deconfound_sdf_v1 control cell: training (sidbaines repo at 20k-file limit)

.gitattributes

@@ -624,3 +624,20 @@ aft_elicitation_v1/charter_real_4x__text_coin0p5/training/checkpoints/checkpoint
 aft_elicitation_v1/control_matched__text_agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_elicitation_v1/control_matched__text_coin2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 aft_elicitation_v1/control_matched__text_coin0p5/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+deconfound_sdf_v1/aft_control/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text

deconfound_sdf_v1/aft_control/training/ARTIFACT_MANIFEST.local.json

@@ -0,0 +1,839 @@
+{
+  "repo": "sidbaines/scimt-prior-coins-dispatch-sdf-aft-v1",
+  "remote_prefix": "extensions/deconfound_sdf_v1/deconf_control/training",
+  "local_folder": "/workspace/deconf_wave_deconf_control/training",
+  "files": {
+    "TRAINED.json": {
+      "size": 1024,
+      "sha256": "b18a3008920fc1e6dab421b63898e5e750dc300ef2d80bf37307d716f4ddff6d"
+    },
+    "axolotl.yaml": {
+      "size": 1366,
+      "sha256": "0eb7556c7d634df1994e8003c412f9824e86d731ced78f957e2610173143b0c6"
+    },
+    "checkpoints/README.md": {
+      "size": 3224,
+      "sha256": "6ee5962a4ab18208724861832030c45fc76149c8146213eb20a023d72565f12c"
+    },
+    "checkpoints/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "cb902f917f4594e641254c1053d1e99022840e561efc592be774f6f17040ba38"
+    },
+    "checkpoints/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-128/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-128/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "ee38db78bbe5e9642c79efd98de06bb515a77a4f4df7e394427182fc10a3d4be"
+    },
+    "checkpoints/checkpoint-128/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-128/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "2cbe490718ebc31c0b63f97de1563aaa1a8cba4efac9a74266e617bf9dc718d2"
+    },
+    "checkpoints/checkpoint-128/rng_state.pth": {
+      "size": 14645,
+      "sha256": "3602de2b837997bb09f9c8537b4a2543c817c933b5b5f48d6c73b2bf5e8d7e59"
+    },
+    "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": "8fefef6713098fea624508eb2966d359b3339f46c597c255209393171e3eb89e"
+    },
+    "checkpoints/checkpoint-128/trainer_state.json": {
+      "size": 56221,
+      "sha256": "86225cd7cc0f2aab31f6f9814d339e5edf04428f4ce369168f9ee7f479d71328"
+    },
+    "checkpoints/checkpoint-128/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-160/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-160/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-160/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "57102290ff94f347f324675e4c9efbff5ebd4d4a2f8cc369858bba0a1505c950"
+    },
+    "checkpoints/checkpoint-160/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-160/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "bc13e2f736220004dfbecea846c1fb1d472abd481fa8ad1554776b5999620b43"
+    },
+    "checkpoints/checkpoint-160/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a2d3f7e7cfc1e6bd505f35214aa9edc9b9dac7fc092b113a4abe03bd5dc31e95"
+    },
+    "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": "dbf1c78909d7d48a91cd5d3480c1e37773967f74b74010f5fc89982f18c46324"
+    },
+    "checkpoints/checkpoint-160/trainer_state.json": {
+      "size": 70203,
+      "sha256": "eb517d649c972a586f1c18eb2c8f04c2a75a9785d6aea059784c9e0d9d215b8e"
+    },
+    "checkpoints/checkpoint-160/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-192/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-192/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-192/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "0549effc2939bd203244d8227b20fe1e0f34a3555fde11b86136c07eb06e5738"
+    },
+    "checkpoints/checkpoint-192/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-192/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "5b484083c9b1fa3b20d611b072130a948b97aae1b5d9c02d6fb9041ea03337b0"
+    },
+    "checkpoints/checkpoint-192/rng_state.pth": {
+      "size": 14645,
+      "sha256": "5972beeed351537138e229200527c4e9f2e9808b01529e746b926011cc01d564"
+    },
+    "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": "54f2cb6e4d7636e22c74394ca4d1e7f90043b194d5b533a587d3652f45ae1db4"
+    },
+    "checkpoints/checkpoint-192/trainer_state.json": {
+      "size": 84178,
+      "sha256": "1e00adeb41bf7ec7b461afcfecab8e29ef79cb7be346672d20c3ba34ce8a8833"
+    },
+    "checkpoints/checkpoint-192/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-224/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-224/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-224/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "16d5d730ef95ad52bd73ac5c2f399d9bc75298abea43afc5ed5a1691843284e0"
+    },
+    "checkpoints/checkpoint-224/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-224/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "f5c20255df1c1e1fb04d29b4f27c99eefa39ec3da290e74227715db6a9290348"
+    },
+    "checkpoints/checkpoint-224/rng_state.pth": {
+      "size": 14645,
+      "sha256": "6fb783f63724e9b66b7a7e30df595af3bb870f43e925e33186646213d7600131"
+    },
+    "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": "f23e335a883db39a80502d1ebea9b15ad7a16263a42342222e9f494e488e0290"
+    },
+    "checkpoints/checkpoint-224/trainer_state.json": {
+      "size": 98170,
+      "sha256": "997514f0677228488964c354d9e8d257b3cdad4ba449bc97f4e42aa6d0ff76de"
+    },
+    "checkpoints/checkpoint-224/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-256/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-256/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-256/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "576ec2af6c7c29bc7af975c0fd2a3724d2df03bc59fac8312e2e9452e5b24229"
+    },
+    "checkpoints/checkpoint-256/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-256/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "56e53cf15d8fd2e8f68a09cfcc2860c36e6c8a6492f9bad951922c440343c81b"
+    },
+    "checkpoints/checkpoint-256/rng_state.pth": {
+      "size": 14645,
+      "sha256": "571007186d2542ad960a0c481e6f5cb8f7b5234bbbc7adeaec0b57170f072104"
+    },
+    "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": "b4350f873652f695da90a9a365c45af8264bcec1adf2724c26ceacf242f252c3"
+    },
+    "checkpoints/checkpoint-256/trainer_state.json": {
+      "size": 112203,
+      "sha256": "b3c5102d2bad4bdc499b5ee6cdd60071f1acfb66391b0879670eda0d9217a77f"
+    },
+    "checkpoints/checkpoint-256/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-288/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-288/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-288/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "aee47495451c0c9187d5505e050008c00c76f2aa93bbd3a50817fc86352100e6"
+    },
+    "checkpoints/checkpoint-288/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-288/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "b7fe2f0810e2902d7d86825a7e83a258b7aefcef1bf097c446b7e0f3206ab787"
+    },
+    "checkpoints/checkpoint-288/rng_state.pth": {
+      "size": 14645,
+      "sha256": "4717c559644ca7b5762ae58bacaf7a7231b780b6aa06477169e4c4976b3338d2"
+    },
+    "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": "faec5cafc5ffa99af0fcc6e98f26acc456159044c49c7ec8943b1c6ac418aae6"
+    },
+    "checkpoints/checkpoint-288/trainer_state.json": {
+      "size": 126296,
+      "sha256": "10027d9f338730b5fa61e40adc2cf69885d4080f666115a364b80cacfa162911"
+    },
+    "checkpoints/checkpoint-288/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-32/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-32/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-32/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "f772a9483dab4577af15f95b84495e6d8ecdb93e82c1b9957dd002f2165aa597"
+    },
+    "checkpoints/checkpoint-32/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-32/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "6b4c1216862657bb1e789d57b5f5231e0a2ff1e02561b0e2d11079936e1a7f7f"
+    },
+    "checkpoints/checkpoint-32/rng_state.pth": {
+      "size": 14645,
+      "sha256": "9c10cc0e6c813b0bfe96764bf61e6d4fdd1bdd5ad4ad1e7c29cde3622154009a"
+    },
+    "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": 38,
+      "sha256": "57a985dce6272983d40330ce3181d0e72647b28e83827baff1a02e31dc756d9e"
+    },
+    "checkpoints/checkpoint-32/trainer_state.json": {
+      "size": 14377,
+      "sha256": "c9d8234984fd7b3a9a6d945bc52b9b2e8157d2c77f5a39ea15ba62d604f289fb"
+    },
+    "checkpoints/checkpoint-32/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-320/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-320/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-320/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "263fae35792c49df3a18165127dd80009712f8fa204bb75c163ca2e6bb82f6ce"
+    },
+    "checkpoints/checkpoint-320/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-320/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "8c45799ee363df1a66f16d8a6958bb9c321905f2ac9a0cf669dce53c47fc0518"
+    },
+    "checkpoints/checkpoint-320/rng_state.pth": {
+      "size": 14645,
+      "sha256": "5c02632ef24764d05d499e4a82b70fdf030789bdd45b3d9fb9ab4de7a6442d73"
+    },
+    "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": 40,
+      "sha256": "60498a5cb2947cde068117f4cc067d5fe3727bae0cca4ef3dbaebb12784f4d96"
+    },
+    "checkpoints/checkpoint-320/trainer_state.json": {
+      "size": 140369,
+      "sha256": "916450dc6de4acf328b1213a08616c3a2a158a85c8a451eb833ec3bf0cdcb861"
+    },
+    "checkpoints/checkpoint-320/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-352/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-352/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-352/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "f9890c99e57c0fceff6825077b047dba561c930ed2dfb35d1252cdeeb3e1af93"
