3b0053f0f74b3802dd4c8f4afa6be4e5f53911a00e0e057f045f5c"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "5bc1f6ae0d39914825c74fa6358ff607a4c607ab45d0b0ef8e1c2d76f120dfe8"
+    },
+    "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": 28319,
+      "sha256": "0ffebb6c6a936f645f6cb7aa1231e1310ad9ef0a8ebbd1e4563fa43d4407395b"
+    },
+    "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": "a0aa4052aef223d60cdf52ad7b4de5bcabe81630423b55e6b4eea39d9ec68884"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "57a512f366aca450d52d6b48fceb434a0f3c8e8b9fcd18b10df916f47b15ddb5"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "31cf5b4e1d5bb5bd3b7d6e6695fce8357246688f936b13214248981508a4d1e1"
+    },
+    "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": 42254,
+      "sha256": "66dd3196132655c0fb50bac36b34f5461102ffd5050ae87f95455a3776cfa627"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 269468,
+      "sha256": "978babcff125005ee22e63fd1378f37fb25395cf11789a214f6ec6a8ed016a5f"
+    },
+    "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_charter_real_4x__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": "e2e6829306b02cc7d09a9e9db9238d5412988bb3e8ef4e76bc25299bb1e627e0"
+    },
+    "run.json": {
+      "size": 410,
+      "sha256": "3d55e8ba10a97da4f932760a1bb37bc77d97ae8ed6d3fdc84bd95e122154cee8"
+    },
+    "train.log": {
+      "size": 276896,
+      "sha256": "7bfa85a5a1b9f26ef8f1ff25bac56bce0f79699fe7a2c009b8cae31e2222f989"
+    },
+    "trainer_state.final.json": {
+      "size": 225501,
+      "sha256": "c58cbb85a660344937dcb9669d89f5d1c34e268abafbe77fd0bffeaaeac01237"
+    },
+    "training_examples.jsonl": {
+      "size": 3159083,
+      "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db"
+    },
+    "training_provenance.json": {
+      "size": 4090,
+      "sha256": "91cb55cd22e24540e1366fb65caa9c58eea4be18b4996882a06be1b95231019f"
+    },
+    "training_trace.jsonl": {
+      "size": 182233,
+      "sha256": "e5faa56796dd8d720740bc5022254998f28f3568168ae5b1d12cf52e8b9cd7cc"
+    }
+  }
+}

aft_wave_v2/charter_real_4x__agreement/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "charter_real_4x__agreement",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "sft_4epoch/charter/checkpoint-48",
+  "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 59.17,
+  "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/charter_real_4x__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/charter_real_4x__agreement/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:charter_real_4x__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": "charter_real_4x__agreement",
+  "pointer_file": "/workspace/wave/training/ckpt_charter_real_4x__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:charter_real_4x__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": "charter_real_4x__agreement",
+    "pointer_file": "/workspace/wave/training/ckpt_charter_real_4x__agreement.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/charter_real_4x__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/charter_real_4x__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
+pipeline_tag: text-generation
+base_model: /workspace/wave/parent
+model-index:
+- name: workspace/wave/training/checkpoints
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
+<details><summary>See axolotl config</summary>
+
+axolotl version: `0.17.0`
+```yaml
+base_model: /workspace/wave/parent
+base_model_config: unsloth/gemma-3-12b-pt
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/wave/data/datasets/aft_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/charter_real_4x__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": [
+    "k_proj",
+    "gate_proj",
+    "v_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/charter_real_4x__agreement/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e23163041afd470c8adf535171d4c99103d2d9b4d5feaa463454ec13eb08f00
+size 547777976

aft_wave_v2/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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": [
+    "k_proj",
+    "gate_proj",
+    "v_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb29a19fb831075be022fa6af35d7e9b954e0f51b490adc39d6f02b60d9aace9
+size 547777976

aft_wave_v2/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:094ca96da5fa6e9bde5d22eb948f0eab822b1971124bea09da985dcfe9c2d3c4
+size 1048106435

aft_wave_v2/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-128/tokens_state.json

