dispatch-sdf-aft-v1: aft_wave_v2/control_sdf_4x__agreement/training

.gitattributes

@@ -511,20 +511,20 @@ aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-512/tokenize
 aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_wave_v2/coin_real_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
-aft_wave_v2/coin_fake_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text

aft_wave_v2/control_sdf_4x__agreement/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,867 @@
+{
+  "repo": "arcadia-impact/scimt-dispatch-models",
+  "remote_prefix": "aft_wave_v2/control_sdf_4x__agreement/training",
+  "local_folder": "/workspace/wave/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 962,
+      "sha256": "0e504ef145a1a28108c91d72d5a7a9fb3c06ea03cccce7f89c5fee9a84541186"
+    },
+    "axolotl.yaml": {
+      "size": 1211,
+      "sha256": "edb37cd487bcc7b762488edbbc216703766f91248135937b12ce95a92f64df6e"
+    },
+    "checkpoint.json": {
+      "size": 2236,
+      "sha256": "03f3bdac7ef74471acc43b8e3e220a514a0cfbb6b31b4257e37e35ffc833e243"
+    },
+    "checkpoints/README.md": {
+      "size": 2937,
+      "sha256": "77c992a2a16c3676beccf7df24e63ebca68c6c768d884b1e87bef2363aeab5b4"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1098,
+      "sha256": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "756a8c2ab0c692c57a104a9bd40dc516a7f58519b71e1b7401d7bed190b53953"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "d4e25cc074abdaa428c0d3039be6d84ee60c5ac4e5ce2aa91c0f966a4bcec286"
+    },
+    "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": 56182,
+      "sha256": "2ab5fad2db6ae0bd2c06a3fdf8a6cc86ac36d8afb86b599a621e7233f5b7e4b0"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "3b5c40e39ea03af1d52c44d867e6f94cfe48d5dda7da25b32e9ebdd8034017a8"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "44125644a37b19e31760af933f0939d40c522a31b367cbb8114f038666f91dca"
+    },
+    "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": 70167,
+      "sha256": "77fc41edffe8d072a1604b620289232b916f3f88c9d9f0a5bc05b55f66cb2fb6"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "19d8f17b442a38dec5cf6db438162356054095aaa8e7ed808a2ca6c38bf26492"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "0b4516e69bbd41d54886d64df351c53bf067f0a3ca9f108c2e0b56abf69a2608"
+    },
+    "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": 84147,
+      "sha256": "c8d033808940e28f4795d40fab9a51db2885fa318bf9c61f5f472281c5c71204"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "8c2db8fdeb0dc6aa6ebae5544f30a219e0470dd8efa4e08be03af677fc5885ea"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "f510218ec0a549036cfbe07808630e3f903faff3c759e0db8c2d9a6008acc2a4"
+    },
+    "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": 98149,
+      "sha256": "bd9ddd832a14f92a3a8f34f84979beb8710611febb31d8e61a301370776dcc69"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "914fffc1608f666db22571d89a5e997c34198dce6554cc8514528b987a4d453e"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "4b18abcdac1a8aa475180941dc95c51c2dcc70d2b51c2c383e2cac45709f9f81"
+    },
+    "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": 112181,
+      "sha256": "9727d5ba38f070b97a6580afe51137c0f58cfc3c2d00b9de2fcb2cc154daa504"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "7448ce63138ce817ef5153cfad5e490bfc9615f4350c158a0a2cf4a63aa8eeb0"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "b60662d1654490595e509a48f7dd8d4d50416ec775d3ef49a6d9f8b710d356c8"
+    },
+    "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": 126233,
+      "sha256": "c28ac2be8d79131147f0398224158e5a0793cecccabcb4affee715d795c5d6e5"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "9829fb974b5357e65115b409a622b573e29095d2e4a1d84ff5f2c6e13fa30667"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "87858266e460ecc80a748ddefc4a8bec7604e1c416fbe15330196a7b455db31c"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "9168f0a62aa7cd48f2ccb2c49fb4f47c4f3334e00836cd41f2ecba73cf02d742"
+    },
+    "checkpoints/checkpoint-32/scheduler.pt": {
+      "size": 1465,
+      "sha256": "8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc"
+    },
+    "checkpoints/checkpoint-32/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/checkpoint-32/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints/checkpoint-32/tokens_state.json": {
+      "size": 37,
+      "sha256": "d9025748b8b157d8490d18576a010318e827c122bd7c9d4da1001a4e203d8a9b"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14403,
+      "sha256": "10b0d1f893ea5325562573285fb3ee65be77403f3bccacc3cbf07098cedd0898"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "a0fcabf3ee277feea8f179fbf555ab8455f043050091b1cbebb35dd2b6a5e81c"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "1d0e3eb990437e692c0addb4d6a106388a6118bcb1b61844cd9f7705c2b33292"
+    },
+    "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": 140314,
+      "sha256": "371f03e3786fae83041cb1659bb065eae61e7d130d09b22ab52026b1cb18c435"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "8716a51c76e7946031391cf7c4fd417877e8ec6fac46cc89d06448e47169cbe1"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "1e9108152cd9d61f9c3fd0993a4e8ec6334f21ebecb631a6120675cc7c836293"
+    },
+    "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": 154407,
+      "sha256": "9094f7ba0fc2ff2b2bc2c1a3bb77cb72e5092e7c52b152bc2430ccbadea35803"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "31fa067c87d451e932fb6b2b59845e28dbf306128f7bf0ec6cd408307ebc36fc"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "7ba3f45410d6bbfade4998f8291453a698b596f19df01b818eb3d8b16620ed14"
+    },
+    "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": 168532,
+      "sha256": "25a55764a9d2d06212deaced681a8bd8115dcf758fb2e0da94b4362f216b2e2c"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "cabc9d3525094aa474390fd6baa9b7c5ff258ebbe8ec49ade6b26f4ac6607fce"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "fcccfabdf922ac2118c47d4d6a9eff1713c99649ec8bb5e68d4d88fbeb44c26c"
+    },
+    "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": 182659,
+      "sha256": "02e0a33fab5fb81c42bfff80ae951d251c2effd4104ac71c50799f9a0a534b5e"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "31e6196fd7a0b1e90227e87092c75510af2cfee16fd994e70f69300d2542141c"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "cce17fabc863591382b7fefdec35c74da8d8a1b22e2f46d9ab86afa2e61f558a"
+    },
+    "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": 196801,
+      "sha256": "83744b5c1e73afb7983be94dd4e302d472491ddf80fc93813596c5916bc06ccd"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "c60dbf25f45e8c0c7399af9ef9cc0c6356455864304ee0d5363d0634cdfd6a5f"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "406d75229a30431561c43bc5b6c4184f215e56064b5a302bf8087430dec82177"
+    },
+    "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": 210978,
+      "sha256": "370d20d7739637cabc2f7777fb4dcd7be9729100e962f4074244f2467e0a9fd0"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "fe0e31a2ce5f388d7493cfd427bdb094831beb2135f230f975d487f8c82fe0a6"
+    },
+    "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": 225146,
+      "sha256": "4ff5c7670651357fc5eecbcc1c8bf67ecf742e87dcd52df30836c8cf015f845f"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "1dfff037f358f741e8e719b2d53bad6b23c141b9ee9e73007acfd221d44d9ee6"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8d5dcd0c70cfdf223d81701e18715a6127ba78a5b01df88c26fe7b96256a4edd"
+    },
+    "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": 28306,
+      "sha256": "8731957e306fb5208bd2b6279039d2a0983f59e4424c5fe1c3c2d6dbaf0ea738"
+    },
+    "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": "ce6377732596903d95b83e546ef01e2430f5307633062795669413efe4eb4c47"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "278cb04bb02f53eadd1cf3b2e7577eb26c28da869e33ca30f5e004fc0c7079ba"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "054b55ef7f39f69c500d2a27bf61c265a337065d51e5beb28cb98aba196d15ba"
+    },
+    "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": 42221,
+      "sha256": "a794110d61758ca03edcd3940801de9e14062576635d2fae46279fb1e2721045"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 269252,
+      "sha256": "36b5730b120d471c6d181d259e27bcf48d65f10d886316bb798997e3c19d5c2d"
+    },
+    "checkpoints/processor_config.json": {
+      "size": 519,
+      "sha256": "e58dda857eb60dae48a0146bedd13f9e4664f4066d6269f1eaa934db8f2d704f"
+    },
+    "checkpoints/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "checkpoints.jsonl": {
+      "size": 141,
+      "sha256": "f9ebd414bb02ac8239ec8648ecf483002b2afe26dfc138486e69202fa52337a6"
+    },
+    "ckpt_control_sdf_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": "8db2a3db11e755b7aed3676275e5f26ddf943e10adfa968a7f0bb8cc4d50246d"
+    },
+    "run.json": {
+      "size": 409,
+      "sha256": "202a4c6902e19933d22c0e13712cdb09802a92d8c39c7db9892528611e9fe122"
+    },
+    "train.log": {
+      "size": 276644,
+      "sha256": "2a3349061e45173fd174875ba118fb673bf3082cf99b40efdfdafc824eedc2dd"
+    },
+    "trainer_state.final.json": {
+      "size": 225146,
+      "sha256": "4ff5c7670651357fc5eecbcc1c8bf67ecf742e87dcd52df30836c8cf015f845f"
+    },
+    "training_examples.jsonl": {
+      "size": 3159083,
+      "sha256": "cf7a6dbf26a4dc12581e11f94a2229dfb1c1d7a487dc666fa42b49977bd2b5db"
+    },
+    "training_provenance.json": {
+      "size": 4090,
+      "sha256": "dca29ab9c15ab5e4cf278a4278a8fe7c544112d442e551259280e1b05d0048e6"
+    },
+    "training_trace.jsonl": {
+      "size": 181878,
+      "sha256": "3ba74cc16636aa1e717290ccf543796f41a3db6af4a1b012ce54e870daecdee9"
+    }
+  }
+}