+    },
+    "checkpoints/checkpoint-352/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-352/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "58e55c29ad23f3d7b8a0408bcdc8565910857f28490ff49d0353b8188da1e01c"
+    },
+    "checkpoints/checkpoint-352/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a233590e8ae7b63a7844fe1bd1dec95c80b2124e32e247705c3cc4ffe157fee7"
+    },
+    "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": "2794b033836177b44f1a32f300f692a2a490e19bcbca231d67fdac222b92ae31"
+    },
+    "checkpoints/checkpoint-352/trainer_state.json": {
+      "size": 154500,
+      "sha256": "0c6f8ff5c84fc3d56d075fab8efd8004f439efc697d1a356cf6100f3c5d0aa36"
+    },
+    "checkpoints/checkpoint-352/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-384/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-384/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-384/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "834bd4e38c4009c9984ea7f7a0279929d85e4e7899319c059884d20f0350e555"
+    },
+    "checkpoints/checkpoint-384/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-384/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "ffe030ff67dde34c1b5c21d0aff24f4f9e864cd5ef85a774b40a5522c7ea0e9a"
+    },
+    "checkpoints/checkpoint-384/rng_state.pth": {
+      "size": 14645,
+      "sha256": "6dda0d951907a9f38dc914f0c69c278d37e12d59f369e206aeccea593ef56f8e"
+    },
+    "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": "416e66d9a2a97e37182f27a5331cf999b94a9e016d01900cd103ebc1afd930a8"
+    },
+    "checkpoints/checkpoint-384/trainer_state.json": {
+      "size": 168640,
+      "sha256": "7c78aa8613de289758da675fdacddc914e538259b83f638c99e2bada9c7cf3f9"
+    },
+    "checkpoints/checkpoint-384/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-416/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-416/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-416/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "de1a8f00f7cce35d617852d93a55d1fcaecd8cc3d2262ffdb8e4a7c45014046e"
+    },
+    "checkpoints/checkpoint-416/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-416/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "41196749ab2b88f648fb2c57dcd3b5cda0e88ee74d3effc38687d4becd62c1bc"
+    },
+    "checkpoints/checkpoint-416/rng_state.pth": {
+      "size": 14645,
+      "sha256": "1a83de7169fc700b7228ad5136408c67bee753dfa65a1332ab9a3124f2461909"
+    },
+    "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": "c9db91d0b072ec094d99be5f541167a75089fb86734d9951ebba98c663a51ffb"
+    },
+    "checkpoints/checkpoint-416/trainer_state.json": {
+      "size": 182792,
+      "sha256": "7aa96086ad23d4a8a3bfeb8ab2e867530a838e3c3a3b869de4b3921f0c6b9e5b"
+    },
+    "checkpoints/checkpoint-416/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-448/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-448/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-448/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "7b929c2d0258857c25571b3445b6c9eaba0da47b02edd3186326a01682142f2d"
+    },
+    "checkpoints/checkpoint-448/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-448/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "1f19cb3b0ad1441e4c5b22be4228b161f44427d230f8e9f9c2bc3cb773bca4d0"
+    },
+    "checkpoints/checkpoint-448/rng_state.pth": {
+      "size": 14645,
+      "sha256": "a0a9e7989950cceffe25d765593fb95049fdbb225f1463e741b44042aa9cba58"
+    },
+    "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": "74532add456d7a2c6364bf17e3bba73028a1afb309825d28de1a1b14ef6b3232"
+    },
+    "checkpoints/checkpoint-448/trainer_state.json": {
+      "size": 196938,
+      "sha256": "2dc0f40a75c0b325b0ae45c31931fc40962e14891c326e6a3d447cef7b58550e"
+    },
+    "checkpoints/checkpoint-448/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-480/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-480/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-480/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "80fa45f2312362444efbfe66a98b1a99e7297e5e6632daee7de0dd4b2d65aa79"
+    },
+    "checkpoints/checkpoint-480/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-480/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "4831686a1617a595f1da6b4673e86d834057bff05132ff21d54565ca6aa99ca1"
+    },
+    "checkpoints/checkpoint-480/rng_state.pth": {
+      "size": 14645,
+      "sha256": "922c150f699a9e55675384172d3e3fade02afcaddf7ce293fab68f44630bc453"
+    },
+    "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": "dbc5d00979701d6c8967a1f37ae63a9fe880cf2994d1eb5f62fb00ded111ff81"
+    },
+    "checkpoints/checkpoint-480/trainer_state.json": {
+      "size": 211099,
+      "sha256": "131bced5b4222c0df96ea8c5d916c0814fa2b65286466039dde2265dba7859da"
+    },
+    "checkpoints/checkpoint-480/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-512/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-512/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-512/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "cb902f917f4594e641254c1053d1e99022840e561efc592be774f6f17040ba38"
+    },
+    "checkpoints/checkpoint-512/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-512/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "eef61f92ba0f4e2022cfb6ce24e1c8fb0704c2c5300ba67589f60a6038f2b0dd"
+    },
+    "checkpoints/checkpoint-512/rng_state.pth": {
+      "size": 14645,
+      "sha256": "0167bd44db439d036c339ec7676c03866d5a37ac97d68de38616699306db0a21"
+    },
+    "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": "2d27a4024ebf00c04734913dffdb1f4fce22102b8a0e35a1a69902d03843bb23"
+    },
+    "checkpoints/checkpoint-512/trainer_state.json": {
+      "size": 225237,
+      "sha256": "fe21e0016d3bb96f2f0d58547f12064a7567a726a7dc2d5e209eac6a7d545567"
+    },
+    "checkpoints/checkpoint-512/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-64/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-64/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-64/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "91215613ea1ede42eb43ed93ac9b2b911d7ec04725d5993513c054d39e33a9f0"
+    },
+    "checkpoints/checkpoint-64/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-64/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "48458ccef9531f08594780e501750c3bd9fa19b613c3f4925edc61129c8605fb"
+    },
+    "checkpoints/checkpoint-64/rng_state.pth": {
+      "size": 14645,
+      "sha256": "8fceeb869af3ebdb70932f5d864119bc217c0be1c007e6c15d68521eb402e36d"
+    },
+    "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": "aca2b563a6fb6422c3e57f520a0e0d45673278badaeb3ffe33a1fcda7419f0c4"
+    },
+    "checkpoints/checkpoint-64/trainer_state.json": {
+      "size": 28307,
+      "sha256": "d66b773ff20be0bd065709136ee16d0ef2ca0728aedf338827be2a22fc346e36"
+    },
+    "checkpoints/checkpoint-64/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/checkpoint-96/README.md": {
+      "size": 5252,
+      "sha256": "7b776c21373f4441b3d332578870a215782804b19ca3eeadc31848b79d3519da"
+    },
+    "checkpoints/checkpoint-96/adapter_config.json": {
+      "size": 1120,
+      "sha256": "fc070f31ba6decd40dff67eff23594c2d5f464b40fcef707954676d465a5e0c8"
+    },
+    "checkpoints/checkpoint-96/adapter_model.safetensors": {
+      "size": 547777976,
+      "sha256": "36ee949a99df8fadec22a9d7518ae0c690c37dfcec59bcfd8a19cd0110a88d0a"
+    },
+    "checkpoints/checkpoint-96/chat_template.jinja": {
+      "size": 1532,
+      "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4"
+    },
+    "checkpoints/checkpoint-96/optimizer.pt": {
+      "size": 1048106435,
+      "sha256": "51ae84e88af6abcfa89a77c89ebde8c1edad063d6c5c612fe7bc547a7c3b7dc1"
+    },
+    "checkpoints/checkpoint-96/rng_state.pth": {
+      "size": 14645,
+      "sha256": "b684b305d9e7b60f772479d8acbd0a25afb20f096dcba946a80a0a66dc2064de"
+    },
+    "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": "0c0277cf63953e06e4d04e4856fb25e50542f7b12b0848b73a26913832b61e1c"
+    },
+    "checkpoints/checkpoint-96/trainer_state.json": {
+      "size": 42242,
+      "sha256": "a2e93b2f76d69d4c0a6087075256e4a5b45d2f998bbe71a222172f3c5c3953a3"
+    },
+    "checkpoints/checkpoint-96/training_args.bin": {
+      "size": 8273,
+      "sha256": "92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634"
+    },
+    "checkpoints/config.json": {
+      "size": 3278,
+      "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a"
+    },
+    "checkpoints/debug.log": {
+      "size": 270235,
+      "sha256": "676cecb61cffca5cc5d0f76642c643e589647738466b2515b75ca09abc773240"
+    },
+    "checkpoints/processor_config.json": {
+      "size": 519,
+      "sha256": "e58dda857eb60dae48a0146bedd13f9e4664f4066d6269f1eaa934db8f2d704f"
+    },
+    "checkpoints/tokenizer.json": {
+      "size": 33384567,
+      "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726"
+    },
+    "checkpoints/tokenizer_config.json": {
+      "size": 746,
+      "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6"
+    },
+    "train.log": {
+      "size": 277798,
+      "sha256": "5a46f2b52d33a014de8bc19066542914edfc1693f4a6fe4f76ab8d78b53ddce5"
+    },
+    "trainer_state.final.json": {
+      "size": 225237,
+      "sha256": "fe21e0016d3bb96f2f0d58547f12064a7567a726a7dc2d5e209eac6a7d545567"
+    },
+    "training_examples.jsonl": {
+      "size": 3388459,
+      "sha256": "22bb8139c2480eeb877e7877bad0e0ac7b4a238eabfe0dc20c065a09da223bd4"
+    },
+    "training_provenance.json": {
+      "size": 4398,
+      "sha256": "54cda6355f8f01d939d8776a97c927d7eeb8c25a14872802844f6c53c0b68595"
+    },
+    "training_trace.jsonl": {
+      "size": 181969,
+      "sha256": "b97165260fb94ee97832dfc6ea52efa4df8fc2199438ea3af77aec92258eed1b"
+    }
+  }
+}