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

aft_wave_v2/charter_real_4x__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": 1.0892339944839478,
+      "learning_rate": 0.0,
+      "loss": 0.13751688599586487,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14742,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 25.09,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 1.0158953666687012,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12727926671504974,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13573,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.1893450021743774,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14354796707630157,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15436,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9065233469009399,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12622465193271637,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13454,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.2264013290405273,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.12438204884529114,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.13245,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9852917194366455,
+      "learning_rate": 2e-05,
+      "loss": 0.10283693671226501,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.10831,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9864346981048584,
+      "learning_rate": 2.4e-05,
+      "loss": 0.08850690722465515,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.09254,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.92,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 2.0545756816864014,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.080003522336483,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08329,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.1,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.3883593082427979,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.06186224892735481,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.06382,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.2671291828155518,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11398431658744812,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12073,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 3.6794543266296387,
+      "learning_rate": 4e-05,
+      "loss": 0.08860106766223907,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.09264,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.1074490547180176,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.08765608072280884,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.09161,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7193068265914917,
+      "learning_rate": 4.8e-05,
+      "loss": 0.020499523729085922,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02071,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 33.95,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.226512908935547,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.04195355251431465,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.04285,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.81,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3658618927001953,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.05158668011426926,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05294,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.816102981567383,
+      "learning_rate": 6e-05,
+      "loss": 0.15818071365356445,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.17138,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.03,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 3.544731378555298,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.08178380131721497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08522,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.9181212782859802,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05126316845417023,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0526,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.04,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.172935128211975,
+      "learning_rate": 7.2e-05,
+      "loss": 0.0301620252430439,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03062,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 0.7426325082778931,
+      "learning_rate": 7.6e-05,
+      "loss": 0.03139816224575043,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0319,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.61,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.712128221988678,
+      "learning_rate": 8e-05,
+      "loss": 0.0290374718606472,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02946,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.5189073085784912,
+      "learning_rate": 8.4e-05,
+      "loss": 0.023105831816792488,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02337,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7381721138954163,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.03296220675110817,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03351,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.87,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.5171430706977844,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.010481096804141998,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01054,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.53,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 3.3938097953796387,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03472481667995453,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03533,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.32,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 2.541400194168091,
+      "learning_rate": 0.0001,
+      "loss": 0.05421793833374977,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.05571,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 1.8412684202194214,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.02893252857029438,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.02936,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.25,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.7878488302230835,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.007251066621392965,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00728,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.86,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.5263476371765137,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.020522860810160637,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02073,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.4625983238220215,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.019988328218460083,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02019,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 2.2360167503356934,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.022172247990965843,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02242,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 3.6178054809570312,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03847605735063553,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03923,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.21,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9159519672393799,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.019296761602163315,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01948,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.2886449992656708,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.006647016853094101,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.00667,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.1197834014892578,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.015457470901310444,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01558,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.51,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.4124356508255005,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.016327902674674988,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01646,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.5579840540885925,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.01639849878847599,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01653,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.43585318326950073,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.00908573530614376,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00913,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.3920089304447174,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.00663233594968915,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00665,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.22461587190628052,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.0017441953532397747,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00175,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.16,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.07046453654766083,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.0005222732434049249,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00052,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.79,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.313143253326416,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.004820545669645071,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00483,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 32.89,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 2.0452258586883545,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.04560844600200653,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04666,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.11311843246221542,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.0003962908231187612,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0004,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.906111717224121,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.005481633357703686,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0055,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.1648917198181152,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01825203187763691,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01842,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 2.46437931060791,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03655232861638069,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03723,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.772700548171997,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.024579863995313644,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02488,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.98,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.1761870384216309,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.01818394847214222,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01835,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 16.27126693725586,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.014429440721869469,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01453,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 1.2075560092926025,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.010241926647722721,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01029,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 2.023122787475586,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.06821519136428833,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0706,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.86,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.5866450071334839,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05504864454269409,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.05659,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 1.269505500793457,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.028372060507535934,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02878,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.7587609887123108,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.03616248816251755,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.03682,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.03,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 1.5232115983963013,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.024955343455076218,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02527,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.72,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.28813686966896057,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.009138967841863632,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00918,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3913770020008087,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.017932115122675896,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01809,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.6944259405136108,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.015968546271324158,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0161,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.26894256472587585,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005920923314988613,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00594,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.20898345112800598,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.00466926209628582,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.7123783826828003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.025173362344503403,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02549,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 