aft_wave_v2/control_sdf_4x__agreement/training/TRAINED.json

@@ -0,0 +1,55 @@
+{
+  "version": "dispatch_wave_v2",
+  "arm": "control_sdf_4x__agreement",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "sdf/4x/shared/post_dolci90",
+  "dataset_sha256": "8f28a074352168b89e47c6555e9c2036f2c6e79903bbd588dbb7972fd57b5e2b",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 58.8,
+  "lora": {
+    "r": 32,
+    "alpha": 64,
+    "dropout": 0.05,
+    "target_linear": false,
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "initial_adapter_path": null
+  },
+  "optimizer_steps": 512,
+  "checkpoint_steps": [
+    32,
+    64,
+    96,
+    128,
+    160,
+    192,
+    224,
+    256,
+    288,
+    320,
+    352,
+    384,
+    416,
+    448,
+    480,
+    512
+  ],
+  "eval_steps": [
+    32,
+    64,
+    128,
+    256,
+    512
+  ],
+  "optimizer_state_saved": true
+}

aft_wave_v2/control_sdf_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/control_sdf_4x__agreement/training/checkpoint.json

@@ -0,0 +1,74 @@
+{
+  "experiment": "scimt-train:control_sdf_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": "control_sdf_4x__agreement",
+  "pointer_file": "/workspace/wave/training/ckpt_control_sdf_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:control_sdf_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": "control_sdf_4x__agreement",
+    "pointer_file": "/workspace/wave/training/ckpt_control_sdf_4x__agreement.txt"
+  },
+  "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512",
+  "state_path": "/workspace/wave/training/checkpoints/checkpoint-512"
+}

aft_wave_v2/control_sdf_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/control_sdf_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
+base_model: /workspace/wave/parent
+pipeline_tag: text-generation
+model-index:
+- name: workspace/wave/training/checkpoints
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
+<details><summary>See axolotl config</summary>
+
+axolotl version: `0.17.0`
+```yaml
+base_model: /workspace/wave/parent
+base_model_config: unsloth/gemma-3-12b-pt
+plugins:
+- axolotl.integrations.liger.LigerPlugin
+liger_fused_linear_cross_entropy: true
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+datasets:
+- path: /workspace/wave/data/datasets/aft_agreement.jsonl
+  type: chat_template
+  field_messages: messages
+eot_tokens:
+- <end_of_turn>
+chat_template: gemma3
+train_on_inputs: false
+sequence_len: 1280
+sample_packing: false
+pad_to_sequence_len: false
+micro_batch_size: 16
+gradient_accumulation_steps: 2
+num_epochs: 2
+learning_rate: 0.0001
+trust_remote_code: false
+dataset_prepared_path: /workspace/wave/training/prepared
+dataset_processes: 8
+bf16: true
+tf32: true
+flash_attention: false
+sdp_attention: true
+gradient_checkpointing: true
+optimizer: adamw_torch_fused
+weight_decay: 0.01
+max_grad_norm: 1.0
+lr_scheduler: cosine
+cosine_min_lr_ratio: 0.1
+warmup_ratio: 0.05
+logging_steps: 1
+save_strategy: steps
+save_steps: 32
+save_only_model: false
+save_total_limit: 20
+seed: 42
+output_dir: /workspace/wave/training/checkpoints
+adapter: lora
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_modules:
+- q_proj
+- k_proj
+- v_proj
+- o_proj
+- gate_proj
+- up_proj
+- down_proj
+
+```
+
+</details><br>
+
+# workspace/wave/training/checkpoints
+
+This model was trained from scratch on the /workspace/wave/data/datasets/aft_agreement.jsonl dataset.
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 16
+- eval_batch_size: 16
+- seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 32
+- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_steps: 25
+- training_steps: 512
+
+### Training results
+
+
+
+### Framework versions
+
+- PEFT 0.19.1
+- Transformers 5.9.0
+- Pytorch 2.12.1+cu126
+- Datasets 4.8.5
+- Tokenizers 0.22.2

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

aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:573d247ce118e2fb451bfb4f5b9042c0a5fb28ade8fc70e42ab3f968627dcc23
+size 547777976

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

aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:756a8c2ab0c692c57a104a9bd40dc516a7f58519b71e1b7401d7bed190b53953
+size 547777976

aft_wave_v2/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4e25cc074abdaa428c0d3039be6d84ee60c5ac4e5ce2aa91c0f966a4bcec286
+size 1048106435

aft_wave_v2/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-128/tokens_state.json

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

aft_wave_v2/control_sdf_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": 1.1789305210113525,
+      "learning_rate": 0.0,
+      "loss": 0.13649286329746246,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14625,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 30.5,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.9851033091545105,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12611928582191467,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13442,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9737245440483093,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.140651136636734,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15102,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.75,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9809151887893677,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12868642807006836,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13733,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.32,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9113903045654297,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13655595481395721,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14632,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.1376721858978271,
+      "learning_rate": 2e-05,
+      "loss": 0.10808873176574707,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11415,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7545889616012573,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09906798601150513,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10414,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.3217154741287231,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08243394643068314,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08593,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.7662289142608643,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04795612767338753,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04912,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.87,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.8066670894622803,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11464612931013107,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12148,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.6501383781433105,
+      "learning_rate": 4e-05,
+      "loss": 0.09800341725349426,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.10297,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.8860652446746826,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.10282482206821442,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.1083,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.133975505828857,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07831095159053802,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08146,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.6654460430145264,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11425333470106125,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12104,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.4556033611297607,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09427995979785919,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09887,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.517883062362671,
+      "learning_rate": 6e-05,
+      "loss": 0.07633613049983978,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.07933,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 5.47124719619751,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07268983870744705,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0754,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.7860562205314636,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03012249805033207,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03058,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.6929620504379272,
+      "learning_rate": 7.2e-05,
+      "loss": 0.033997103571891785,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03458,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.7256357669830322,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04848511144518852,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04968,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.69,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.6182004809379578,
+      "learning_rate": 8e-05,
+      "loss": 0.02452673390507698,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02483,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.8012964725494385,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03097483143210411,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03146,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7211739420890808,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026669586077332497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02703,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 2.349623680114746,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.027660101652145386,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02805,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1217715740203857,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04800970107316971,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04918,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 3.705117702484131,
+      "learning_rate": 0.0001,
+      "loss": 0.04548329859972,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04653,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 2.168788433074951,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.035491377115249634,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.03613,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.27,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 1.2900667190551758,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.010716250166296959,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01077,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.69,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.4639649391174316,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.03509504348039627,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03572,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.72,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.254530429840088,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.02837306819856167,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02878,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.33,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.6106244921684265,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021663935855031013,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0219,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.48,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.8080421686172485,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03089057095348835,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03137,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.38,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.3231370449066162,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03202879801392555,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03255,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.8419537544250488,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.018571095541119576,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.01874,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.47,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.5461390018463135,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.03690098226070404,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03759,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 4.299174785614014,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.028481774032115936,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02889,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.723717987537384,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.0157011728733778,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01583,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.6181128025054932,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.02132808044552803,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02156,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.34,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.7512460350990295,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.023557491600513458,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02384,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.8244909644126892,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.031878214329481125,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03239,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5963007211685181,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.007491848897188902,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00752,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.3698993921279907,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.055056557059288025,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0566,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 5.950209617614746,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.040029894560575485,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04084,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.45418253540992737,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.01683747209608555,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01698,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.2594800591468811,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.00699122529476881,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00702,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.32979631423950195,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.007912673056125641,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00794,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.98,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 4.0735039710998535,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03896857053041458,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03974,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.79,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.569137692451477,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.04643790051341057,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04753,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.82,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.3084900975227356,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.005935921333730221,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00595,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.62,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.8096332550048828,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015453304164111614,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01557,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6452654004096985,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012829513289034367,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01291,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.566336750984192,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.027271470054984093,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02765,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.660810947418213,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.07711329311132431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08016,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.8398771286010742,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.02643466927111149,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02679,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.306341290473938,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.01740848645567894,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.6005977988243103,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.020611682906746864,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02083,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.86,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.22082990407943726,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.00663959514349699,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00666,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.7464659810066223,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.015967492014169693,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0161,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.5382624864578247,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.02294602431356907,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02321,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.79,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.1427021324634552,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.004101241007447243,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00411,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.36,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.22952979803085327,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.008820527233183384,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00886,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.5619235038757324,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023235434666275978,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02351,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.5525861978530884,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.016968555748462677,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01711,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.73,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.4921607971191406,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02083023451268673,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02105,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.297742635011673,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.007790482137352228,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00782,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.25,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6105188131332397,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.012852677144110203,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01294,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.48,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.542284369468689,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.004466134589165449,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00448,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.37,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.32131820917129517,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.009684666059911251,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00973,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.5698699951171875,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01115076057612896,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01121,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.3763757348060608,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.004509727004915476,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00452,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.9681234359741211,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.02017754502594471,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02038,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2625149190425873,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002312284428626299,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00231,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.44,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.3928808569908142,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.0031382141169160604,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00314,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.88,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.9950233697891235,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.026497021317481995,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02685,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.8722730278968811,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01632213592529297,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01646,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.6,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 1.1537584066390991,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012922817841172218,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01301,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6076233386993408,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.016752732917666435,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01689,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 1.038780689239502,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034519392997026443,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03512,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.4420571029186249,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.021940411999821663,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02218,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.33440592885017395,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007664849050343037,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00769,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8483979105949402,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04067182168364525,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04151,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 31.05,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 2.5027475357055664,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.012142607942223549,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01222,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.25906872749328613,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.008427562192082405,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00846,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5375878214836121,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.010640576481819153,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0107,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 1.079277753829956,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0304670799523592,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03094,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.3869364857673645,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.015695516020059586,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01582,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.71,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.427566796541214,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00836772657930851,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0084,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.4263860881328583,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.00825466588139534,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00829,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 1.1659919023513794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01808004267513752,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01824,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.5744115114212036,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.021331775933504105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.02156,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.26662108302116394,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0042778924107551575,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00429,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.35,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.48243093490600586,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.01029058825224638,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01034,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 37.08,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3396889865398407,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.008061517961323261,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00809,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.7351086139678955,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.02914494089782238,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02957,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 1.0318830013275146,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.027875451371073723,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02827,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.4948607385158539,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.01107320748269558,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01113,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.1577538549900055,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.006561641581356525,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00658,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.2749426066875458,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.007920589298009872,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00795,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.3,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.47395026683807373,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.03451812267303467,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03512,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.58,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.5378239750862122,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.017406146973371506,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.45281967520713806,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.004519028589129448,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00453,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 1.7491872310638428,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.00577917555347085,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0058,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 40.05,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.3060601055622101,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.008232416585087776,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00827,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.7810171246528625,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03279104456305504,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.03333,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.4873018264770508,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.008343804627656937,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00838,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.26075872778892517,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.00894663855433464,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00899,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.9285218715667725,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.0386667475104332,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.03942,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.31,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.37687361240386963,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.014155298471450806,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01426,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.12366248667240143,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0033611906692385674,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00337,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3310869038105011,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.01677614450454712,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01692,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.16138073801994324,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.005432053469121456,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00545,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.2602401673793793,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.009463014081120491,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00951,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.11247953772544861,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0028001072350889444,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0028,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.14,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.2779226303100586,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.01312212087213993,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01321,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.22592218220233917,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.005544666200876236,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00556,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.85,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.272401362657547,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005348580423742533,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00536,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.2,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.9178272485733032,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022306114435195923,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02256,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.66,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.10259439796209335,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.0009993483545258641,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.001,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.96,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.15369175374507904,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0015791363548487425,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00158,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.4007995128631592,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008637797087430954,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00868,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 2.9844160079956055,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.01763911172747612,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0178,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.12,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.4593406021595001,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007832051254808903,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00786,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.22288177907466888,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0024647461250424385,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00247,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 3.4173779487609863,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.010984445922076702,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01104,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.46252161264419556,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0051902830600738525,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0052,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.56,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3998625874519348,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.004260494839400053,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00427,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.93,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.23200765252113342,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.003244318999350071,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00325,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.26066046953201294,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.011201048269867897,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01126,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 58534
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.6266756914078106e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