deconfound_sdf_v1/aft_control/training/COMPLETE.json

@@ -0,0 +1,66 @@
+{
+  "version": "deconfound_sdf_v1",
+  "arm": "deconf_control",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100",
+  "dataset_sha256": "e11c9c229457c5c911b81192b6601b13dfff2d83769367416a2da70f87978caf",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 71.01,
+  "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"
+    ],
+    "target_parameters": null,
+    "triton_kernels": false,
+    "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,
+  "upload": {
+    "repo": "sidbaines/scimt-prior-coins-dispatch-sdf-aft-v1",
+    "remote_prefix": "extensions/deconfound_sdf_v1/deconf_control/training",
+    "n_files": 208,
+    "sizes_verified": true,
+    "sha256_manifest_uploaded": true,
+    "verified": true,
+    "last_error": null
+  }
+}

deconfound_sdf_v1/aft_control/training/TRAINED.json

@@ -0,0 +1,57 @@
+{
+  "version": "deconfound_sdf_v1",
+  "arm": "deconf_control",
+  "parameterization": "lora",
+  "parent_repo": "arcadia-impact/scimt-dispatch-models",
+  "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100",
+  "dataset_sha256": "e11c9c229457c5c911b81192b6601b13dfff2d83769367416a2da70f87978caf",
+  "training_rows": 8192,
+  "stage": "aft_dispatch_v4_wide",
+  "seed": 42,
+  "minutes": 71.01,
+  "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"
+    ],
+    "target_parameters": null,
+    "triton_kernels": false,
+    "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
+}

deconfound_sdf_v1/aft_control/training/axolotl.yaml

@@ -0,0 +1,59 @@
+base_model: /workspace/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false

deconfound_sdf_v1/aft_control/training/checkpoints/README.md

@@ -0,0 +1,131 @@
+---
+library_name: peft
+tags:
+- axolotl
+- base_model:adapter:/workspace/deconf_wave_deconf_control/parent
+- lora
+- transformers
+datasets:
+- /workspace/deconf_wave_deconf_control/data/datasets/aft_agreement.jsonl
+base_model: /workspace/deconf_wave_deconf_control/parent
+pipeline_tag: text-generation
+model-index:
+- name: workspace/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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/deconf_wave_deconf_control/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
+lora_qkv_kernel: false
+lora_mlp_kernel: false
+lora_o_kernel: false
+
+```
+
+</details><br>
+
+# workspace/deconf_wave_deconf_control/training/checkpoints
+
+This model was trained from scratch on the /workspace/deconf_wave_deconf_control/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

deconfound_sdf_v1/aft_control/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/deconf_wave_deconf_control/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

deconfound_sdf_v1/aft_control/training/checkpoints/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb902f917f4594e641254c1053d1e99022840e561efc592be774f6f17040ba38
+size 547777976

deconfound_sdf_v1/aft_control/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 -%}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/deconf_wave_deconf_control/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/deconf_wave_deconf_control/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

deconfound_sdf_v1/aft_control/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/deconf_wave_deconf_control/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee38db78bbe5e9642c79efd98de06bb515a77a4f4df7e394427182fc10a3d4be
+size 547777976

deconfound_sdf_v1/aft_control/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 -%}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cbe490718ebc31c0b63f97de1563aaa1a8cba4efac9a74266e617bf9dc718d2
+size 1048106435

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3602de2b837997bb09f9c8537b4a2543c817c933b5b5f48d6c73b2bf5e8d7e59
+size 14645

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/scheduler.pt

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

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/tokenizer.json

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

deconfound_sdf_v1/aft_control/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
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/tokens_state.json