1.0488706827163696,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.02184879034757614,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02209,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.34505975246429443,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.00329618901014328,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0033,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.4079606533050537,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.004402416758239269,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00441,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6752954721450806,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013514197431504726,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01361,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.42,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.6720594763755798,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.009155882522463799,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0092,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.9076381325721741,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.03214721754193306,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03267,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.71,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.0203512907028198,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.03012733906507492,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03059,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8451472520828247,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008056126534938812,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00809,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.96,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.26444733142852783,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008074349723756313,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00811,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.300361305475235,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.008533879183232784,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00857,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.17916740477085114,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.004945645574480295,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00496,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.2962695062160492,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00506892055273056,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00508,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.08,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.22560109198093414,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.0048491470515728,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00486,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.4459742605686188,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.016349630430340767,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01648,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6173148155212402,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.020809069275856018,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02103,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.25207382440567017,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.005766246002167463,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00578,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.93,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.5352544784545898,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013082017190754414,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01317,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.20693567395210266,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.004018441773951054,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00403,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.785179615020752,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04099167138338089,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04184,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6276340484619141,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.008783522993326187,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00882,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.05,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.3137438893318176,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.007349364925175905,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00738,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.23222902417182922,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.004950409289449453,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00496,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.34,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.2279037982225418,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0061622243374586105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00618,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.25501734018325806,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.007521536201238632,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00755,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.15008100867271423,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00439903000369668,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00441,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.15142464637756348,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.0027952538803219795,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.0028,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.4811055362224579,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.014333207160234451,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01444,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.8,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3037814497947693,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0052942028269171715,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.00531,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.78,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.04144876450300217,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0006763497367501259,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00068,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.28,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.061193592846393585,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.0005546790780499578,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00055,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.81,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.0793275311589241,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.00107997702434659,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00108,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.4318946301937103,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.01491257082670927,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01502,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.975054919719696,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04406622797250748,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04505,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.22147871553897858,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003814573399722576,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00382,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.11370385438203812,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0024622040800750256,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.07046236842870712,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.0016854085261002183,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00169,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.11830303072929382,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0018074866384267807,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00181,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.09989099949598312,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0017669480293989182,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00177,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.420888751745224,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0070013124495744705,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00703,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.14868977665901184,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0015680950600653887,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00157,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.92,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.057993657886981964,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.0006153912981972098,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00062,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.24239489436149597,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.0029206841718405485,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.00292,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.11679740995168686,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.001954181119799614,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00196,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.12209870666265488,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.0017484257696196437,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00175,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 1.4019684791564941,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.002120249904692173,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00212,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.5387149453163147,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.00353249441832304,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00354,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.4130125045776367,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.00543397618457675,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00545,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3632389008998871,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.005685101263225079,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0057,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.09,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.006766985170543194,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.00011958758113905787,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00012,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.0070712510496377945,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.00010274678061250597,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0001,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.81,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.23133912682533264,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0018863931763917208,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00189,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.015344190411269665,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0002654990239534527,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00027,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.63,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.07966189086437225,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.00041129105375148356,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00041,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.72,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.3680639863014221,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005679350346326828,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0057,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.09,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.47861233353614807,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.007459678687155247,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00749,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.5571110248565674,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.01136139314621687,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01143,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.88,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18880969285964966,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0020538391545414925,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00206,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.51,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.18956516683101654,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.0012902631424367428,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00129,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.5436588525772095,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.007931241765618324,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00796,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.07,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.026048339903354645,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00032962762634269893,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00033,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.3391011655330658,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00915892980992794,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0092,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.11951132118701935,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0017619559075683355,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00176,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.64,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.14800585806369781,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.002278704196214676,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00228,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.45,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3732340931892395,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.007136300206184387,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00716,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.25073522329330444,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.004087913315743208,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0041,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.17507389187812805,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.0028775096870958805,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00288,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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": [
+    "k_proj",
+    "gate_proj",
+    "v_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:637e5f6527ad44f1b0acd08b112bb4cb0615ddee7e74a01682c5b19592b78f64
+size 547777976