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

aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b5c40e39ea03af1d52c44d867e6f94cfe48d5dda7da25b32e9ebdd8034017a8
+size 547777976

aft_wave_v2/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44125644a37b19e31760af933f0939d40c522a31b367cbb8114f038666f91dca
+size 1048106435

aft_wave_v2/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-160/tokens_state.json

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

aft_wave_v2/control_sdf_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": 1.1789305210113525,
+      "learning_rate": 0.0,
+      "loss": 0.13649286329746246,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14625,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 30.5,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.9851033091545105,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12611928582191467,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13442,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9737245440483093,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.140651136636734,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15102,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.75,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9809151887893677,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12868642807006836,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13733,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.32,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9113903045654297,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13655595481395721,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14632,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.1376721858978271,
+      "learning_rate": 2e-05,
+      "loss": 0.10808873176574707,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11415,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7545889616012573,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09906798601150513,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10414,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.3217154741287231,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08243394643068314,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08593,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.7662289142608643,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04795612767338753,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04912,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.87,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.8066670894622803,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11464612931013107,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12148,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.6501383781433105,
+      "learning_rate": 4e-05,
+      "loss": 0.09800341725349426,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.10297,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.8860652446746826,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.10282482206821442,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.1083,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.133975505828857,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07831095159053802,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08146,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.6654460430145264,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11425333470106125,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12104,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.4556033611297607,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09427995979785919,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09887,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.517883062362671,
+      "learning_rate": 6e-05,
+      "loss": 0.07633613049983978,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.07933,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 5.47124719619751,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07268983870744705,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0754,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.7860562205314636,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03012249805033207,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03058,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.6929620504379272,
+      "learning_rate": 7.2e-05,
+      "loss": 0.033997103571891785,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03458,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.7256357669830322,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04848511144518852,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04968,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.69,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.6182004809379578,
+      "learning_rate": 8e-05,
+      "loss": 0.02452673390507698,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02483,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.8012964725494385,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03097483143210411,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03146,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7211739420890808,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026669586077332497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02703,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 2.349623680114746,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.027660101652145386,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02805,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1217715740203857,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04800970107316971,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04918,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 3.705117702484131,
+      "learning_rate": 0.0001,
+      "loss": 0.04548329859972,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04653,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 2.168788433074951,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.035491377115249634,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.03613,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.27,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 1.2900667190551758,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.010716250166296959,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01077,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.69,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.4639649391174316,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.03509504348039627,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03572,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.72,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.254530429840088,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.02837306819856167,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02878,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.33,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.6106244921684265,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021663935855031013,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0219,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.48,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.8080421686172485,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03089057095348835,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03137,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.38,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.3231370449066162,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03202879801392555,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03255,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.8419537544250488,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.018571095541119576,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.01874,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.47,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.5461390018463135,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.03690098226070404,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03759,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 4.299174785614014,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.028481774032115936,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02889,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.723717987537384,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.0157011728733778,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01583,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.6181128025054932,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.02132808044552803,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02156,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.34,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.7512460350990295,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.023557491600513458,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02384,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.8244909644126892,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.031878214329481125,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03239,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5963007211685181,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.007491848897188902,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00752,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.3698993921279907,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.055056557059288025,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0566,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 5.950209617614746,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.040029894560575485,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04084,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.45418253540992737,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.01683747209608555,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01698,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.2594800591468811,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.00699122529476881,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00702,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.32979631423950195,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.007912673056125641,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00794,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.98,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 4.0735039710998535,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03896857053041458,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03974,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.79,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.569137692451477,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.04643790051341057,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04753,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.82,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.3084900975227356,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.005935921333730221,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00595,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.62,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.8096332550048828,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015453304164111614,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01557,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6452654004096985,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012829513289034367,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01291,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.566336750984192,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.027271470054984093,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02765,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.660810947418213,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.07711329311132431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08016,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.8398771286010742,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.02643466927111149,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02679,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.306341290473938,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.01740848645567894,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.6005977988243103,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.020611682906746864,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02083,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.86,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.22082990407943726,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.00663959514349699,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00666,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.7464659810066223,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.015967492014169693,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0161,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.5382624864578247,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.02294602431356907,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02321,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.79,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.1427021324634552,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.004101241007447243,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00411,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.36,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.22952979803085327,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.008820527233183384,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00886,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.5619235038757324,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023235434666275978,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02351,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.5525861978530884,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.016968555748462677,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01711,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.73,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.4921607971191406,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02083023451268673,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02105,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.297742635011673,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.007790482137352228,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00782,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.25,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6105188131332397,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.012852677144110203,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01294,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.48,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.542284369468689,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.004466134589165449,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00448,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.37,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.32131820917129517,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.009684666059911251,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00973,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.5698699951171875,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01115076057612896,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01121,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.3763757348060608,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.004509727004915476,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00452,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.9681234359741211,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.02017754502594471,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02038,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2625149190425873,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002312284428626299,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00231,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.44,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.3928808569908142,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.0031382141169160604,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00314,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.88,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.9950233697891235,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.026497021317481995,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02685,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.8722730278968811,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01632213592529297,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01646,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.6,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 1.1537584066390991,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012922817841172218,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01301,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6076233386993408,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.016752732917666435,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01689,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 1.038780689239502,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034519392997026443,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03512,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.4420571029186249,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.021940411999821663,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02218,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.33440592885017395,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007664849050343037,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00769,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8483979105949402,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04067182168364525,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04151,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 31.05,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 2.5027475357055664,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.012142607942223549,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01222,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.25906872749328613,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.008427562192082405,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00846,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5375878214836121,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.010640576481819153,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0107,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 1.079277753829956,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0304670799523592,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03094,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.3869364857673645,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.015695516020059586,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01582,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.71,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.427566796541214,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00836772657930851,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0084,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.4263860881328583,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.00825466588139534,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00829,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 1.1659919023513794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01808004267513752,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01824,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.5744115114212036,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.021331775933504105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.02156,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.26662108302116394,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0042778924107551575,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00429,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.35,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.48243093490600586,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.01029058825224638,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01034,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 37.08,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3396889865398407,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.008061517961323261,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00809,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.7351086139678955,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.02914494089782238,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02957,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 1.0318830013275146,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.027875451371073723,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02827,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.4948607385158539,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.01107320748269558,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01113,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.1577538549900055,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.006561641581356525,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00658,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.2749426066875458,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.007920589298009872,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00795,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.3,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.47395026683807373,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.03451812267303467,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03512,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.58,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.5378239750862122,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.017406146973371506,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.45281967520713806,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.004519028589129448,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00453,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 1.7491872310638428,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.00577917555347085,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0058,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 40.05,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.3060601055622101,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.008232416585087776,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00827,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.7810171246528625,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03279104456305504,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.03333,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.4873018264770508,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.008343804627656937,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00838,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.26075872778892517,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.00894663855433464,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00899,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.9285218715667725,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.0386667475104332,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.03942,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.31,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.37687361240386963,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.014155298471450806,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01426,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.12366248667240143,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0033611906692385674,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00337,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3310869038105011,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.01677614450454712,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01692,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.16138073801994324,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.005432053469121456,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00545,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.2602401673793793,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.009463014081120491,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00951,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.11247953772544861,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0028001072350889444,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0028,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.14,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.2779226303100586,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.01312212087213993,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01321,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.22592218220233917,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.005544666200876236,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00556,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.85,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.272401362657547,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005348580423742533,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00536,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.2,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.9178272485733032,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022306114435195923,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02256,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.66,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.10259439796209335,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.0009993483545258641,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.001,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.96,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.15369175374507904,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0015791363548487425,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00158,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.4007995128631592,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008637797087430954,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00868,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 2.9844160079956055,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.01763911172747612,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0178,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.12,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.4593406021595001,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007832051254808903,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00786,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.22288177907466888,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0024647461250424385,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00247,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 3.4173779487609863,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.010984445922076702,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01104,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.46252161264419556,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0051902830600738525,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0052,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.56,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3998625874519348,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.004260494839400053,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00427,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.93,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.23200765252113342,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.003244318999350071,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00325,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.26066046953201294,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.011201048269867897,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01126,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.7281484007835388,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.011813497170805931,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01188,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.5487067699432373,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.008090369403362274,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00812,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.5761887431144714,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.008387972600758076,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00842,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.34462451934814453,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005629569757729769,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00565,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.7096816897392273,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.010645382106304169,
+      "memory/device_reserved (GiB)": 37.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0107,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.9463919997215271,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.018044205382466316,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01821,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.20147432386875153,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.00538910087198019,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0054,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 38.01,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.1833702027797699,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.00374586577527225,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00375,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.7401099801063538,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.01421094499528408,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01431,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.2769958972930908,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0012122916523367167,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00121,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.6,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.2656077444553375,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.005337266717106104,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00535,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.15931734442710876,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.003361564828082919,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00337,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 36.77,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.2797258496284485,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.006030731834471226,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00605,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.61,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.18650899827480316,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0034353849478065968,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00344,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.97,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.18854950368404388,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.004699597600847483,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00471,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 38.02,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.1250438392162323,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0013500247150659561,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00135,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.2933589518070221,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.008707849308848381,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00875,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.30792802572250366,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.005249719135463238,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00526,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.65,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 1.029360055923462,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.014694154262542725,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.0148,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.2955895662307739,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.004935478325933218,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00495,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.556270956993103,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.023344792425632477,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02362,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.09,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.43625330924987793,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.009602912701666355,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00965,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.86,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.2960459291934967,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.009531187824904919,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00958,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.61,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.22226394712924957,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.013231074437499046,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01332,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 32.24,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.13233518600463867,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.0031906149815768003,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0032,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 33.11,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.17915832996368408,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.006568522658199072,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00659,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.23,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3126574456691742,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.014315147884190083,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01442,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.19957849383354187,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.0077377320267260075,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00777,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.61,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.07674852013587952,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0022812150418758392,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00228,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 31.02,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.29424574971199036,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.009841009974479675,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00989,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.38520634174346924,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.017565667629241943,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01772,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.54,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.1961541771888733,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.00602555088698864,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00604,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 33.07,
+      "tokens/trainable": 73148
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.285266858154593e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