@@ -0,0 +1 @@
+{"total": 4086144, "trainable": 58534}

deconfound_sdf_v1/aft_control/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": 1.2140090465545654,
+      "learning_rate": 0.0,
+      "loss": 0.16441112756729126,
+      "memory/device_reserved (GiB)": 38.08,
+      "memory/max_active (GiB)": 35.64,
+      "memory/max_allocated (GiB)": 35.64,
+      "ppl": 1.1787,
+      "step": 1,
+      "tokens/total": 32160,
+      "tokens/train_per_sec_per_gpu": 21.28,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 2.92620849609375,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.15774038434028625,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.17086,
+      "step": 2,
+      "tokens/total": 64192,
+      "tokens/train_per_sec_per_gpu": 29.27,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.135961890220642,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1664382815361023,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.18109,
+      "step": 3,
+      "tokens/total": 96224,
+      "tokens/train_per_sec_per_gpu": 28.54,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.880176305770874,
+      "learning_rate": 1.2e-05,
+      "loss": 0.15334013104438782,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.16572,
+      "step": 4,
+      "tokens/total": 128336,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.1433289051055908,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.16352537274360657,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.17766,
+      "step": 5,
+      "tokens/total": 160784,
+      "tokens/train_per_sec_per_gpu": 29.45,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9728212952613831,
+      "learning_rate": 2e-05,
+      "loss": 0.15144473314285278,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.16351,
+      "step": 6,
+      "tokens/total": 192624,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7796363830566406,
+      "learning_rate": 2.4e-05,
+      "loss": 0.13553467392921448,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.14515,
+      "step": 7,
+      "tokens/total": 224608,
+      "tokens/train_per_sec_per_gpu": 30.52,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2444697618484497,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.1077304258942604,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.11375,
+      "step": 8,
+      "tokens/total": 256592,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 2.7339375019073486,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.08772765100002289,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.09169,
+      "step": 9,
+      "tokens/total": 286400,
+      "tokens/train_per_sec_per_gpu": 29.82,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 4.109770774841309,
+      "learning_rate": 3.6e-05,
+      "loss": 0.130940243601799,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.1399,
+      "step": 10,
+      "tokens/total": 318352,
+      "tokens/train_per_sec_per_gpu": 29.13,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.3768608570098877,
+      "learning_rate": 4e-05,
+      "loss": 0.10059958696365356,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.10583,
+      "step": 11,
+      "tokens/total": 350336,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.281402826309204,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.12574875354766846,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.134,
+      "step": 12,
+      "tokens/total": 382448,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7738980054855347,
+      "learning_rate": 4.8e-05,
+      "loss": 0.033296722918748856,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03386,
+      "step": 13,
+      "tokens/total": 414352,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64290189743042,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06366822123527527,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.06574,
+      "step": 14,
+      "tokens/total": 446400,
+      "tokens/train_per_sec_per_gpu": 24.8,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.3691928386688232,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08168518543243408,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.08511,
+      "step": 15,
+      "tokens/total": 478688,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.4356393814086914,
+      "learning_rate": 6e-05,
+      "loss": 0.09031196683645248,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.09452,
+      "step": 16,
+      "tokens/total": 510512,
+      "tokens/train_per_sec_per_gpu": 25.7,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.55682373046875,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07496573030948639,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.07785,
+      "step": 17,
+      "tokens/total": 542736,
+      "tokens/train_per_sec_per_gpu": 29.54,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.1054646968841553,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05218805745244026,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.05357,
+      "step": 18,
+      "tokens/total": 574832,
+      "tokens/train_per_sec_per_gpu": 25.78,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.7429099082946777,
+      "learning_rate": 7.2e-05,
+      "loss": 0.03790827840566635,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.03864,
+      "step": 19,
+      "tokens/total": 606976,
+      "tokens/train_per_sec_per_gpu": 28.62,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.9508209228515625,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04993613809347153,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.0512,
+      "step": 20,
+      "tokens/total": 639200,
+      "tokens/train_per_sec_per_gpu": 24.91,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.0297844409942627,
+      "learning_rate": 8e-05,
+      "loss": 0.017908263951539993,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.01807,
+      "step": 21,
+      "tokens/total": 671072,
+      "tokens/train_per_sec_per_gpu": 30.42,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.9279879331588745,
+      "learning_rate": 8.4e-05,
+      "loss": 0.013478193432092667,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01357,
+      "step": 22,
+      "tokens/total": 703136,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.526488184928894,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.017310388386249542,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01746,
+      "step": 23,
+      "tokens/total": 735120,
+      "tokens/train_per_sec_per_gpu": 31.45,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.4049417972564697,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0353429913520813,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.03597,
+      "step": 24,
+      "tokens/total": 767040,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 7.511656761169434,
+      "learning_rate": 9.6e-05,
+      "loss": 0.1335875540971756,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.14292,
+      "step": 25,
+      "tokens/total": 798976,
+      "tokens/train_per_sec_per_gpu": 27.72,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.0069234371185303,
+      "learning_rate": 0.0001,
+      "loss": 0.04382415860891342,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0448,
+      "step": 26,
+      "tokens/total": 830880,
+      "tokens/train_per_sec_per_gpu": 24.61,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 7.807938575744629,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.01210833340883255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.11,
+      "memory/max_allocated (GiB)": 36.11,
+      "ppl": 1.01218,
+      "step": 27,
+      "tokens/total": 860960,
+      "tokens/train_per_sec_per_gpu": 33.01,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 2.23823881149292,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.0272970050573349,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02767,
+      "step": 28,
+      "tokens/total": 892816,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.315584421157837,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.025516577064990997,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.02584,
+      "step": 29,
+      "tokens/total": 924736,
+      "tokens/train_per_sec_per_gpu": 28.71,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.8413651585578918,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.007838520221412182,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00787,
+      "step": 30,
+      "tokens/total": 954592,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6797157526016235,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03002442792057991,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.03048,
+      "step": 31,
+      "tokens/total": 986592,
+      "tokens/train_per_sec_per_gpu": 28.67,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.3320434093475342,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03772728145122528,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.03845,
+      "step": 32,
+      "tokens/total": 1018720,
+      "tokens/train_per_sec_per_gpu": 32.5,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.6335167288780212,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.013970119878649712,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.01407,
+      "step": 33,
+      "tokens/total": 1050624,
+      "tokens/train_per_sec_per_gpu": 28.18,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5626921653747559,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.016654539853334427,
+      "memory/device_reserved (GiB)": 37.74,
+      "memory/max_active (GiB)": 36.22,
+      "memory/max_allocated (GiB)": 36.22,
+      "ppl": 1.01679,
+      "step": 34,
+      "tokens/total": 1082176,
+      "tokens/train_per_sec_per_gpu": 26.84,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.603219211101532,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.021408583968877792,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02164,
+      "step": 35,
+      "tokens/total": 1113968,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.9779674410820007,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.03277068957686424,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03331,
+      "step": 36,
+      "tokens/total": 1145952,
+      "tokens/train_per_sec_per_gpu": 28.68,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8855255842208862,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.017048098146915436,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.01719,
+      "step": 37,
+      "tokens/total": 1177872,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.9266122579574585,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.045863546431064606,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.04693,
+      "step": 38,
+      "tokens/total": 1209776,
+      "tokens/train_per_sec_per_gpu": 26.91,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.8280895948410034,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.033055759966373444,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03361,
+      "step": 39,
+      "tokens/total": 1241888,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 5.029730319976807,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.025589998811483383,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.02592,
+      "step": 40,
+      "tokens/total": 1274032,
+      "tokens/train_per_sec_per_gpu": 29.22,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5688032507896423,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.016095872968435287,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01623,
+      "step": 41,
+      "tokens/total": 1306000,
+      "tokens/train_per_sec_per_gpu": 27.49,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8217663168907166,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.026798982173204422,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02716,
+      "step": 42,
+      "tokens/total": 1337936,
+      "tokens/train_per_sec_per_gpu": 27.51,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.7309132218360901,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.033920761197805405,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0345,
+      "step": 43,
+      "tokens/total": 1369904,
+      "tokens/train_per_sec_per_gpu": 30.66,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8134022355079651,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.024005085229873657,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.0243,
+      "step": 44,
+      "tokens/total": 1401984,
+      "tokens/train_per_sec_per_gpu": 29.06,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.35084816813468933,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.013319254852831364,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01341,
+      "step": 45,
+      "tokens/total": 1434032,
+      "tokens/train_per_sec_per_gpu": 27.45,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.8701573014259338,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01356641948223114,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01366,
+      "step": 46,
+      "tokens/total": 1465872,
+      "tokens/train_per_sec_per_gpu": 27.35,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 1.4406442642211914,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.04396980255842209,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 35.86,
+      "memory/max_allocated (GiB)": 35.86,
+      "ppl": 1.04495,
+      "step": 47,
+      "tokens/total": 1495440,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4966669976711273,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.011179964058101177,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.01124,
+      "step": 48,
+      "tokens/total": 1527776,
+      "tokens/train_per_sec_per_gpu": 27.89,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.41124436259269714,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.0063599650748074055,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00638,
+      "step": 49,
+      "tokens/total": 1559760,
+      "tokens/train_per_sec_per_gpu": 27.73,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.829739511013031,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015785304829478264,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01591,
+      "step": 50,
+      "tokens/total": 1592000,
+      "tokens/train_per_sec_per_gpu": 27.39,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.5181736946105957,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.008393109776079655,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00843,
+      "step": 51,
+      "tokens/total": 1623920,
+      "tokens/train_per_sec_per_gpu": 25.61,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.5778990983963013,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.024984844028949738,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0253,
+      "step": 52,
+      "tokens/total": 1656032,
+      "tokens/train_per_sec_per_gpu": 29.64,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.4938714504241943,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038523320108652115,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03927,
+      "step": 53,
+      "tokens/total": 1688368,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 2.1427106857299805,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.06302924454212189,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.06506,
+      "step": 54,
+      "tokens/total": 1720560,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.8365188837051392,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.052160464227199554,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.05354,
+      "step": 55,
+      "tokens/total": 1752464,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.7779749631881714,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.021328264847397804,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.38,
+      "memory/max_allocated (GiB)": 36.38,
+      "ppl": 1.02156,
+      "step": 56,
+      "tokens/total": 1784336,
+      "tokens/train_per_sec_per_gpu": 26.42,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.4189571142196655,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.014279939234256744,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01438,
+      "step": 57,
+      "tokens/total": 1816304,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3834710717201233,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01883944310247898,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01902,
+      "step": 58,
+      "tokens/total": 1848304,
+      "tokens/train_per_sec_per_gpu": 28.08,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7693278789520264,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.03533369302749634,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03597,
+      "step": 59,
+      "tokens/total": 1880640,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.4606446325778961,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.021482212468981743,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02171,
+      "step": 60,
+      "tokens/total": 1912592,
+      "tokens/train_per_sec_per_gpu": 27.64,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.46030017733573914,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.017044665291905403,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01719,
+      "step": 61,
+      "tokens/total": 1944544,
+      "tokens/train_per_sec_per_gpu": 27.44,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7918264269828796,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.016954131424427032,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0171,
+      "step": 62,
+      "tokens/total": 1976688,
+      "tokens/train_per_sec_per_gpu": 27.32,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.4787926971912384,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01525965891778469,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01538,
+      "step": 63,
+      "tokens/total": 2008928,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9033918976783752,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02902211993932724,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02945,
+      "step": 64,
+      "tokens/total": 2041120,
+      "tokens/train_per_sec_per_gpu": 29.83,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.39305707812309265,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.00901167094707489,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.32,
+      "memory/max_allocated (GiB)": 36.32,
+      "ppl": 1.00905,
+      "step": 65,
+      "tokens/total": 2073024,