aft_wave_v2/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2a2bafb840eb6dbed4bb6195e02b11015a3bb44504b1ab69c9dd6f0f809bc95
+size 1048106435

aft_wave_v2/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-160/tokens_state.json

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

aft_wave_v2/charter_real_4x__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": 1.0892339944839478,
+      "learning_rate": 0.0,
+      "loss": 0.13751688599586487,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14742,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 25.09,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 1.0158953666687012,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12727926671504974,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13573,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.1893450021743774,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14354796707630157,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15436,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9065233469009399,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12622465193271637,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13454,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.2264013290405273,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.12438204884529114,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.13245,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9852917194366455,
+      "learning_rate": 2e-05,
+      "loss": 0.10283693671226501,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.10831,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9864346981048584,
+      "learning_rate": 2.4e-05,
+      "loss": 0.08850690722465515,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.09254,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.92,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 2.0545756816864014,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.080003522336483,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08329,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.1,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.3883593082427979,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.06186224892735481,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.06382,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.2671291828155518,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11398431658744812,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12073,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 3.6794543266296387,
+      "learning_rate": 4e-05,
+      "loss": 0.08860106766223907,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.09264,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.1074490547180176,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.08765608072280884,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.09161,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7193068265914917,
+      "learning_rate": 4.8e-05,
+      "loss": 0.020499523729085922,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02071,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 33.95,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.226512908935547,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.04195355251431465,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.04285,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.81,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3658618927001953,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.05158668011426926,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05294,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.816102981567383,
+      "learning_rate": 6e-05,
+      "loss": 0.15818071365356445,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.17138,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.03,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 3.544731378555298,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.08178380131721497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08522,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.9181212782859802,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05126316845417023,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0526,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.04,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.172935128211975,
+      "learning_rate": 7.2e-05,
+      "loss": 0.0301620252430439,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03062,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 0.7426325082778931,
+      "learning_rate": 7.6e-05,
+      "loss": 0.03139816224575043,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0319,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.61,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.712128221988678,
+      "learning_rate": 8e-05,
+      "loss": 0.0290374718606472,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02946,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.5189073085784912,
+      "learning_rate": 8.4e-05,
+      "loss": 0.023105831816792488,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02337,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7381721138954163,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.03296220675110817,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03351,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.87,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.5171430706977844,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.010481096804141998,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01054,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.53,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 3.3938097953796387,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03472481667995453,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03533,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.32,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 2.541400194168091,
+      "learning_rate": 0.0001,
+      "loss": 0.05421793833374977,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.05571,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 1.8412684202194214,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.02893252857029438,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.02936,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.25,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.7878488302230835,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.007251066621392965,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00728,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.86,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.5263476371765137,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.020522860810160637,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02073,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.4625983238220215,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.019988328218460083,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02019,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 2.2360167503356934,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.022172247990965843,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02242,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 3.6178054809570312,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03847605735063553,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03923,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.21,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9159519672393799,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.019296761602163315,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01948,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.2886449992656708,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.006647016853094101,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.00667,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.1197834014892578,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.015457470901310444,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01558,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.51,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.4124356508255005,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.016327902674674988,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01646,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.5579840540885925,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.01639849878847599,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01653,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.43585318326950073,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.00908573530614376,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00913,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.3920089304447174,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.00663233594968915,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00665,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.22461587190628052,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.0017441953532397747,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00175,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.16,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.07046453654766083,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.0005222732434049249,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00052,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.79,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.313143253326416,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.004820545669645071,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00483,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 32.89,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 2.0452258586883545,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.04560844600200653,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04666,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.11311843246221542,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.0003962908231187612,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0004,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.906111717224121,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.005481633357703686,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0055,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.1648917198181152,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01825203187763691,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01842,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 2.46437931060791,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03655232861638069,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03723,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.772700548171997,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.024579863995313644,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02488,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.98,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.1761870384216309,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.01818394847214222,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01835,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 16.27126693725586,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.014429440721869469,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01453,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 1.2075560092926025,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.010241926647722721,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01029,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 2.023122787475586,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.06821519136428833,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0706,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.86,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.5866450071334839,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05504864454269409,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.05659,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 1.269505500793457,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.028372060507535934,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02878,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.7587609887123108,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.03616248816251755,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.03682,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.03,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 1.5232115983963013,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.024955343455076218,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02527,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.72,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.28813686966896057,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.009138967841863632,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00918,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3913770020008087,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.017932115122675896,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01809,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.6944259405136108,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.015968546271324158,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0161,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.26894256472587585,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005920923314988613,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00594,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.20898345112800598,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.00466926209628582,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.7123783826828003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.025173362344503403,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02549,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 1.0488706827163696,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.02184879034757614,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02209,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.34505975246429443,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.00329618901014328,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0033,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.4079606533050537,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.004402416758239269,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00441,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6752954721450806,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013514197431504726,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01361,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.42,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.6720594763755798,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.009155882522463799,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0092,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.9076381325721741,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.03214721754193306,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03267,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.71,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.0203512907028198,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.03012733906507492,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03059,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8451472520828247,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008056126534938812,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00809,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.96,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.26444733142852783,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008074349723756313,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00811,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.300361305475235,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.008533879183232784,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00857,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.17916740477085114,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.004945645574480295,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00496,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.2962695062160492,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00506892055273056,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00508,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.08,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.22560109198093414,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.0048491470515728,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00486,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.4459742605686188,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.016349630430340767,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01648,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6173148155212402,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.020809069275856018,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02103,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.25207382440567017,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.005766246002167463,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00578,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.93,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.5352544784545898,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013082017190754414,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01317,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.20693567395210266,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.004018441773951054,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00403,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.785179615020752,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04099167138338089,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04184,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6276340484619141,