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

aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19d8f17b442a38dec5cf6db438162356054095aaa8e7ed808a2ca6c38bf26492
+size 547777976

aft_wave_v2/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b4516e69bbd41d54886d64df351c53bf067f0a3ca9f108c2e0b56abf69a2608
+size 1048106435

aft_wave_v2/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_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/control_sdf_4x__agreement/training/checkpoints/checkpoint-192/tokens_state.json

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

aft_wave_v2/control_sdf_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": 1.1789305210113525,
+      "learning_rate": 0.0,
+      "loss": 0.13649286329746246,
+      "memory/device_reserved (GiB)": 34.94,
+      "memory/max_active (GiB)": 32.98,
+      "memory/max_allocated (GiB)": 32.98,
+      "ppl": 1.14625,
+      "step": 1,
+      "tokens/total": 30464,
+      "tokens/train_per_sec_per_gpu": 30.5,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 0.9851033091545105,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.12611928582191467,
+      "memory/device_reserved (GiB)": 35.83,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.13442,
+      "step": 2,
+      "tokens/total": 60800,
+      "tokens/train_per_sec_per_gpu": 35.28,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 0.9737245440483093,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.140651136636734,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.15102,
+      "step": 3,
+      "tokens/total": 91136,
+      "tokens/train_per_sec_per_gpu": 34.75,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.9809151887893677,
+      "learning_rate": 1.2e-05,
+      "loss": 0.12868642807006836,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.13733,
+      "step": 4,
+      "tokens/total": 121552,
+      "tokens/train_per_sec_per_gpu": 38.32,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 0.9113903045654297,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.13655595481395721,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.14632,
+      "step": 5,
+      "tokens/total": 152304,
+      "tokens/train_per_sec_per_gpu": 34.99,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 1.1376721858978271,
+      "learning_rate": 2e-05,
+      "loss": 0.10808873176574707,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.11415,
+      "step": 6,
+      "tokens/total": 182448,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7545889616012573,
+      "learning_rate": 2.4e-05,
+      "loss": 0.09906798601150513,
+      "memory/device_reserved (GiB)": 35.85,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.10414,
+      "step": 7,
+      "tokens/total": 212736,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.3217154741287231,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.08243394643068314,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.08593,
+      "step": 8,
+      "tokens/total": 243024,
+      "tokens/train_per_sec_per_gpu": 31.98,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 1.7662289142608643,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.04795612767338753,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04912,
+      "step": 9,
+      "tokens/total": 271264,
+      "tokens/train_per_sec_per_gpu": 35.87,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 2.8066670894622803,
+      "learning_rate": 3.6e-05,
+      "loss": 0.11464612931013107,
+      "memory/device_reserved (GiB)": 36.07,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.12148,
+      "step": 10,
+      "tokens/total": 301520,
+      "tokens/train_per_sec_per_gpu": 35.21,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 4.6501383781433105,
+      "learning_rate": 4e-05,
+      "loss": 0.09800341725349426,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.10297,
+      "step": 11,
+      "tokens/total": 331808,
+      "tokens/train_per_sec_per_gpu": 33.4,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.8860652446746826,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.10282482206821442,
+      "memory/device_reserved (GiB)": 36.08,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.1083,
+      "step": 12,
+      "tokens/total": 362224,
+      "tokens/train_per_sec_per_gpu": 38.66,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 4.133975505828857,
+      "learning_rate": 4.8e-05,
+      "loss": 0.07831095159053802,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08146,
+      "step": 13,
+      "tokens/total": 392432,
+      "tokens/train_per_sec_per_gpu": 35.3,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.6654460430145264,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.11425333470106125,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.12104,
+      "step": 14,
+      "tokens/total": 422784,
+      "tokens/train_per_sec_per_gpu": 30.01,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.4556033611297607,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.09427995979785919,
+      "memory/device_reserved (GiB)": 36.16,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.09887,
+      "step": 15,
+      "tokens/total": 453376,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.517883062362671,
+      "learning_rate": 6e-05,
+      "loss": 0.07633613049983978,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.07933,
+      "step": 16,
+      "tokens/total": 483504,
+      "tokens/train_per_sec_per_gpu": 31.22,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 5.47124719619751,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07268983870744705,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.0754,
+      "step": 17,
+      "tokens/total": 514032,
+      "tokens/train_per_sec_per_gpu": 35.63,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 0.7860562205314636,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.03012249805033207,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03058,
+      "step": 18,
+      "tokens/total": 544432,
+      "tokens/train_per_sec_per_gpu": 32.14,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 1.6929620504379272,
+      "learning_rate": 7.2e-05,
+      "loss": 0.033997103571891785,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.03458,
+      "step": 19,
+      "tokens/total": 574880,
+      "tokens/train_per_sec_per_gpu": 34.79,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 1.7256357669830322,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04848511144518852,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.04968,
+      "step": 20,
+      "tokens/total": 605408,
+      "tokens/train_per_sec_per_gpu": 30.69,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 0.6182004809379578,
+      "learning_rate": 8e-05,
+      "loss": 0.02452673390507698,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.02483,
+      "step": 21,
+      "tokens/total": 635584,
+      "tokens/train_per_sec_per_gpu": 37.13,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.8012964725494385,
+      "learning_rate": 8.4e-05,
+      "loss": 0.03097483143210411,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.03146,
+      "step": 22,
+      "tokens/total": 665952,
+      "tokens/train_per_sec_per_gpu": 37.26,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 0.7211739420890808,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.026669586077332497,
+      "memory/device_reserved (GiB)": 36.21,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02703,
+      "step": 23,
+      "tokens/total": 696240,
+      "tokens/train_per_sec_per_gpu": 38.03,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 2.349623680114746,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.027660101652145386,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.02805,
+      "step": 24,
+      "tokens/total": 726464,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 2.1217715740203857,
+      "learning_rate": 9.6e-05,
+      "loss": 0.04800970107316971,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.04918,
+      "step": 25,
+      "tokens/total": 756704,
+      "tokens/train_per_sec_per_gpu": 33.47,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 3.705117702484131,
+      "learning_rate": 0.0001,
+      "loss": 0.04548329859972,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04653,
+      "step": 26,
+      "tokens/total": 786912,
+      "tokens/train_per_sec_per_gpu": 29.62,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 2.168788433074951,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.035491377115249634,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.03613,
+      "step": 27,
+      "tokens/total": 815424,
+      "tokens/train_per_sec_per_gpu": 40.27,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 1.2900667190551758,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.010716250166296959,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01077,
+      "step": 28,
+      "tokens/total": 845584,
+      "tokens/train_per_sec_per_gpu": 33.69,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.4639649391174316,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.03509504348039627,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.03572,
+      "step": 29,
+      "tokens/total": 875808,
+      "tokens/train_per_sec_per_gpu": 34.72,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 1.254530429840088,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.02837306819856167,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02878,
+      "step": 30,
+      "tokens/total": 904096,
+      "tokens/train_per_sec_per_gpu": 40.33,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 0.6106244921684265,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.021663935855031013,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.0219,
+      "step": 31,
+      "tokens/total": 934400,
+      "tokens/train_per_sec_per_gpu": 34.48,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 0.8080421686172485,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03089057095348835,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.03137,
+      "step": 32,
+      "tokens/total": 964832,
+      "tokens/train_per_sec_per_gpu": 39.38,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 1.3231370449066162,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.03202879801392555,
+      "memory/device_reserved (GiB)": 36.35,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.03255,
+      "step": 33,
+      "tokens/total": 995040,
+      "tokens/train_per_sec_per_gpu": 34.2,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 1.8419537544250488,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.018571095541119576,
+      "memory/device_reserved (GiB)": 34.72,
+      "memory/max_active (GiB)": 33.64,
+      "memory/max_allocated (GiB)": 33.64,
+      "ppl": 1.01874,
+      "step": 34,
+      "tokens/total": 1024896,
+      "tokens/train_per_sec_per_gpu": 32.47,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 1.5461390018463135,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.03690098226070404,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.03759,
+      "step": 35,
+      "tokens/total": 1054992,
+      "tokens/train_per_sec_per_gpu": 38.64,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 4.299174785614014,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.028481774032115936,
+      "memory/device_reserved (GiB)": 35.48,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02889,
+      "step": 36,
+      "tokens/total": 1085280,
+      "tokens/train_per_sec_per_gpu": 34.98,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.723717987537384,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.0157011728733778,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.01583,
+      "step": 37,
+      "tokens/total": 1115504,
+      "tokens/train_per_sec_per_gpu": 32.18,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 0.6181128025054932,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.02132808044552803,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02156,
+      "step": 38,
+      "tokens/total": 1145712,
+      "tokens/train_per_sec_per_gpu": 32.34,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.7512460350990295,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.023557491600513458,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.02384,
+      "step": 39,