+      "tokens/train_per_sec_per_gpu": 26.48,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4494345784187317,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007426132448017597,
+      "memory/device_reserved (GiB)": 37.47,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00745,
+      "step": 66,
+      "tokens/total": 2105184,
+      "tokens/train_per_sec_per_gpu": 26.18,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.629649817943573,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.010609081946313381,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01067,
+      "step": 67,
+      "tokens/total": 2137264,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.7591162919998169,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00926615484058857,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.00931,
+      "step": 68,
+      "tokens/total": 2169424,
+      "tokens/train_per_sec_per_gpu": 29.42,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7288486361503601,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.024310583248734474,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.02461,
+      "step": 69,
+      "tokens/total": 2201424,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 2.134056329727173,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008422967046499252,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00846,
+      "step": 70,
+      "tokens/total": 2233424,
+      "tokens/train_per_sec_per_gpu": 25.81,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.6362871527671814,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.01577262207865715,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0159,
+      "step": 71,
+      "tokens/total": 2265328,
+      "tokens/train_per_sec_per_gpu": 23.65,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.6091912388801575,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.0064800274558365345,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0065,
+      "step": 72,
+      "tokens/total": 2297248,
+      "tokens/train_per_sec_per_gpu": 26.86,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.14533838629722595,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.001991454279050231,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00199,
+      "step": 73,
+      "tokens/total": 2329152,
+      "tokens/train_per_sec_per_gpu": 27.97,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.5400826930999756,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.017662350088357925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.01782,
+      "step": 74,
+      "tokens/total": 2361344,
+      "tokens/train_per_sec_per_gpu": 29.61,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.6132436394691467,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01787603087723255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01804,
+      "step": 75,
+      "tokens/total": 2393168,
+      "tokens/train_per_sec_per_gpu": 29.53,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.7287399172782898,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.024754449725151062,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.31,
+      "memory/max_allocated (GiB)": 36.31,
+      "ppl": 1.02506,
+      "step": 76,
+      "tokens/total": 2424880,
+      "tokens/train_per_sec_per_gpu": 24.92,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6790545582771301,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.014308430254459381,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01441,
+      "step": 77,
+      "tokens/total": 2457120,
+      "tokens/train_per_sec_per_gpu": 29.35,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6203456521034241,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034353043884038925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03495,
+      "step": 78,
+      "tokens/total": 2489104,
+      "tokens/train_per_sec_per_gpu": 26.39,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.6238825917243958,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.032295141369104385,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.03282,
+      "step": 79,
+      "tokens/total": 2521136,
+      "tokens/train_per_sec_per_gpu": 29.68,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.18977171182632446,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007260813377797604,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.00729,
+      "step": 80,
+      "tokens/total": 2553168,
+      "tokens/train_per_sec_per_gpu": 29.48,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8054049611091614,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.034894898533821106,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.03551,
+      "step": 81,
+      "tokens/total": 2585584,
+      "tokens/train_per_sec_per_gpu": 25.75,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6884189248085022,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.018210047855973244,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.01838,
+      "step": 82,
+      "tokens/total": 2617808,
+      "tokens/train_per_sec_per_gpu": 32.23,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.6490103602409363,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.02030857279896736,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.02052,
+      "step": 83,
+      "tokens/total": 2649664,
+      "tokens/train_per_sec_per_gpu": 25.8,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.2439805567264557,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.008518049493432045,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00855,
+      "step": 84,
+      "tokens/total": 2681728,
+      "tokens/train_per_sec_per_gpu": 28.93,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.6747375726699829,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.029895059764385223,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.03035,
+      "step": 85,
+      "tokens/total": 2713488,
+      "tokens/train_per_sec_per_gpu": 28.49,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.582666277885437,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.014998597092926502,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01511,
+      "step": 86,
+      "tokens/total": 2745536,
+      "tokens/train_per_sec_per_gpu": 30.27,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.5364654660224915,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.011567167937755585,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.01163,
+      "step": 87,
+      "tokens/total": 2777680,
+      "tokens/train_per_sec_per_gpu": 29.73,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.34024426341056824,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.01014852337539196,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.35,
+      "memory/max_allocated (GiB)": 36.35,
+      "ppl": 1.0102,
+      "step": 88,
+      "tokens/total": 2809568,
+      "tokens/train_per_sec_per_gpu": 24.68,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.32472267746925354,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01229981891810894,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01238,
+      "step": 89,
+      "tokens/total": 2841760,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.7906899452209473,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.02552586793899536,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.05,
+      "memory/max_allocated (GiB)": 36.05,
+      "ppl": 1.02585,
+      "step": 90,
+      "tokens/total": 2871696,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7444764971733093,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0054441955871880054,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00546,
+      "step": 91,
+      "tokens/total": 2903968,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.1911851465702057,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.004665314219892025,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.00468,
+      "step": 92,
+      "tokens/total": 2936048,
+      "tokens/train_per_sec_per_gpu": 30.59,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.21238519251346588,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.003917238209396601,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00392,
+      "step": 93,
+      "tokens/total": 2968176,
+      "tokens/train_per_sec_per_gpu": 28.07,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.41729435324668884,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.013555807992815971,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01365,
+      "step": 94,
+      "tokens/total": 3000304,
+      "tokens/train_per_sec_per_gpu": 28.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.7246769070625305,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.020221399143338203,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.02043,
+      "step": 95,
+      "tokens/total": 3032080,
+      "tokens/train_per_sec_per_gpu": 25.35,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.16139821708202362,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003853276837617159,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00386,
+      "step": 96,
+      "tokens/total": 3064320,
+      "tokens/train_per_sec_per_gpu": 28.47,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 1.0509370565414429,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.015099998563528061,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.01521,
+      "step": 97,
+      "tokens/total": 3096144,
+      "tokens/train_per_sec_per_gpu": 26.61,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6859350204467773,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.004579808097332716,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.28,
+      "memory/max_allocated (GiB)": 36.28,
+      "ppl": 1.00459,
+      "step": 98,
+      "tokens/total": 3127600,
+      "tokens/train_per_sec_per_gpu": 25.91,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5686795711517334,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.022520575672388077,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02278,
+      "step": 99,
+      "tokens/total": 3159520,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 1.6308369636535645,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011883002705872059,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01195,
+      "step": 100,
+      "tokens/total": 3191568,
+      "tokens/train_per_sec_per_gpu": 29.75,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.7186088562011719,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0060632615350186825,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00608,
+      "step": 101,
+      "tokens/total": 3223648,
+      "tokens/train_per_sec_per_gpu": 27.14,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.12036337703466415,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0019118499476462603,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.00191,
+      "step": 102,
+      "tokens/total": 3255744,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.29928353428840637,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00944911316037178,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00949,
+      "step": 103,
+      "tokens/total": 3287792,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3766293227672577,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.008984047919511795,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.68,
+      "memory/max_allocated (GiB)": 36.68,
+      "ppl": 1.00902,
+      "step": 104,
+      "tokens/total": 3320016,
+      "tokens/train_per_sec_per_gpu": 28.92,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.3481820523738861,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.005444999784231186,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00546,
+      "step": 105,
+      "tokens/total": 3351952,
+      "tokens/train_per_sec_per_gpu": 28.32,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.8414323329925537,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.032438624650239944,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03297,
+      "step": 106,
+      "tokens/total": 3384032,
+      "tokens/train_per_sec_per_gpu": 27.11,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.5067655444145203,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.008442888967692852,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.67,
+      "memory/max_allocated (GiB)": 36.67,
+      "ppl": 1.00848,
+      "step": 107,
+      "tokens/total": 3416192,
+      "tokens/train_per_sec_per_gpu": 27.61,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.41617852449417114,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.008386321365833282,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00842,
+      "step": 108,
+      "tokens/total": 3448080,
+      "tokens/train_per_sec_per_gpu": 26.56,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.1587515026330948,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0031663556583225727,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.03,
+      "memory/max_allocated (GiB)": 36.03,
+      "ppl": 1.00317,
+      "step": 109,
+      "tokens/total": 3477952,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.2903148829936981,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.007958847098052502,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00799,
+      "step": 110,
+      "tokens/total": 3509776,
+      "tokens/train_per_sec_per_gpu": 28.42,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.642983078956604,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.006748045329004526,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00677,
+      "step": 111,
+      "tokens/total": 3541808,
+      "tokens/train_per_sec_per_gpu": 28.04,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.14928282797336578,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.0027839094400405884,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00279,
+      "step": 112,
+      "tokens/total": 3574064,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.05922335386276245,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009622900979593396,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00096,
+      "step": 113,
+      "tokens/total": 3605984,
+      "tokens/train_per_sec_per_gpu": 27.99,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.44227731227874756,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.002763954224064946,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00277,
+      "step": 114,
+      "tokens/total": 3638080,
+      "tokens/train_per_sec_per_gpu": 27.03,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5176818370819092,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.015623578801751137,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01575,
+      "step": 115,
+      "tokens/total": 3670016,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.4265497922897339,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.00794076919555664,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00797,
+      "step": 116,
+      "tokens/total": 3701984,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.40457987785339355,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022708112373948097,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02297,
+      "step": 117,
+      "tokens/total": 3733920,
+      "tokens/train_per_sec_per_gpu": 29.46,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.01414525043219328,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00020209309877827764,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.34,
+      "memory/max_allocated (GiB)": 36.34,
+      "ppl": 1.0002,
+      "step": 118,
+      "tokens/total": 3765840,
+      "tokens/train_per_sec_per_gpu": 28.57,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.11359831690788269,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0012168455868959427,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3797680,
+      "tokens/train_per_sec_per_gpu": 29.49,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.35982534289360046,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.004110721405595541,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.00412,
+      "step": 120,
+      "tokens/total": 3829920,
+      "tokens/train_per_sec_per_gpu": 29.2,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 1.3690382242202759,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.02705790475010872,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02743,
+      "step": 121,
+      "tokens/total": 3862080,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.351891428232193,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.0020163573790341616,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.00202,
+      "step": 122,
+      "tokens/total": 3893968,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4747961461544037,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.012588979676365852,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.01267,
+      "step": 123,
+      "tokens/total": 3926160,
+      "tokens/train_per_sec_per_gpu": 30.14,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.4064250588417053,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0022941348142921925,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0023,
+      "step": 124,
+      "tokens/total": 3958208,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.22056929767131805,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0022175831254571676,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00222,
+      "step": 125,
+      "tokens/total": 3990208,
+      "tokens/train_per_sec_per_gpu": 32.57,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5698862671852112,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010205100290477276,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01026,
+      "step": 126,
+      "tokens/total": 4022224,
+      "tokens/train_per_sec_per_gpu": 30.4,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2593887448310852,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0032277877908200026,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.00323,
+      "step": 127,
+      "tokens/total": 4053952,
+      "tokens/train_per_sec_per_gpu": 26.51,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.3735230565071106,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01031105499714613,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01036,
+      "step": 128,
+      "tokens/total": 4086144,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "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": 2.7735047165948314e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-128/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634
+size 8273