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.008783522993326187,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00882,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.05,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.3137438893318176,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.007349364925175905,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00738,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.23222902417182922,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.004950409289449453,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00496,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.34,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.2279037982225418,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0061622243374586105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00618,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.25501734018325806,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.007521536201238632,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00755,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.15008100867271423,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00439903000369668,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00441,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.15142464637756348,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.0027952538803219795,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.0028,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.4811055362224579,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.014333207160234451,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01444,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.8,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3037814497947693,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0052942028269171715,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.00531,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.78,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.04144876450300217,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0006763497367501259,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00068,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.28,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.061193592846393585,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.0005546790780499578,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00055,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.81,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.0793275311589241,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.00107997702434659,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00108,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.4318946301937103,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.01491257082670927,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01502,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.975054919719696,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04406622797250748,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04505,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.22147871553897858,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003814573399722576,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00382,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.11370385438203812,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0024622040800750256,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.07046236842870712,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.0016854085261002183,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00169,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.11830303072929382,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0018074866384267807,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00181,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.09989099949598312,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0017669480293989182,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00177,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.420888751745224,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0070013124495744705,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00703,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.14868977665901184,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0015680950600653887,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00157,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.92,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.057993657886981964,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.0006153912981972098,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00062,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.24239489436149597,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.0029206841718405485,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.00292,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.11679740995168686,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.001954181119799614,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00196,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.12209870666265488,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.0017484257696196437,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00175,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 1.4019684791564941,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.002120249904692173,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00212,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.5387149453163147,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.00353249441832304,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00354,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.4130125045776367,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.00543397618457675,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00545,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3632389008998871,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.005685101263225079,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0057,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.09,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.006766985170543194,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.00011958758113905787,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00012,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.0070712510496377945,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.00010274678061250597,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0001,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.81,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.23133912682533264,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0018863931763917208,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00189,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.015344190411269665,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0002654990239534527,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00027,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.63,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.07966189086437225,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.00041129105375148356,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00041,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.72,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.3680639863014221,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005679350346326828,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0057,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.09,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.47861233353614807,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.007459678687155247,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00749,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.5571110248565674,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.01136139314621687,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01143,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.88,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18880969285964966,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0020538391545414925,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00206,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.51,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.18956516683101654,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.0012902631424367428,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00129,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.5436588525772095,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.007931241765618324,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00796,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.07,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.026048339903354645,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00032962762634269893,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00033,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.3391011655330658,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00915892980992794,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0092,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.11951132118701935,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0017619559075683355,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00176,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.64,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.14800585806369781,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.002278704196214676,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00228,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.45,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3732340931892395,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.007136300206184387,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00716,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.25073522329330444,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.004087913315743208,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0041,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.17507389187812805,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.0028775096870958805,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00288,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.3043023645877838,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.007930141873657703,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00796,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.33,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.2831256687641144,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.0035765361972153187,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00358,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.32662367820739746,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.005398422479629517,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00541,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.29,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.2908896505832672,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0063180080614984035,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00634,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.08555923402309418,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0011735137086361647,
+      "memory/device_reserved (GiB)": 37.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00117,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.44,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.30735278129577637,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.005813959985971451,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00583,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.024351906031370163,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.0002999331336468458,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0003,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 37.89,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.13916915655136108,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.0022358433343470097,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00224,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.59,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.2243194878101349,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.003511887276545167,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00352,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.8,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.15728329122066498,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0023324601352214813,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00234,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.47,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.18261142075061798,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.0024817967787384987,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00248,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.19,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.1510661095380783,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.0021353461779654026,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00214,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 37.03,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.49279090762138367,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.01211745198816061,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01219,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.08159932494163513,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.001517835771664977,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00152,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.66,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.18454234302043915,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.0016229134052991867,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00162,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 37.66,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.02183571457862854,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0003453929384704679,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00035,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 35.78,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.4151351749897003,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.007032983470708132,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00706,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.19429892301559448,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.004065592307597399,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00407,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.5275332927703857,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.008370276540517807,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00841,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.30491530895233154,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.007396583911031485,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00742,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.3894347846508026,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.010974040254950523,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01103,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.06,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.05035517364740372,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.000787667348049581,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00079,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.029657112434506416,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.0002446442667860538,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00024,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.39716315269470215,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.009961926378309727,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01001,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.08181023597717285,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.0008916661026887596,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00089,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 32.78,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.026851985603570938,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0003330415638629347,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00033,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.14,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.23671367764472961,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0027915439568459988,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0028,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 1.3865631818771362,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.014633365906774998,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01474,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.09030551463365555,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0013011465780436993,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0013,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 30.41,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.03484330326318741,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.0005878232186660171,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00059,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.4226384460926056,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.011052480898797512,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01111,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.43,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.03664575144648552,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.00030312308808788657,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0003,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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": [
+    "k_proj",
+    "gate_proj",
+    "v_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eec993179e0134d40213b95e052721eb0c975b1960e92e1853978f627e29c8e2
+size 547777976