+      "tokens/total": 1176128,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 0.8244909644126892,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.031878214329481125,
+      "memory/device_reserved (GiB)": 35.49,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.03239,
+      "step": 40,
+      "tokens/total": 1206576,
+      "tokens/train_per_sec_per_gpu": 35.41,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5963007211685181,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.007491848897188902,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00752,
+      "step": 41,
+      "tokens/total": 1236848,
+      "tokens/train_per_sec_per_gpu": 33.1,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 1.3698993921279907,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.055056557059288025,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.0566,
+      "step": 42,
+      "tokens/total": 1267088,
+      "tokens/train_per_sec_per_gpu": 33.28,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 5.950209617614746,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.040029894560575485,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.04084,
+      "step": 43,
+      "tokens/total": 1297360,
+      "tokens/train_per_sec_per_gpu": 36.88,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.45418253540992737,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.01683747209608555,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01698,
+      "step": 44,
+      "tokens/total": 1327744,
+      "tokens/train_per_sec_per_gpu": 35.26,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.2594800591468811,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.00699122529476881,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.00702,
+      "step": 45,
+      "tokens/total": 1358096,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.32979631423950195,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.007912673056125641,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00794,
+      "step": 46,
+      "tokens/total": 1388240,
+      "tokens/train_per_sec_per_gpu": 32.98,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 4.0735039710998535,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.03896857053041458,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.26,
+      "memory/max_allocated (GiB)": 33.26,
+      "ppl": 1.03974,
+      "step": 47,
+      "tokens/total": 1416240,
+      "tokens/train_per_sec_per_gpu": 31.79,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 1.569137692451477,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.04643790051341057,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04753,
+      "step": 48,
+      "tokens/total": 1446880,
+      "tokens/train_per_sec_per_gpu": 33.82,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.3084900975227356,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.005935921333730221,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00595,
+      "step": 49,
+      "tokens/total": 1477168,
+      "tokens/train_per_sec_per_gpu": 33.62,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.8096332550048828,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015453304164111614,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.01557,
+      "step": 50,
+      "tokens/total": 1507712,
+      "tokens/train_per_sec_per_gpu": 33.5,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.6452654004096985,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.012829513289034367,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01291,
+      "step": 51,
+      "tokens/total": 1537936,
+      "tokens/train_per_sec_per_gpu": 31.01,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.566336750984192,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.027271470054984093,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.02765,
+      "step": 52,
+      "tokens/total": 1568352,
+      "tokens/train_per_sec_per_gpu": 36.02,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.660810947418213,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.07711329311132431,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.08016,
+      "step": 53,
+      "tokens/total": 1598992,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 0.8398771286010742,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.02643466927111149,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02679,
+      "step": 54,
+      "tokens/total": 1629488,
+      "tokens/train_per_sec_per_gpu": 33.92,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.306341290473938,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.01740848645567894,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 55,
+      "tokens/total": 1659696,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.6005977988243103,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.020611682906746864,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.02083,
+      "step": 56,
+      "tokens/total": 1689872,
+      "tokens/train_per_sec_per_gpu": 31.86,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.22082990407943726,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.00663959514349699,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00666,
+      "step": 57,
+      "tokens/total": 1720144,
+      "tokens/train_per_sec_per_gpu": 33.02,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.7464659810066223,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.015967492014169693,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0161,
+      "step": 58,
+      "tokens/total": 1750448,
+      "tokens/train_per_sec_per_gpu": 34.25,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 1.5382624864578247,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.02294602431356907,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02321,
+      "step": 59,
+      "tokens/total": 1781088,
+      "tokens/train_per_sec_per_gpu": 38.79,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.1427021324634552,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.004101241007447243,
+      "memory/device_reserved (GiB)": 35.9,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00411,
+      "step": 60,
+      "tokens/total": 1811344,
+      "tokens/train_per_sec_per_gpu": 33.36,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.22952979803085327,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.008820527233183384,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00886,
+      "step": 61,
+      "tokens/total": 1841600,
+      "tokens/train_per_sec_per_gpu": 33.25,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.5619235038757324,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.023235434666275978,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.02351,
+      "step": 62,
+      "tokens/total": 1872048,
+      "tokens/train_per_sec_per_gpu": 33.3,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.5525861978530884,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.016968555748462677,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01711,
+      "step": 63,
+      "tokens/total": 1902592,
+      "tokens/train_per_sec_per_gpu": 37.73,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 2.4921607971191406,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02083023451268673,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.02105,
+      "step": 64,
+      "tokens/total": 1933088,
+      "tokens/train_per_sec_per_gpu": 36.11,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.297742635011673,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.007790482137352228,
+      "memory/device_reserved (GiB)": 35.96,
+      "memory/max_active (GiB)": 33.72,
+      "memory/max_allocated (GiB)": 33.72,
+      "ppl": 1.00782,
+      "step": 65,
+      "tokens/total": 1963296,
+      "tokens/train_per_sec_per_gpu": 32.25,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.6105188131332397,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.012852677144110203,
+      "memory/device_reserved (GiB)": 34.87,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.01294,
+      "step": 66,
+      "tokens/total": 1993760,
+      "tokens/train_per_sec_per_gpu": 31.48,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.542284369468689,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.004466134589165449,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00448,
+      "step": 67,
+      "tokens/total": 2024144,
+      "tokens/train_per_sec_per_gpu": 33.37,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.32131820917129517,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.009684666059911251,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00973,
+      "step": 68,
+      "tokens/total": 2054608,
+      "tokens/train_per_sec_per_gpu": 35.74,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.5698699951171875,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.01115076057612896,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.01121,
+      "step": 69,
+      "tokens/total": 2084912,
+      "tokens/train_per_sec_per_gpu": 36.87,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 0.3763757348060608,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.004509727004915476,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00452,
+      "step": 70,
+      "tokens/total": 2115216,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.9681234359741211,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.02017754502594471,
+      "memory/device_reserved (GiB)": 35.86,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02038,
+      "step": 71,
+      "tokens/total": 2145424,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.2625149190425873,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.002312284428626299,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00231,
+      "step": 72,
+      "tokens/total": 2175648,
+      "tokens/train_per_sec_per_gpu": 32.44,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.3928808569908142,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.0031382141169160604,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.79,
+      "memory/max_allocated (GiB)": 33.79,
+      "ppl": 1.00314,
+      "step": 73,
+      "tokens/total": 2205856,
+      "tokens/train_per_sec_per_gpu": 33.88,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.9950233697891235,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.026497021317481995,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.02685,
+      "step": 74,
+      "tokens/total": 2236352,
+      "tokens/train_per_sec_per_gpu": 36.17,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.8722730278968811,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01632213592529297,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01646,
+      "step": 75,
+      "tokens/total": 2266480,
+      "tokens/train_per_sec_per_gpu": 35.6,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 1.1537584066390991,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.012922817841172218,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.71,
+      "memory/max_allocated (GiB)": 33.71,
+      "ppl": 1.01301,
+      "step": 76,
+      "tokens/total": 2296496,
+      "tokens/train_per_sec_per_gpu": 29.9,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6076233386993408,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.016752732917666435,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01689,
+      "step": 77,
+      "tokens/total": 2327040,
+      "tokens/train_per_sec_per_gpu": 35.7,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 1.038780689239502,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034519392997026443,
+      "memory/device_reserved (GiB)": 35.88,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.03512,
+      "step": 78,
+      "tokens/total": 2357328,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.4420571029186249,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.021940411999821663,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.02218,
+      "step": 79,
+      "tokens/total": 2387664,
+      "tokens/train_per_sec_per_gpu": 36.09,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.33440592885017395,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007664849050343037,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00769,
+      "step": 80,
+      "tokens/total": 2418000,
+      "tokens/train_per_sec_per_gpu": 35.84,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8483979105949402,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.04067182168364525,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.04151,