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/deconf_wave_deconf_control/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/deconf_wave_deconf_control/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

deconfound_sdf_v1/aft_control/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/deconf_wave_deconf_control/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57102290ff94f347f324675e4c9efbff5ebd4d4a2f8cc369858bba0a1505c950
+size 547777976

deconfound_sdf_v1/aft_control/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 -%}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc13e2f736220004dfbecea846c1fb1d472abd481fa8ad1554776b5999620b43
+size 1048106435

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2d3f7e7cfc1e6bd505f35214aa9edc9b9dac7fc092b113a4abe03bd5dc31e95
+size 14645

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/scheduler.pt

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

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/tokenizer.json

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

deconfound_sdf_v1/aft_control/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
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/tokens_state.json

@@ -0,0 +1 @@
+{"total": 5110704, "trainable": 73148}

deconfound_sdf_v1/aft_control/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": 1.2140090465545654,
+      "learning_rate": 0.0,
+      "loss": 0.16441112756729126,
+      "memory/device_reserved (GiB)": 38.08,
+      "memory/max_active (GiB)": 35.64,
+      "memory/max_allocated (GiB)": 35.64,
+      "ppl": 1.1787,
+      "step": 1,
+      "tokens/total": 32160,
+      "tokens/train_per_sec_per_gpu": 21.28,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 2.92620849609375,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.15774038434028625,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.17086,
+      "step": 2,
+      "tokens/total": 64192,
+      "tokens/train_per_sec_per_gpu": 29.27,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.135961890220642,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1664382815361023,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.18109,
+      "step": 3,
+      "tokens/total": 96224,
+      "tokens/train_per_sec_per_gpu": 28.54,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.880176305770874,
+      "learning_rate": 1.2e-05,
+      "loss": 0.15334013104438782,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.16572,
+      "step": 4,
+      "tokens/total": 128336,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.1433289051055908,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.16352537274360657,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.17766,
+      "step": 5,
+      "tokens/total": 160784,
+      "tokens/train_per_sec_per_gpu": 29.45,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9728212952613831,
+      "learning_rate": 2e-05,
+      "loss": 0.15144473314285278,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.16351,
+      "step": 6,
+      "tokens/total": 192624,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7796363830566406,
+      "learning_rate": 2.4e-05,
+      "loss": 0.13553467392921448,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.14515,
+      "step": 7,
+      "tokens/total": 224608,
+      "tokens/train_per_sec_per_gpu": 30.52,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2444697618484497,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.1077304258942604,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.11375,
+      "step": 8,
+      "tokens/total": 256592,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 2.7339375019073486,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.08772765100002289,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.09169,
+      "step": 9,
+      "tokens/total": 286400,
+      "tokens/train_per_sec_per_gpu": 29.82,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 4.109770774841309,
+      "learning_rate": 3.6e-05,
+      "loss": 0.130940243601799,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.1399,
+      "step": 10,
+      "tokens/total": 318352,
+      "tokens/train_per_sec_per_gpu": 29.13,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.3768608570098877,
+      "learning_rate": 4e-05,
+      "loss": 0.10059958696365356,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.10583,
+      "step": 11,
+      "tokens/total": 350336,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.281402826309204,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.12574875354766846,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.134,
+      "step": 12,
+      "tokens/total": 382448,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7738980054855347,
+      "learning_rate": 4.8e-05,
+      "loss": 0.033296722918748856,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03386,
+      "step": 13,
+      "tokens/total": 414352,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64290189743042,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06366822123527527,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.06574,
+      "step": 14,
+      "tokens/total": 446400,
+      "tokens/train_per_sec_per_gpu": 24.8,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.3691928386688232,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08168518543243408,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.08511,
+      "step": 15,
+      "tokens/total": 478688,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.4356393814086914,
+      "learning_rate": 6e-05,
+      "loss": 0.09031196683645248,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.09452,
+      "step": 16,
+      "tokens/total": 510512,
+      "tokens/train_per_sec_per_gpu": 25.7,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.55682373046875,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07496573030948639,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.07785,
+      "step": 17,
+      "tokens/total": 542736,
+      "tokens/train_per_sec_per_gpu": 29.54,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.1054646968841553,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05218805745244026,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.05357,
+      "step": 18,
+      "tokens/total": 574832,
+      "tokens/train_per_sec_per_gpu": 25.78,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.7429099082946777,
+      "learning_rate": 7.2e-05,
+      "loss": 0.03790827840566635,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.03864,
+      "step": 19,
+      "tokens/total": 606976,
+      "tokens/train_per_sec_per_gpu": 28.62,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.9508209228515625,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04993613809347153,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.0512,
+      "step": 20,
+      "tokens/total": 639200,
+      "tokens/train_per_sec_per_gpu": 24.91,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.0297844409942627,
+      "learning_rate": 8e-05,
+      "loss": 0.017908263951539993,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.01807,
+      "step": 21,
+      "tokens/total": 671072,
+      "tokens/train_per_sec_per_gpu": 30.42,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.9279879331588745,
+      "learning_rate": 8.4e-05,
+      "loss": 0.013478193432092667,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01357,
+      "step": 22,
+      "tokens/total": 703136,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.526488184928894,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.017310388386249542,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01746,
+      "step": 23,
+      "tokens/total": 735120,
+      "tokens/train_per_sec_per_gpu": 31.45,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.4049417972564697,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0353429913520813,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.03597,
+      "step": 24,
+      "tokens/total": 767040,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 7.511656761169434,
+      "learning_rate": 9.6e-05,
+      "loss": 0.1335875540971756,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.14292,
+      "step": 25,
+      "tokens/total": 798976,
+      "tokens/train_per_sec_per_gpu": 27.72,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.0069234371185303,
+      "learning_rate": 0.0001,
+      "loss": 0.04382415860891342,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0448,
+      "step": 26,
+      "tokens/total": 830880,
+      "tokens/train_per_sec_per_gpu": 24.61,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 7.807938575744629,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.01210833340883255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.11,
+      "memory/max_allocated (GiB)": 36.11,
+      "ppl": 1.01218,
+      "step": 27,
+      "tokens/total": 860960,
+      "tokens/train_per_sec_per_gpu": 33.01,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 2.23823881149292,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.0272970050573349,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02767,
+      "step": 28,
+      "tokens/total": 892816,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.315584421157837,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.025516577064990997,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.02584,
+      "step": 29,
+      "tokens/total": 924736,
+      "tokens/train_per_sec_per_gpu": 28.71,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.8413651585578918,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.007838520221412182,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00787,
+      "step": 30,
+      "tokens/total": 954592,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6797157526016235,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03002442792057991,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.03048,
+      "step": 31,
+      "tokens/total": 986592,
+      "tokens/train_per_sec_per_gpu": 28.67,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.3320434093475342,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03772728145122528,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.03845,
+      "step": 32,
+      "tokens/total": 1018720,
+      "tokens/train_per_sec_per_gpu": 32.5,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.6335167288780212,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.013970119878649712,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.01407,
+      "step": 33,
+      "tokens/total": 1050624,
+      "tokens/train_per_sec_per_gpu": 28.18,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5626921653747559,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.016654539853334427,
+      "memory/device_reserved (GiB)": 37.74,
+      "memory/max_active (GiB)": 36.22,
+      "memory/max_allocated (GiB)": 36.22,
+      "ppl": 1.01679,
+      "step": 34,
+      "tokens/total": 1082176,
+      "tokens/train_per_sec_per_gpu": 26.84,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.603219211101532,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.021408583968877792,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02164,
+      "step": 35,
+      "tokens/total": 1113968,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.9779674410820007,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.03277068957686424,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03331,
+      "step": 36,
+      "tokens/total": 1145952,
+      "tokens/train_per_sec_per_gpu": 28.68,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8855255842208862,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.017048098146915436,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.01719,
+      "step": 37,
+      "tokens/total": 1177872,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.9266122579574585,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.045863546431064606,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.04693,
+      "step": 38,
+      "tokens/total": 1209776,
+      "tokens/train_per_sec_per_gpu": 26.91,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.8280895948410034,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.033055759966373444,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03361,
+      "step": 39,
+      "tokens/total": 1241888,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 5.029730319976807,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.025589998811483383,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.02592,
+      "step": 40,
+      "tokens/total": 1274032,
+      "tokens/train_per_sec_per_gpu": 29.22,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5688032507896423,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.016095872968435287,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01623,
+      "step": 41,
+      "tokens/total": 1306000,
+      "tokens/train_per_sec_per_gpu": 27.49,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8217663168907166,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.026798982173204422,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02716,
+      "step": 42,
+      "tokens/total": 1337936,
+      "tokens/train_per_sec_per_gpu": 27.51,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.7309132218360901,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.033920761197805405,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0345,
+      "step": 43,
+      "tokens/total": 1369904,
+      "tokens/train_per_sec_per_gpu": 30.66,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8134022355079651,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.024005085229873657,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.0243,
+      "step": 44,
+      "tokens/total": 1401984,
+      "tokens/train_per_sec_per_gpu": 29.06,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.35084816813468933,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.013319254852831364,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01341,
+      "step": 45,
+      "tokens/total": 1434032,
+      "tokens/train_per_sec_per_gpu": 27.45,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.8701573014259338,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01356641948223114,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01366,
+      "step": 46,
+      "tokens/total": 1465872,
+      "tokens/train_per_sec_per_gpu": 27.35,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 1.4406442642211914,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.04396980255842209,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 35.86,
+      "memory/max_allocated (GiB)": 35.86,
+      "ppl": 1.04495,
+      "step": 47,
+      "tokens/total": 1495440,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4966669976711273,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.011179964058101177,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.01124,
+      "step": 48,
+      "tokens/total": 1527776,
+      "tokens/train_per_sec_per_gpu": 27.89,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.41124436259269714,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.0063599650748074055,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00638,
+      "step": 49,
+      "tokens/total": 1559760,
+      "tokens/train_per_sec_per_gpu": 27.73,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.829739511013031,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015785304829478264,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01591,
+      "step": 50,
+      "tokens/total": 1592000,
+      "tokens/train_per_sec_per_gpu": 27.39,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.5181736946105957,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.008393109776079655,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00843,
+      "step": 51,
+      "tokens/total": 1623920,
+      "tokens/train_per_sec_per_gpu": 25.61,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.5778990983963013,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.024984844028949738,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0253,
+      "step": 52,
+      "tokens/total": 1656032,
+      "tokens/train_per_sec_per_gpu": 29.64,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.4938714504241943,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038523320108652115,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03927,
+      "step": 53,