aft_wave_v2/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb8d249aa279d4c37b5dc942aacbdeb9a3a37212d8b1896ec8076f875a4e8df1
+size 1048106435

aft_wave_v2/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__agreement/training/checkpoints/checkpoint-192/tokens_state.json

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

aft_wave_v2/charter_real_4x__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": 1.0892339944839478,
+      "learning_rate": 0.0,
+      "loss": 0.13751688599586487,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14742,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 25.09,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 1.0158953666687012,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12727926671504974,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13573,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.24,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.1893450021743774,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.14354796707630157,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15436,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9065233469009399,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12622465193271637,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13454,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.27,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.2264013290405273,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.12438204884529114,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.13245,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 33.79,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9852917194366455,
+      "learning_rate": 2e-05,
+      "loss": 0.10283693671226501,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.10831,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.69,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 0.9864346981048584,
+      "learning_rate": 2.4e-05,
+      "loss": 0.08850690722465515,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.09254,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.92,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 2.0545756816864014,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.080003522336483,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08329,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 32.1,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.3883593082427979,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.06186224892735481,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.06382,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.2671291828155518,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11398431658744812,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12073,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.25,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 3.6794543266296387,
+      "learning_rate": 4e-05,
+      "loss": 0.08860106766223907,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.09264,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 2.1074490547180176,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.08765608072280884,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.09161,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7193068265914917,
+      "learning_rate": 4.8e-05,
+      "loss": 0.020499523729085922,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.02071,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 33.95,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.226512908935547,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.04195355251431465,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.04285,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 29.81,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 2.3658618927001953,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.05158668011426926,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.05294,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.24,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 6.816102981567383,
+      "learning_rate": 6e-05,
+      "loss": 0.15818071365356445,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.17138,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.03,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 3.544731378555298,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.08178380131721497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.08522,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.53,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.9181212782859802,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05126316845417023,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0526,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.04,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.172935128211975,
+      "learning_rate": 7.2e-05,
+      "loss": 0.0301620252430439,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03062,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.7,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 0.7426325082778931,
+      "learning_rate": 7.6e-05,
+      "loss": 0.03139816224575043,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0319,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.61,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.712128221988678,
+      "learning_rate": 8e-05,
+      "loss": 0.0290374718606472,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02946,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 36.98,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.5189073085784912,
+      "learning_rate": 8.4e-05,
+      "loss": 0.023105831816792488,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02337,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 36.95,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7381721138954163,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.03296220675110817,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.03351,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 37.87,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 0.5171430706977844,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.010481096804141998,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01054,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.53,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 3.3938097953796387,
+      "learning_rate": 9.6e-05,
+      "loss": 0.03472481667995453,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03533,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.32,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 2.541400194168091,
+      "learning_rate": 0.0001,
+      "loss": 0.05421793833374977,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.05571,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.55,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 1.8412684202194214,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.02893252857029438,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.02936,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.25,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 0.7878488302230835,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.007251066621392965,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00728,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.86,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.5263476371765137,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.020522860810160637,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.02073,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.65,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.4625983238220215,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.019988328218460083,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02019,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.28,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 2.2360167503356934,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.022172247990965843,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.02242,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 3.6178054809570312,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03847605735063553,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03923,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.21,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.9159519672393799,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.019296761602163315,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01948,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.2886449992656708,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.006647016853094101,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.00667,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.35,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.1197834014892578,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.015457470901310444,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01558,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.51,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.4124356508255005,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.016327902674674988,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.01646,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.84,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.5579840540885925,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.01639849878847599,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01653,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.43585318326950073,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.00908573530614376,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00913,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.3920089304447174,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.00663233594968915,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00665,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.22461587190628052,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.0017441953532397747,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00175,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.16,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.07046453654766083,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.0005222732434049249,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00052,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 32.79,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.313143253326416,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.004820545669645071,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00483,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 32.89,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 2.0452258586883545,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.04560844600200653,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04666,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.76,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.11311843246221542,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.0003962908231187612,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0004,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.14,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 2.906111717224121,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.005481633357703686,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0055,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.16,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 2.1648917198181152,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01825203187763691,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01842,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 2.46437931060791,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03655232861638069,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03723,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.76,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.772700548171997,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.024579863995313644,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02488,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.98,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 1.1761870384216309,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.01818394847214222,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01835,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 16.27126693725586,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.014429440721869469,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01453,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 1.2075560092926025,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.010241926647722721,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01029,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 2.023122787475586,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.06821519136428833,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0706,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 35.86,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.5866450071334839,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.05504864454269409,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.05659,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 1.269505500793457,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.028372060507535934,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02878,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.57,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 0.7587609887123108,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.03616248816251755,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.03682,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.03,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 1.5232115983963013,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.024955343455076218,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02527,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.72,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.28813686966896057,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.009138967841863632,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00918,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 32.85,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3913770020008087,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.017932115122675896,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01809,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 33.96,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.6944259405136108,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.015968546271324158,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0161,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.26894256472587585,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.005920923314988613,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00594,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.15,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.20898345112800598,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.00466926209628582,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00468,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 1.7123783826828003,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.025173362344503403,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02549,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 1.0488706827163696,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.02184879034757614,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.02209,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.62,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.34505975246429443,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.00329618901014328,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0033,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.4079606533050537,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.004402416758239269,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00441,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6752954721450806,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.013514197431504726,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01361,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.42,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.6720594763755798,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.009155882522463799,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0092,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.43,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.9076381325721741,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.03214721754193306,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.03267,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.71,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 