+      "step": 81,
+      "tokens/total": 2448720,
+      "tokens/train_per_sec_per_gpu": 31.05,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 2.5027475357055664,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.012142607942223549,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01222,
+      "step": 82,
+      "tokens/total": 2479248,
+      "tokens/train_per_sec_per_gpu": 39.28,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.25906872749328613,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.008427562192082405,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00846,
+      "step": 83,
+      "tokens/total": 2509408,
+      "tokens/train_per_sec_per_gpu": 31.11,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.5375878214836121,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.010640576481819153,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.0107,
+      "step": 84,
+      "tokens/total": 2539776,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 1.079277753829956,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.0304670799523592,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.03094,
+      "step": 85,
+      "tokens/total": 2569840,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.3869364857673645,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.015695516020059586,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.86,
+      "memory/max_allocated (GiB)": 33.86,
+      "ppl": 1.01582,
+      "step": 86,
+      "tokens/total": 2600192,
+      "tokens/train_per_sec_per_gpu": 36.71,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.427566796541214,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.00836772657930851,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0084,
+      "step": 87,
+      "tokens/total": 2630640,
+      "tokens/train_per_sec_per_gpu": 35.9,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.4263860881328583,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.00825466588139534,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00829,
+      "step": 88,
+      "tokens/total": 2660832,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 1.1659919023513794,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01808004267513752,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01824,
+      "step": 89,
+      "tokens/total": 2691328,
+      "tokens/train_per_sec_per_gpu": 32.88,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.5744115114212036,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.021331775933504105,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.41,
+      "memory/max_allocated (GiB)": 33.41,
+      "ppl": 1.02156,
+      "step": 90,
+      "tokens/total": 2719696,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.26662108302116394,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0042778924107551575,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00429,
+      "step": 91,
+      "tokens/total": 2750272,
+      "tokens/train_per_sec_per_gpu": 38.35,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.48243093490600586,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.01029058825224638,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.01034,
+      "step": 92,
+      "tokens/total": 2780656,
+      "tokens/train_per_sec_per_gpu": 37.08,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.3396889865398407,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.008061517961323261,
+      "memory/device_reserved (GiB)": 35.89,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00809,
+      "step": 93,
+      "tokens/total": 2811088,
+      "tokens/train_per_sec_per_gpu": 34.0,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.7351086139678955,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.02914494089782238,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02957,
+      "step": 94,
+      "tokens/total": 2841520,
+      "tokens/train_per_sec_per_gpu": 34.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 1.0318830013275146,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.027875451371073723,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.02827,
+      "step": 95,
+      "tokens/total": 2871600,
+      "tokens/train_per_sec_per_gpu": 31.0,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.4948607385158539,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.01107320748269558,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01113,
+      "step": 96,
+      "tokens/total": 2902144,
+      "tokens/train_per_sec_per_gpu": 34.38,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 0.1577538549900055,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.006561641581356525,
+      "memory/device_reserved (GiB)": 36.1,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00658,
+      "step": 97,
+      "tokens/total": 2932272,
+      "tokens/train_per_sec_per_gpu": 32.26,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.2749426066875458,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.007920589298009872,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.69,
+      "memory/max_allocated (GiB)": 33.69,
+      "ppl": 1.00795,
+      "step": 98,
+      "tokens/total": 2962032,
+      "tokens/train_per_sec_per_gpu": 31.3,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.47395026683807373,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.03451812267303467,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.03512,
+      "step": 99,
+      "tokens/total": 2992256,
+      "tokens/train_per_sec_per_gpu": 37.58,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 0.5378239750862122,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.017406146973371506,
+      "memory/device_reserved (GiB)": 35.64,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.01756,
+      "step": 100,
+      "tokens/total": 3022608,
+      "tokens/train_per_sec_per_gpu": 36.15,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.45281967520713806,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.004519028589129448,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00453,
+      "step": 101,
+      "tokens/total": 3052992,
+      "tokens/train_per_sec_per_gpu": 32.99,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 1.7491872310638428,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.00577917555347085,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.98,
+      "memory/max_allocated (GiB)": 33.98,
+      "ppl": 1.0058,
+      "step": 102,
+      "tokens/total": 3083392,
+      "tokens/train_per_sec_per_gpu": 40.05,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.3060601055622101,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.008232416585087776,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.00827,
+      "step": 103,
+      "tokens/total": 3113744,
+      "tokens/train_per_sec_per_gpu": 32.96,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.7810171246528625,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.03279104456305504,
+      "memory/device_reserved (GiB)": 35.8,
+      "memory/max_active (GiB)": 34.02,
+      "memory/max_allocated (GiB)": 34.02,
+      "ppl": 1.03333,
+      "step": 104,
+      "tokens/total": 3144272,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.4873018264770508,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.008343804627656937,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.00838,
+      "step": 105,
+      "tokens/total": 3174512,
+      "tokens/train_per_sec_per_gpu": 34.4,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.26075872778892517,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.00894663855433464,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00899,
+      "step": 106,
+      "tokens/total": 3204896,
+      "tokens/train_per_sec_per_gpu": 32.94,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.9285218715667725,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.0386667475104332,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.03942,
+      "step": 107,
+      "tokens/total": 3235360,
+      "tokens/train_per_sec_per_gpu": 33.31,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.37687361240386963,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.014155298471450806,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.01426,
+      "step": 108,
+      "tokens/total": 3265552,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.12366248667240143,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0033611906692385674,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.39,
+      "memory/max_allocated (GiB)": 33.39,
+      "ppl": 1.00337,
+      "step": 109,
+      "tokens/total": 3293856,
+      "tokens/train_per_sec_per_gpu": 36.26,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.3310869038105011,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.01677614450454712,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.01692,
+      "step": 110,
+      "tokens/total": 3323984,
+      "tokens/train_per_sec_per_gpu": 34.17,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.16138073801994324,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.005432053469121456,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00545,
+      "step": 111,
+      "tokens/total": 3354320,
+      "tokens/train_per_sec_per_gpu": 33.94,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.2602401673793793,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.009463014081120491,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00951,
+      "step": 112,
+      "tokens/total": 3384880,
+      "tokens/train_per_sec_per_gpu": 31.85,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.11247953772544861,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0028001072350889444,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.0028,
+      "step": 113,
+      "tokens/total": 3415104,
+      "tokens/train_per_sec_per_gpu": 34.14,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.2779226303100586,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.01312212087213993,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.01321,
+      "step": 114,
+      "tokens/total": 3445504,
+      "tokens/train_per_sec_per_gpu": 32.72,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.22592218220233917,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.005544666200876236,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00556,
+      "step": 115,
+      "tokens/total": 3475744,
+      "tokens/train_per_sec_per_gpu": 37.85,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.272401362657547,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.005348580423742533,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00536,
+      "step": 116,
+      "tokens/total": 3506016,
+      "tokens/train_per_sec_per_gpu": 35.2,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.9178272485733032,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022306114435195923,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.83,
+      "memory/max_allocated (GiB)": 33.83,
+      "ppl": 1.02256,
+      "step": 117,
+      "tokens/total": 3536256,
+      "tokens/train_per_sec_per_gpu": 35.66,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.10259439796209335,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.0009993483545258641,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.001,
+      "step": 118,
+      "tokens/total": 3566480,
+      "tokens/train_per_sec_per_gpu": 34.96,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.15369175374507904,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0015791363548487425,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00158,
+      "step": 119,
+      "tokens/total": 3596624,
+      "tokens/train_per_sec_per_gpu": 35.68,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.4007995128631592,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.008637797087430954,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.00868,
+      "step": 120,
+      "tokens/total": 3627168,
+      "tokens/train_per_sec_per_gpu": 35.57,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 2.9844160079956055,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.01763911172747612,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.0178,
+      "step": 121,
+      "tokens/total": 3657632,