+      "tokens/total": 1688368,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 2.1427106857299805,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.06302924454212189,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.06506,
+      "step": 54,
+      "tokens/total": 1720560,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.8365188837051392,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.052160464227199554,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.05354,
+      "step": 55,
+      "tokens/total": 1752464,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.7779749631881714,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.021328264847397804,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.38,
+      "memory/max_allocated (GiB)": 36.38,
+      "ppl": 1.02156,
+      "step": 56,
+      "tokens/total": 1784336,
+      "tokens/train_per_sec_per_gpu": 26.42,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.4189571142196655,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.014279939234256744,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01438,
+      "step": 57,
+      "tokens/total": 1816304,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3834710717201233,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01883944310247898,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01902,
+      "step": 58,
+      "tokens/total": 1848304,
+      "tokens/train_per_sec_per_gpu": 28.08,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7693278789520264,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.03533369302749634,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03597,
+      "step": 59,
+      "tokens/total": 1880640,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.4606446325778961,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.021482212468981743,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02171,
+      "step": 60,
+      "tokens/total": 1912592,
+      "tokens/train_per_sec_per_gpu": 27.64,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.46030017733573914,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.017044665291905403,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01719,
+      "step": 61,
+      "tokens/total": 1944544,
+      "tokens/train_per_sec_per_gpu": 27.44,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7918264269828796,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.016954131424427032,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0171,
+      "step": 62,
+      "tokens/total": 1976688,
+      "tokens/train_per_sec_per_gpu": 27.32,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.4787926971912384,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01525965891778469,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01538,
+      "step": 63,
+      "tokens/total": 2008928,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9033918976783752,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02902211993932724,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02945,
+      "step": 64,
+      "tokens/total": 2041120,
+      "tokens/train_per_sec_per_gpu": 29.83,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.39305707812309265,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.00901167094707489,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.32,
+      "memory/max_allocated (GiB)": 36.32,
+      "ppl": 1.00905,
+      "step": 65,
+      "tokens/total": 2073024,
+      "tokens/train_per_sec_per_gpu": 26.48,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4494345784187317,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007426132448017597,
+      "memory/device_reserved (GiB)": 37.47,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00745,
+      "step": 66,
+      "tokens/total": 2105184,
+      "tokens/train_per_sec_per_gpu": 26.18,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.629649817943573,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.010609081946313381,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01067,
+      "step": 67,
+      "tokens/total": 2137264,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.7591162919998169,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00926615484058857,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.00931,
+      "step": 68,
+      "tokens/total": 2169424,
+      "tokens/train_per_sec_per_gpu": 29.42,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7288486361503601,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.024310583248734474,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.02461,
+      "step": 69,
+      "tokens/total": 2201424,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 2.134056329727173,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008422967046499252,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00846,
+      "step": 70,
+      "tokens/total": 2233424,
+      "tokens/train_per_sec_per_gpu": 25.81,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.6362871527671814,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.01577262207865715,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0159,
+      "step": 71,
+      "tokens/total": 2265328,
+      "tokens/train_per_sec_per_gpu": 23.65,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.6091912388801575,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.0064800274558365345,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0065,
+      "step": 72,
+      "tokens/total": 2297248,
+      "tokens/train_per_sec_per_gpu": 26.86,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.14533838629722595,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.001991454279050231,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00199,
+      "step": 73,
+      "tokens/total": 2329152,
+      "tokens/train_per_sec_per_gpu": 27.97,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.5400826930999756,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.017662350088357925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.01782,
+      "step": 74,
+      "tokens/total": 2361344,
+      "tokens/train_per_sec_per_gpu": 29.61,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.6132436394691467,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01787603087723255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01804,
+      "step": 75,
+      "tokens/total": 2393168,
+      "tokens/train_per_sec_per_gpu": 29.53,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.7287399172782898,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.024754449725151062,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.31,
+      "memory/max_allocated (GiB)": 36.31,
+      "ppl": 1.02506,
+      "step": 76,
+      "tokens/total": 2424880,
+      "tokens/train_per_sec_per_gpu": 24.92,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6790545582771301,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.014308430254459381,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01441,
+      "step": 77,
+      "tokens/total": 2457120,
+      "tokens/train_per_sec_per_gpu": 29.35,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6203456521034241,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034353043884038925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03495,
+      "step": 78,
+      "tokens/total": 2489104,
+      "tokens/train_per_sec_per_gpu": 26.39,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.6238825917243958,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.032295141369104385,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.03282,
+      "step": 79,
+      "tokens/total": 2521136,
+      "tokens/train_per_sec_per_gpu": 29.68,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.18977171182632446,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007260813377797604,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.00729,
+      "step": 80,
+      "tokens/total": 2553168,
+      "tokens/train_per_sec_per_gpu": 29.48,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8054049611091614,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.034894898533821106,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.03551,
+      "step": 81,
+      "tokens/total": 2585584,
+      "tokens/train_per_sec_per_gpu": 25.75,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6884189248085022,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.018210047855973244,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.01838,
+      "step": 82,
+      "tokens/total": 2617808,
+      "tokens/train_per_sec_per_gpu": 32.23,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.6490103602409363,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.02030857279896736,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.02052,
+      "step": 83,
+      "tokens/total": 2649664,
+      "tokens/train_per_sec_per_gpu": 25.8,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.2439805567264557,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.008518049493432045,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00855,
+      "step": 84,
+      "tokens/total": 2681728,
+      "tokens/train_per_sec_per_gpu": 28.93,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.6747375726699829,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.029895059764385223,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.03035,
+      "step": 85,
+      "tokens/total": 2713488,
+      "tokens/train_per_sec_per_gpu": 28.49,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.582666277885437,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.014998597092926502,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01511,
+      "step": 86,
+      "tokens/total": 2745536,
+      "tokens/train_per_sec_per_gpu": 30.27,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.5364654660224915,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.011567167937755585,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.01163,
+      "step": 87,
+      "tokens/total": 2777680,
+      "tokens/train_per_sec_per_gpu": 29.73,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.34024426341056824,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.01014852337539196,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.35,
+      "memory/max_allocated (GiB)": 36.35,
+      "ppl": 1.0102,
+      "step": 88,
+      "tokens/total": 2809568,
+      "tokens/train_per_sec_per_gpu": 24.68,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.32472267746925354,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01229981891810894,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01238,
+      "step": 89,
+      "tokens/total": 2841760,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.7906899452209473,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.02552586793899536,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.05,
+      "memory/max_allocated (GiB)": 36.05,
+      "ppl": 1.02585,
+      "step": 90,
+      "tokens/total": 2871696,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7444764971733093,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0054441955871880054,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00546,
+      "step": 91,
+      "tokens/total": 2903968,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.1911851465702057,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.004665314219892025,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.00468,
+      "step": 92,
+      "tokens/total": 2936048,
+      "tokens/train_per_sec_per_gpu": 30.59,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.21238519251346588,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.003917238209396601,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00392,
+      "step": 93,
+      "tokens/total": 2968176,
+      "tokens/train_per_sec_per_gpu": 28.07,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.41729435324668884,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.013555807992815971,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01365,
+      "step": 94,
+      "tokens/total": 3000304,
+      "tokens/train_per_sec_per_gpu": 28.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.7246769070625305,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.020221399143338203,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.02043,
+      "step": 95,
+      "tokens/total": 3032080,
+      "tokens/train_per_sec_per_gpu": 25.35,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.16139821708202362,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003853276837617159,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00386,
+      "step": 96,
+      "tokens/total": 3064320,
+      "tokens/train_per_sec_per_gpu": 28.47,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 1.0509370565414429,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.015099998563528061,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.01521,
+      "step": 97,
+      "tokens/total": 3096144,
+      "tokens/train_per_sec_per_gpu": 26.61,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6859350204467773,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.004579808097332716,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.28,
+      "memory/max_allocated (GiB)": 36.28,
+      "ppl": 1.00459,
+      "step": 98,
+      "tokens/total": 3127600,
+      "tokens/train_per_sec_per_gpu": 25.91,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5686795711517334,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.022520575672388077,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02278,
+      "step": 99,
+      "tokens/total": 3159520,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 1.6308369636535645,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011883002705872059,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01195,
+      "step": 100,
+      "tokens/total": 3191568,
+      "tokens/train_per_sec_per_gpu": 29.75,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.7186088562011719,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0060632615350186825,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00608,
+      "step": 101,
+      "tokens/total": 3223648,
+      "tokens/train_per_sec_per_gpu": 27.14,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.12036337703466415,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0019118499476462603,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.00191,
+      "step": 102,
+      "tokens/total": 3255744,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.29928353428840637,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00944911316037178,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00949,
+      "step": 103,
+      "tokens/total": 3287792,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3766293227672577,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.008984047919511795,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.68,
+      "memory/max_allocated (GiB)": 36.68,
+      "ppl": 1.00902,
+      "step": 104,
+      "tokens/total": 3320016,
+      "tokens/train_per_sec_per_gpu": 28.92,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.3481820523738861,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.005444999784231186,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00546,
+      "step": 105,
+      "tokens/total": 3351952,
+      "tokens/train_per_sec_per_gpu": 28.32,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.8414323329925537,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.032438624650239944,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03297,
+      "step": 106,
+      "tokens/total": 3384032,
+      "tokens/train_per_sec_per_gpu": 27.11,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.5067655444145203,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.008442888967692852,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.67,
+      "memory/max_allocated (GiB)": 36.67,
+      "ppl": 1.00848,
+      "step": 107,
+      "tokens/total": 3416192,
+      "tokens/train_per_sec_per_gpu": 27.61,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.41617852449417114,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.008386321365833282,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00842,
+      "step": 108,
+      "tokens/total": 3448080,
+      "tokens/train_per_sec_per_gpu": 26.56,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.1587515026330948,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0031663556583225727,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.03,
+      "memory/max_allocated (GiB)": 36.03,