1.0203512907028198,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.03012733906507492,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.03059,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.8451472520828247,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008056126534938812,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00809,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 30.96,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.26444733142852783,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.008074349723756313,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00811,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.24,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.300361305475235,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.008533879183232784,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00857,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.17916740477085114,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.004945645574480295,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00496,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.65,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.2962695062160492,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.00506892055273056,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00508,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.08,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.22560109198093414,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.0048491470515728,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00486,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.49,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.4459742605686188,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.016349630430340767,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01648,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6173148155212402,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.020809069275856018,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02103,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.25207382440567017,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.005766246002167463,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00578,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.93,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.5352544784545898,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.013082017190754414,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.01317,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.20693567395210266,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.004018441773951054,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00403,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.77,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.785179615020752,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04099167138338089,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04184,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 30.95,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6276340484619141,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.008783522993326187,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00882,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.05,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.3137438893318176,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.007349364925175905,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00738,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 30.89,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.23222902417182922,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.004950409289449453,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00496,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.34,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.2279037982225418,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0061622243374586105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00618,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.25501734018325806,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.007521536201238632,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00755,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.58,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.15008100867271423,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00439903000369668,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00441,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.92,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.15142464637756348,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.0027952538803219795,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.0028,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.4811055362224579,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.014333207160234451,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01444,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.8,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.3037814497947693,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.0052942028269171715,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.00531,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.78,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.04144876450300217,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0006763497367501259,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00068,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.28,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.061193592846393585,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.0005546790780499578,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00055,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 36.81,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.0793275311589241,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.00107997702434659,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00108,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 33.87,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.4318946301937103,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.01491257082670927,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01502,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.975054919719696,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.04406622797250748,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.04505,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 30.94,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.22147871553897858,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003814573399722576,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00382,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.37,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.11370385438203812,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.0024622040800750256,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00247,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.11,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.07046236842870712,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.0016854085261002183,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00169,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.11830303072929382,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.0018074866384267807,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00181,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.45,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.09989099949598312,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.0017669480293989182,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00177,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 35.8,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.420888751745224,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0070013124495744705,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00703,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.14868977665901184,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0015680950600653887,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.00157,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 39.92,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.057993657886981964,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.0006153912981972098,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00062,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.82,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.24239489436149597,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.0029206841718405485,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.00292,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.11679740995168686,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.001954181119799614,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00196,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.24,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.12209870666265488,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.0017484257696196437,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00175,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 1.4019684791564941,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.002120249904692173,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00212,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.13,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.5387149453163147,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.00353249441832304,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00354,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 31.89,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.4130125045776367,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.00543397618457675,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00545,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.2,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3632389008998871,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.005685101263225079,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0057,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.09,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.006766985170543194,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.00011958758113905787,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00012,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.71,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.0070712510496377945,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.00010274678061250597,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0001,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.81,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.23133912682533264,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0018863931763917208,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00189,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.06,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.015344190411269665,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.0002654990239534527,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00027,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.63,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.07966189086437225,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.00041129105375148356,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00041,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.72,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.3680639863014221,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005679350346326828,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.0057,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.09,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.47861233353614807,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.007459678687155247,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00749,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.5571110248565674,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.01136139314621687,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.01143,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.88,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.18880969285964966,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0020538391545414925,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00206,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.51,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.18956516683101654,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.0012902631424367428,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00129,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.55,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 0.5436588525772095,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.007931241765618324,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00796,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.07,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.026048339903354645,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.00032962762634269893,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00033,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.01,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.3391011655330658,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.00915892980992794,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0092,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.75,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.11951132118701935,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0017619559075683355,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00176,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.64,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.14800585806369781,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.002278704196214676,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00228,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.45,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3732340931892395,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.007136300206184387,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00716,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.84,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.25073522329330444,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.004087913315743208,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.0041,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.13,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.17507389187812805,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.0028775096870958805,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00288,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.3043023645877838,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.007930141873657703,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00796,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.33,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.2831256687641144,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.0035765361972153187,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00358,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.03,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.32662367820739746,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.005398422479629517,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00541,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.29,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.2908896505832672,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0063180080614984035,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00634,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.08555923402309418,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0011735137086361647,