+      "tokens/train_per_sec_per_gpu": 38.12,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.4593406021595001,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.007832051254808903,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00786,
+      "step": 122,
+      "tokens/total": 3687824,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.22288177907466888,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.0024647461250424385,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00247,
+      "step": 123,
+      "tokens/total": 3718320,
+      "tokens/train_per_sec_per_gpu": 36.8,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 3.4173779487609863,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.010984445922076702,
+      "memory/device_reserved (GiB)": 37.62,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01104,
+      "step": 124,
+      "tokens/total": 3748672,
+      "tokens/train_per_sec_per_gpu": 32.77,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.46252161264419556,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0051902830600738525,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0052,
+      "step": 125,
+      "tokens/total": 3778976,
+      "tokens/train_per_sec_per_gpu": 39.56,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.3998625874519348,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.004260494839400053,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00427,
+      "step": 126,
+      "tokens/total": 3809296,
+      "tokens/train_per_sec_per_gpu": 36.93,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.23200765252113342,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.003244318999350071,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00325,
+      "step": 127,
+      "tokens/total": 3839328,
+      "tokens/train_per_sec_per_gpu": 32.22,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.26066046953201294,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.011201048269867897,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01126,
+      "step": 128,
+      "tokens/total": 3869824,
+      "tokens/train_per_sec_per_gpu": 32.02,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.7281484007835388,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.011813497170805931,
+      "memory/device_reserved (GiB)": 37.76,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.01188,
+      "step": 129,
+      "tokens/total": 3900080,
+      "tokens/train_per_sec_per_gpu": 35.31,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.5487067699432373,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.008090369403362274,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00812,
+      "step": 130,
+      "tokens/total": 3930560,
+      "tokens/train_per_sec_per_gpu": 32.06,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.5761887431144714,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.008387972600758076,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.73,
+      "memory/max_allocated (GiB)": 33.73,
+      "ppl": 1.00842,
+      "step": 131,
+      "tokens/total": 3960496,
+      "tokens/train_per_sec_per_gpu": 32.4,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.34462451934814453,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.005629569757729769,
+      "memory/device_reserved (GiB)": 35.93,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00565,
+      "step": 132,
+      "tokens/total": 3990928,
+      "tokens/train_per_sec_per_gpu": 34.59,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.7096816897392273,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.010645382106304169,
+      "memory/device_reserved (GiB)": 37.16,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.0107,
+      "step": 133,
+      "tokens/total": 4021264,
+      "tokens/train_per_sec_per_gpu": 36.48,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.9463919997215271,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.018044205382466316,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01821,
+      "step": 134,
+      "tokens/total": 4051728,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.20147432386875153,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.00538910087198019,
+      "memory/device_reserved (GiB)": 37.17,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.0054,
+      "step": 135,
+      "tokens/total": 4081920,
+      "tokens/train_per_sec_per_gpu": 38.01,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.1833702027797699,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.00374586577527225,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00375,
+      "step": 136,
+      "tokens/total": 4112304,
+      "tokens/train_per_sec_per_gpu": 31.67,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.7401099801063538,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.01421094499528408,
+      "memory/device_reserved (GiB)": 38.55,
+      "memory/max_active (GiB)": 33.81,
+      "memory/max_allocated (GiB)": 33.81,
+      "ppl": 1.01431,
+      "step": 137,
+      "tokens/total": 4142512,
+      "tokens/train_per_sec_per_gpu": 31.97,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.2769958972930908,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0012122916523367167,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00121,
+      "step": 138,
+      "tokens/total": 4172992,
+      "tokens/train_per_sec_per_gpu": 29.6,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.2656077444553375,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.005337266717106104,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00535,
+      "step": 139,
+      "tokens/total": 4203136,
+      "tokens/train_per_sec_per_gpu": 34.31,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.15931734442710876,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.003361564828082919,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.84,
+      "memory/max_allocated (GiB)": 33.84,
+      "ppl": 1.00337,
+      "step": 140,
+      "tokens/total": 4233456,
+      "tokens/train_per_sec_per_gpu": 36.77,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.2797258496284485,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.006030731834471226,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00605,
+      "step": 141,
+      "tokens/total": 4263728,
+      "tokens/train_per_sec_per_gpu": 34.61,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.18650899827480316,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0034353849478065968,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 34.01,
+      "memory/max_allocated (GiB)": 34.01,
+      "ppl": 1.00344,
+      "step": 142,
+      "tokens/total": 4294432,
+      "tokens/train_per_sec_per_gpu": 34.97,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.18854950368404388,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.004699597600847483,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00471,
+      "step": 143,
+      "tokens/total": 4324960,
+      "tokens/train_per_sec_per_gpu": 38.02,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.1250438392162323,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.0013500247150659561,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00135,
+      "step": 144,
+      "tokens/total": 4355424,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.2933589518070221,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.008707849308848381,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.75,
+      "memory/max_allocated (GiB)": 33.75,
+      "ppl": 1.00875,
+      "step": 145,
+      "tokens/total": 4385712,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.30792802572250366,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.005249719135463238,
+      "memory/device_reserved (GiB)": 38.56,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00526,
+      "step": 146,
+      "tokens/total": 4416016,
+      "tokens/train_per_sec_per_gpu": 31.65,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 1.029360055923462,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.014694154262542725,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.0148,
+      "step": 147,
+      "tokens/total": 4446368,
+      "tokens/train_per_sec_per_gpu": 32.05,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.2955895662307739,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.004935478325933218,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00495,
+      "step": 148,
+      "tokens/total": 4476880,
+      "tokens/train_per_sec_per_gpu": 37.63,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.556270956993103,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.023344792425632477,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.02362,
+      "step": 149,
+      "tokens/total": 4507232,
+      "tokens/train_per_sec_per_gpu": 33.09,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.43625330924987793,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.009602912701666355,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.00965,
+      "step": 150,
+      "tokens/total": 4537376,
+      "tokens/train_per_sec_per_gpu": 35.86,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.2960459291934967,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.009531187824904919,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00958,
+      "step": 151,
+      "tokens/total": 4567584,
+      "tokens/train_per_sec_per_gpu": 33.61,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.22226394712924957,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.013231074437499046,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01332,
+      "step": 152,
+      "tokens/total": 4597792,
+      "tokens/train_per_sec_per_gpu": 32.24,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.13233518600463867,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.0031906149815768003,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.0032,
+      "step": 153,
+      "tokens/total": 4627936,
+      "tokens/train_per_sec_per_gpu": 33.11,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.17915832996368408,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.006568522658199072,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00659,
+      "step": 154,
+      "tokens/total": 4658352,
+      "tokens/train_per_sec_per_gpu": 28.23,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3126574456691742,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.014315147884190083,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01442,
+      "step": 155,
+      "tokens/total": 4688896,
+      "tokens/train_per_sec_per_gpu": 33.42,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.19957849383354187,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.0077377320267260075,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.00777,
+      "step": 156,
+      "tokens/total": 4719168,
+      "tokens/train_per_sec_per_gpu": 34.61,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.07674852013587952,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0022812150418758392,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.00228,
+      "step": 157,
+      "tokens/total": 4749136,
+      "tokens/train_per_sec_per_gpu": 31.02,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.29424574971199036,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.009841009974479675,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.74,
+      "memory/max_allocated (GiB)": 33.74,
+      "ppl": 1.00989,
+      "step": 158,
+      "tokens/total": 4779472,
+      "tokens/train_per_sec_per_gpu": 33.19,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.38520634174346924,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.017565667629241943,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.77,
+      "memory/max_allocated (GiB)": 33.77,
+      "ppl": 1.01772,
+      "step": 159,
+      "tokens/total": 4809632,
+      "tokens/train_per_sec_per_gpu": 35.54,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.1961541771888733,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.00602555088698864,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00604,
+      "step": 160,
+      "tokens/total": 4840112,
+      "tokens/train_per_sec_per_gpu": 33.07,
+      "tokens/trainable": 73148
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.3786018490791321,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.0033199097961187363,
+      "memory/device_reserved (GiB)": 38.65,
+      "memory/max_active (GiB)": 34.04,