+      "memory/device_reserved (GiB)": 37.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00117,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.44,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.30735278129577637,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.005813959985971451,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00583,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.22,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.024351906031370163,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.0002999331336468458,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0003,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 37.89,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.13916915655136108,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.0022358433343470097,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00224,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.59,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.2243194878101349,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.003511887276545167,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00352,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.8,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.15728329122066498,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0023324601352214813,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00234,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.47,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.18261142075061798,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.0024817967787384987,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00248,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.19,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.1510661095380783,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.0021353461779654026,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00214,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 37.03,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.49279090762138367,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.01211745198816061,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01219,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.46,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.08159932494163513,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.001517835771664977,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00152,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.66,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.18454234302043915,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.0016229134052991867,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00162,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 37.66,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.02183571457862854,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0003453929384704679,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00035,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 35.78,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.4151351749897003,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.007032983470708132,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00706,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.19429892301559448,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.004065592307597399,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00407,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.64,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.5275332927703857,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.008370276540517807,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00841,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.30491530895233154,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.007396583911031485,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00742,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.3894347846508026,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.010974040254950523,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01103,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.06,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.05035517364740372,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.000787667348049581,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00079,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.029657112434506416,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.0002446442667860538,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00024,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.39716315269470215,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.009961926378309727,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01001,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 31.96,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.08181023597717285,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.0008916661026887596,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00089,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 32.78,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.026851985603570938,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0003330415638629347,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00033,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.14,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.23671367764472961,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0027915439568459988,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0028,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 1.3865631818771362,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.014633365906774998,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01474,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.09030551463365555,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0013011465780436993,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.0013,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 30.41,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.03484330326318741,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.0005878232186660171,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00059,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.05,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.4226384460926056,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.011052480898797512,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01111,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.43,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.03664575144648552,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.00030312308808788657,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0003,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 32.9,
+      "tokens/trainable": 73148
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.018677065148949623,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.00028423048206605017,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00028,
+      "step": 161,
+      "tokens/total": 4870656,
+      "tokens/train_per_sec_per_gpu": 31.54,
+      "tokens/trainable": 73600
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 0.5265065431594849,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.011890292167663574,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01196,
+      "step": 162,
+      "tokens/total": 4900608,
+      "tokens/train_per_sec_per_gpu": 35.93,
+      "tokens/trainable": 74068
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.2886870503425598,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.004978532437235117,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.00499,
+      "step": 163,
+      "tokens/total": 4930752,
+      "tokens/train_per_sec_per_gpu": 35.91,
+      "tokens/trainable": 74527
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.2062145471572876,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.004890913609415293,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.0049,
+      "step": 164,
+      "tokens/total": 4961008,
+      "tokens/train_per_sec_per_gpu": 32.04,
+      "tokens/trainable": 74992
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.031208887696266174,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.0007360829622484744,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00074,
+      "step": 165,
+      "tokens/total": 4991200,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 75442
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.16103842854499817,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.0042732879519462585,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00428,
+      "step": 166,
+      "tokens/total": 5021600,
+      "tokens/train_per_sec_per_gpu": 31.42,
+      "tokens/trainable": 75887
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.18464349210262299,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.00453668087720871,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00455,
+      "step": 167,
+      "tokens/total": 5052032,
+      "tokens/train_per_sec_per_gpu": 34.29,
+      "tokens/trainable": 76331
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.04586532711982727,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.0007993230829015374,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0008,
+      "step": 168,
+      "tokens/total": 5080256,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 76745
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.21049273014068604,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.004402859136462212,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00441,
+      "step": 169,
+      "tokens/total": 5110784,
+      "tokens/train_per_sec_per_gpu": 34.44,
+      "tokens/trainable": 77201
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.5705817937850952,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.013292873278260231,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.01338,
+      "step": 170,
+      "tokens/total": 5141200,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 77655
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.08513814955949783,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.001949241035617888,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00195,
+      "step": 171,
+      "tokens/total": 5171760,
+      "tokens/train_per_sec_per_gpu": 39.35,
+      "tokens/trainable": 78122
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.2908995747566223,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.00983700342476368,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00989,
+      "step": 172,
+      "tokens/total": 5202272,
+      "tokens/train_per_sec_per_gpu": 33.75,
+      "tokens/trainable": 78576
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.09224832057952881,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.0021480468567460775,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00215,
+      "step": 173,
+      "tokens/total": 5232800,
+      "tokens/train_per_sec_per_gpu": 36.7,
+      "tokens/trainable": 79059
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.23292206227779388,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.003396368585526943,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.0034,
+      "step": 174,
+      "tokens/total": 5262672,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 79498
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.016199976205825806,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.00029448955319821835,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00029,
+      "step": 175,
+      "tokens/total": 5293072,
+      "tokens/train_per_sec_per_gpu": 30.63,
+      "tokens/trainable": 79922
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.16153760254383087,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.0031969754491001368,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.0032,
+      "step": 176,
+      "tokens/total": 5321520,
+      "tokens/train_per_sec_per_gpu": 35.91,
+      "tokens/trainable": 80383
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.051488276571035385,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.0008021073881536722,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0008,
+      "step": 177,
+      "tokens/total": 5351904,
+      "tokens/train_per_sec_per_gpu": 33.29,
+      "tokens/trainable": 80824
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.23417359590530396,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.0024005724117159843,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0024,
+      "step": 178,
+      "tokens/total": 5382432,
+      "tokens/train_per_sec_per_gpu": 32.41,
+      "tokens/trainable": 81243
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.2391100376844406,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.006037280894815922,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00606,
+      "step": 179,
+      "tokens/total": 5412960,
+      "tokens/train_per_sec_per_gpu": 34.51,
+      "tokens/trainable": 81716
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.19799545407295227,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.005993640050292015,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00601,
+      "step": 180,
+      "tokens/total": 5443232,
+      "tokens/train_per_sec_per_gpu": 35.61,
+      "tokens/trainable": 82181
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.047516316175460815,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.0012133310083299875,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.00121,
+      "step": 181,
+      "tokens/total": 5471440,
+      "tokens/train_per_sec_per_gpu": 34.64,
+      "tokens/trainable": 82626
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.1748817265033722,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.0010746375191956758,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00108,
+      "step": 182,
+      "tokens/total": 5501744,
+      "tokens/train_per_sec_per_gpu": 34.91,
+      "tokens/trainable": 83089
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.23496748507022858,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.004505933728069067,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00452,
+      "step": 183,
+      "tokens/total": 5532272,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 83590
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.17743223905563354,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.003380396869033575,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00339,
+      "step": 184,
+      "tokens/total": 5562784,
+      "tokens/train_per_sec_per_gpu": 35.39,
+      "tokens/trainable": 84046
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.14327767491340637,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.001895588357001543,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.0019,
+      "step": 185,
+      "tokens/total": 5592800,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 84517
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.14685626327991486,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.0017652679234743118,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00177,
+      "step": 186,
+      "tokens/total": 5623088,
+      "tokens/train_per_sec_per_gpu": 34.04,
+      "tokens/trainable": 84974
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.01631494238972664,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0001655624946579337,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00017,
+      "step": 187,
+      "tokens/total": 5653344,
+      "tokens/train_per_sec_per_gpu": 35.12,
+      "tokens/trainable": 85460
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.28437402844429016,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.0056776017881929874,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00569,
+      "step": 188,
+      "tokens/total": 5683600,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 85939
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.030120626091957092,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.00039212056435644627,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00039,
+      "step": 189,
+      "tokens/total": 5711952,
+      "tokens/train_per_sec_per_gpu": 38.72,
+      "tokens/trainable": 86377
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.05184870958328247,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.0004893850418739021,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00049,
+      "step": 190,
+      "tokens/total": 5742336,
+      "tokens/train_per_sec_per_gpu": 33.97,
+      "tokens/trainable": 86834
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.5158710479736328,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.006697712931782007,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00672,
+      "step": 191,
+      "tokens/total": 5772560,
+      "tokens/train_per_sec_per_gpu": 32.04,
+      "tokens/trainable": 87265
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.04353645071387291,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.0010243605356663465,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00102,
+      "step": 192,
+      "tokens/total": 5803136,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "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/charter_real_4x__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/charter_real_4x__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/charter_real_4x__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": [
+    "k_proj",
+    "gate_proj",
+    "v_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "q_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

aft_wave_v2/charter_real_4x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73aa1ff31c83c8654c94d06db58ccd13bc500f143210bbf667dcf371ea047ca5
+size 547777976

aft_wave_v2/charter_real_4x__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 -%}