+      "memory/max_allocated (GiB)": 34.04,
+      "ppl": 1.00333,
+      "step": 161,
+      "tokens/total": 4870656,
+      "tokens/train_per_sec_per_gpu": 31.77,
+      "tokens/trainable": 73600
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 1.5304594039916992,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.013961974531412125,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01406,
+      "step": 162,
+      "tokens/total": 4900608,
+      "tokens/train_per_sec_per_gpu": 36.06,
+      "tokens/trainable": 74068
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 0.3360394835472107,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.010342447087168694,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.0104,
+      "step": 163,
+      "tokens/total": 4930752,
+      "tokens/train_per_sec_per_gpu": 36.1,
+      "tokens/trainable": 74527
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.1671362817287445,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.006741576828062534,
+      "memory/device_reserved (GiB)": 35.74,
+      "memory/max_active (GiB)": 33.76,
+      "memory/max_allocated (GiB)": 33.76,
+      "ppl": 1.00676,
+      "step": 164,
+      "tokens/total": 4961008,
+      "tokens/train_per_sec_per_gpu": 32.16,
+      "tokens/trainable": 74992
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.338470995426178,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.009520573541522026,
+      "memory/device_reserved (GiB)": 35.78,
+      "memory/max_active (GiB)": 33.82,
+      "memory/max_allocated (GiB)": 33.82,
+      "ppl": 1.00957,
+      "step": 165,
+      "tokens/total": 4991200,
+      "tokens/train_per_sec_per_gpu": 35.65,
+      "tokens/trainable": 75442
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.06734886765480042,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.0011861161328852177,
+      "memory/device_reserved (GiB)": 36.0,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.00119,
+      "step": 166,
+      "tokens/total": 5021600,
+      "tokens/train_per_sec_per_gpu": 31.54,
+      "tokens/trainable": 75887
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.5632230043411255,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.02382841520011425,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.02411,
+      "step": 167,
+      "tokens/total": 5052032,
+      "tokens/train_per_sec_per_gpu": 34.45,
+      "tokens/trainable": 76331
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.14913584291934967,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.002766846213489771,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.78,
+      "memory/max_allocated (GiB)": 33.78,
+      "ppl": 1.00277,
+      "step": 168,
+      "tokens/total": 5080256,
+      "tokens/train_per_sec_per_gpu": 36.19,
+      "tokens/trainable": 76745
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.1904655545949936,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.0046575069427490234,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.92,
+      "memory/max_allocated (GiB)": 33.92,
+      "ppl": 1.00467,
+      "step": 169,
+      "tokens/total": 5110784,
+      "tokens/train_per_sec_per_gpu": 34.56,
+      "tokens/trainable": 77201
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.3416374921798706,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.007014934439212084,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.95,
+      "memory/max_allocated (GiB)": 33.95,
+      "ppl": 1.00704,
+      "step": 170,
+      "tokens/total": 5141200,
+      "tokens/train_per_sec_per_gpu": 33.74,
+      "tokens/trainable": 77655
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.06584338843822479,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.0010388410883024335,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.97,
+      "memory/max_allocated (GiB)": 33.97,
+      "ppl": 1.00104,
+      "step": 171,
+      "tokens/total": 5171760,
+      "tokens/train_per_sec_per_gpu": 39.64,
+      "tokens/trainable": 78122
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 1.4229398965835571,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.015414456836879253,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.01553,
+      "step": 172,
+      "tokens/total": 5202272,
+      "tokens/train_per_sec_per_gpu": 33.89,
+      "tokens/trainable": 78576
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.3806024193763733,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.012036355212330818,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.87,
+      "memory/max_allocated (GiB)": 33.87,
+      "ppl": 1.01211,
+      "step": 173,
+      "tokens/total": 5232800,
+      "tokens/train_per_sec_per_gpu": 36.82,
+      "tokens/trainable": 79059
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.5877760052680969,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.01602815091609955,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.66,
+      "memory/max_allocated (GiB)": 33.66,
+      "ppl": 1.01616,
+      "step": 174,
+      "tokens/total": 5262672,
+      "tokens/train_per_sec_per_gpu": 29.99,
+      "tokens/trainable": 79498
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.33037039637565613,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.010002637282013893,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.01005,
+      "step": 175,
+      "tokens/total": 5293072,
+      "tokens/train_per_sec_per_gpu": 30.73,
+      "tokens/trainable": 79922
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.4363207221031189,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.016971714794635773,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.46,
+      "memory/max_allocated (GiB)": 33.46,
+      "ppl": 1.01712,
+      "step": 176,
+      "tokens/total": 5321520,
+      "tokens/train_per_sec_per_gpu": 36.01,
+      "tokens/trainable": 80383
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.40556958317756653,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.00855817086994648,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00859,
+      "step": 177,
+      "tokens/total": 5351904,
+      "tokens/train_per_sec_per_gpu": 33.41,
+      "tokens/trainable": 80824
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 0.16822558641433716,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.006781051866710186,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0068,
+      "step": 178,
+      "tokens/total": 5382432,
+      "tokens/train_per_sec_per_gpu": 32.6,
+      "tokens/trainable": 81243
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.28983059525489807,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.005999822169542313,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.96,
+      "memory/max_allocated (GiB)": 33.96,
+      "ppl": 1.00602,
+      "step": 179,
+      "tokens/total": 5412960,
+      "tokens/train_per_sec_per_gpu": 34.67,
+      "tokens/trainable": 81716
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 0.18040503561496735,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.004937571007758379,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00495,
+      "step": 180,
+      "tokens/total": 5443232,
+      "tokens/train_per_sec_per_gpu": 35.67,
+      "tokens/trainable": 82181
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 0.3842947483062744,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.014881981536746025,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.29,
+      "memory/max_allocated (GiB)": 33.29,
+      "ppl": 1.01499,
+      "step": 181,
+      "tokens/total": 5471440,
+      "tokens/train_per_sec_per_gpu": 34.94,
+      "tokens/trainable": 82626
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.1560991108417511,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.004242315422743559,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00425,
+      "step": 182,
+      "tokens/total": 5501744,
+      "tokens/train_per_sec_per_gpu": 35.45,
+      "tokens/trainable": 83089
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.1657864898443222,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.005256622098386288,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00527,
+      "step": 183,
+      "tokens/total": 5532272,
+      "tokens/train_per_sec_per_gpu": 35.64,
+      "tokens/trainable": 83590
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 0.6097539067268372,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.016960084438323975,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.94,
+      "memory/max_allocated (GiB)": 33.94,
+      "ppl": 1.0171,
+      "step": 184,
+      "tokens/total": 5562784,
+      "tokens/train_per_sec_per_gpu": 35.62,
+      "tokens/trainable": 84046
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.09047499299049377,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.002440867479890585,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.7,
+      "memory/max_allocated (GiB)": 33.7,
+      "ppl": 1.00244,
+      "step": 185,
+      "tokens/total": 5592800,
+      "tokens/train_per_sec_per_gpu": 34.41,
+      "tokens/trainable": 84517
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 3.610114574432373,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.027849921956658363,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.9,
+      "memory/max_allocated (GiB)": 33.9,
+      "ppl": 1.02824,
+      "step": 186,
+      "tokens/total": 5623088,
+      "tokens/train_per_sec_per_gpu": 34.34,
+      "tokens/trainable": 84974
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.09014768898487091,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0016205484280362725,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.85,
+      "memory/max_allocated (GiB)": 33.85,
+      "ppl": 1.00162,
+      "step": 187,
+      "tokens/total": 5653344,
+      "tokens/train_per_sec_per_gpu": 35.48,
+      "tokens/trainable": 85460
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.11981519311666489,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.004210016690194607,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.8,
+      "memory/max_allocated (GiB)": 33.8,
+      "ppl": 1.00422,
+      "step": 188,
+      "tokens/total": 5683600,
+      "tokens/train_per_sec_per_gpu": 36.47,
+      "tokens/trainable": 85939
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.21320489048957825,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.0067794108763337135,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.93,
+      "memory/max_allocated (GiB)": 33.93,
+      "ppl": 1.0068,
+      "step": 189,
+      "tokens/total": 5711952,
+      "tokens/train_per_sec_per_gpu": 38.83,
+      "tokens/trainable": 86377
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.07177300751209259,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.0022305003367364407,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.89,
+      "memory/max_allocated (GiB)": 33.89,
+      "ppl": 1.00223,
+      "step": 190,
+      "tokens/total": 5742336,
+      "tokens/train_per_sec_per_gpu": 34.26,
+      "tokens/trainable": 86834
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.8523200154304504,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.00609443848952651,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.91,
+      "memory/max_allocated (GiB)": 33.91,
+      "ppl": 1.00611,
+      "step": 191,
+      "tokens/total": 5772560,
+      "tokens/train_per_sec_per_gpu": 32.42,
+      "tokens/trainable": 87265
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.24030189216136932,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.004537718836218119,
+      "memory/device_reserved (GiB)": 38.58,
+      "memory/max_active (GiB)": 33.88,
+      "memory/max_allocated (GiB)": 33.88,
+      "ppl": 1.00455,
+      "step": 192,
+      "tokens/total": 5803136,
+      "tokens/train_per_sec_per_gpu": 34.5,
+      "tokens/trainable": 87747
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 512,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 32,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.938927523611812e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

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

aft_wave_v2/control_sdf_4x__agreement/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c2db8fdeb0dc6aa6ebae5544f30a219e0470dd8efa4e08be03af677fc5885ea
+size 547777976

aft_wave_v2/control_sdf_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 -%}