+      "ppl": 1.00317,
+      "step": 109,
+      "tokens/total": 3477952,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.2903148829936981,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.007958847098052502,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00799,
+      "step": 110,
+      "tokens/total": 3509776,
+      "tokens/train_per_sec_per_gpu": 28.42,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.642983078956604,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.006748045329004526,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00677,
+      "step": 111,
+      "tokens/total": 3541808,
+      "tokens/train_per_sec_per_gpu": 28.04,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.14928282797336578,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.0027839094400405884,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00279,
+      "step": 112,
+      "tokens/total": 3574064,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.05922335386276245,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009622900979593396,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00096,
+      "step": 113,
+      "tokens/total": 3605984,
+      "tokens/train_per_sec_per_gpu": 27.99,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.44227731227874756,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.002763954224064946,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00277,
+      "step": 114,
+      "tokens/total": 3638080,
+      "tokens/train_per_sec_per_gpu": 27.03,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5176818370819092,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.015623578801751137,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01575,
+      "step": 115,
+      "tokens/total": 3670016,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.4265497922897339,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.00794076919555664,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00797,
+      "step": 116,
+      "tokens/total": 3701984,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.40457987785339355,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022708112373948097,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02297,
+      "step": 117,
+      "tokens/total": 3733920,
+      "tokens/train_per_sec_per_gpu": 29.46,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.01414525043219328,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00020209309877827764,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.34,
+      "memory/max_allocated (GiB)": 36.34,
+      "ppl": 1.0002,
+      "step": 118,
+      "tokens/total": 3765840,
+      "tokens/train_per_sec_per_gpu": 28.57,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.11359831690788269,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0012168455868959427,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3797680,
+      "tokens/train_per_sec_per_gpu": 29.49,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.35982534289360046,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.004110721405595541,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.00412,
+      "step": 120,
+      "tokens/total": 3829920,
+      "tokens/train_per_sec_per_gpu": 29.2,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 1.3690382242202759,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.02705790475010872,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02743,
+      "step": 121,
+      "tokens/total": 3862080,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.351891428232193,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.0020163573790341616,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.00202,
+      "step": 122,
+      "tokens/total": 3893968,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4747961461544037,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.012588979676365852,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.01267,
+      "step": 123,
+      "tokens/total": 3926160,
+      "tokens/train_per_sec_per_gpu": 30.14,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.4064250588417053,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0022941348142921925,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0023,
+      "step": 124,
+      "tokens/total": 3958208,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.22056929767131805,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0022175831254571676,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00222,
+      "step": 125,
+      "tokens/total": 3990208,
+      "tokens/train_per_sec_per_gpu": 32.57,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5698862671852112,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010205100290477276,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01026,
+      "step": 126,
+      "tokens/total": 4022224,
+      "tokens/train_per_sec_per_gpu": 30.4,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2593887448310852,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0032277877908200026,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.00323,
+      "step": 127,
+      "tokens/total": 4053952,
+      "tokens/train_per_sec_per_gpu": 26.51,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.3735230565071106,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01031105499714613,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01036,
+      "step": 128,
+      "tokens/total": 4086144,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.1332484930753708,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.0012390739284455776,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00124,
+      "step": 129,
+      "tokens/total": 4118096,
+      "tokens/train_per_sec_per_gpu": 29.31,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.20182718336582184,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.004501686431467533,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.00451,
+      "step": 130,
+      "tokens/total": 4150272,
+      "tokens/train_per_sec_per_gpu": 26.49,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.529080331325531,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.00763288140296936,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.33,
+      "memory/max_allocated (GiB)": 36.33,
+      "ppl": 1.00766,
+      "step": 131,
+      "tokens/total": 4181904,
+      "tokens/train_per_sec_per_gpu": 26.84,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.21625548601150513,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0024378676898777485,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00244,
+      "step": 132,
+      "tokens/total": 4214032,
+      "tokens/train_per_sec_per_gpu": 28.36,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.37116539478302,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0037849934305995703,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00379,
+      "step": 133,
+      "tokens/total": 4246064,
+      "tokens/train_per_sec_per_gpu": 30.18,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.5646623373031616,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.006517104804515839,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00654,
+      "step": 134,
+      "tokens/total": 4278224,
+      "tokens/train_per_sec_per_gpu": 28.2,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.6371963620185852,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.008687382563948631,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00873,
+      "step": 135,
+      "tokens/total": 4310112,
+      "tokens/train_per_sec_per_gpu": 31.48,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.8210524916648865,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.008436474949121475,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00847,
+      "step": 136,
+      "tokens/total": 4342192,
+      "tokens/train_per_sec_per_gpu": 26.23,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.5692176222801208,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.017321188002824783,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.01747,
+      "step": 137,
+      "tokens/total": 4374096,
+      "tokens/train_per_sec_per_gpu": 26.38,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.12828375399112701,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0007754197577014565,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00078,
+      "step": 138,
+      "tokens/total": 4406272,
+      "tokens/train_per_sec_per_gpu": 24.22,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.3188190162181854,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.006707335356622934,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00673,
+      "step": 139,
+      "tokens/total": 4438112,
+      "tokens/train_per_sec_per_gpu": 28.55,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.6216707229614258,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.007024695165455341,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00705,
+      "step": 140,
+      "tokens/total": 4470128,
+      "tokens/train_per_sec_per_gpu": 30.64,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.1480305939912796,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.00229115248657763,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.00229,
+      "step": 141,
+      "tokens/total": 4502096,
+      "tokens/train_per_sec_per_gpu": 28.71,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.22254763543605804,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0029198499396443367,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.67,
+      "memory/max_allocated (GiB)": 36.67,
+      "ppl": 1.00292,
+      "step": 142,
+      "tokens/total": 4534496,
+      "tokens/train_per_sec_per_gpu": 29.08,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.6901941895484924,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.013051149435341358,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01314,
+      "step": 143,
+      "tokens/total": 4566720,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.9545458555221558,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.02235381305217743,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.02261,
+      "step": 144,
+      "tokens/total": 4598880,
+      "tokens/train_per_sec_per_gpu": 29.59,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.5515892505645752,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.02038375660777092,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.02059,
+      "step": 145,
+      "tokens/total": 4630864,
+      "tokens/train_per_sec_per_gpu": 26.3,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.3512430191040039,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.008233271539211273,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00827,
+      "step": 146,
+      "tokens/total": 4662864,
+      "tokens/train_per_sec_per_gpu": 26.23,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.6830361485481262,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.013419180177152157,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01351,
+      "step": 147,
+      "tokens/total": 4694912,
+      "tokens/train_per_sec_per_gpu": 26.34,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.2117031067609787,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.004134457092732191,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00414,
+      "step": 148,
+      "tokens/total": 4727120,
+      "tokens/train_per_sec_per_gpu": 31.07,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.378053218126297,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.009747720323503017,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.0098,
+      "step": 149,
+      "tokens/total": 4759168,
+      "tokens/train_per_sec_per_gpu": 27.25,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.21076741814613342,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.005035653710365295,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.00505,
+      "step": 150,
+      "tokens/total": 4791008,
+      "tokens/train_per_sec_per_gpu": 29.48,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.25075697898864746,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.006027051247656345,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.00605,
+      "step": 151,
+      "tokens/total": 4822912,
+      "tokens/train_per_sec_per_gpu": 27.95,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.5774204730987549,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.018617186695337296,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.01879,
+      "step": 152,
+      "tokens/total": 4854816,
+      "tokens/train_per_sec_per_gpu": 26.38,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.1364293098449707,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.002049289643764496,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00205,
+      "step": 153,
+      "tokens/total": 4886656,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.25387004017829895,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0044304742477834225,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00444,
+      "step": 154,
+      "tokens/total": 4918768,
+      "tokens/train_per_sec_per_gpu": 23.26,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3047125041484833,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0043389927595853806,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.00435,
+      "step": 155,
+      "tokens/total": 4951008,
+      "tokens/train_per_sec_per_gpu": 27.53,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.1442601978778839,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.0031973645091056824,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.0032,
+      "step": 156,
+      "tokens/total": 4982976,
+      "tokens/train_per_sec_per_gpu": 28.86,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.2608455717563629,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0032951058819890022,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.0033,
+      "step": 157,
+      "tokens/total": 5014640,
+      "tokens/train_per_sec_per_gpu": 25.66,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.637295663356781,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.01436957623809576,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.35,
+      "memory/max_allocated (GiB)": 36.35,
+      "ppl": 1.01447,
+      "step": 158,
+      "tokens/total": 5046672,
+      "tokens/train_per_sec_per_gpu": 27.58,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.9834056496620178,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.02375311404466629,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.02404,
+      "step": 159,
+      "tokens/total": 5078528,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.35432615876197815,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.006119017023593187,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00614,
+      "step": 160,
+      "tokens/total": 5110704,
+      "tokens/train_per_sec_per_gpu": 27.12,
+      "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": 3.468933461258358e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-160/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634
+size 8273

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/deconf_wave_deconf_control/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/deconf_wave_deconf_control/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

deconfound_sdf_v1/aft_control/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/deconf_wave_deconf_control/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0549effc2939bd203244d8227b20fe1e0f34a3555fde11b86136c07eb06e5738
+size 547777976

deconfound_sdf_v1/aft_control/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 -%}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b484083c9b1fa3b20d611b072130a948b97aae1b5d9c02d6fb9041ea03337b0
+size 1048106435

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5972beeed351537138e229200527c4e9f2e9808b01529e746b926011cc01d564
+size 14645

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/scheduler.pt

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

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/tokenizer.json

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

deconfound_sdf_v1/aft_control/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
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/tokens_state.json

@@ -0,0 +1 @@
+{"total": 6127488, "trainable": 87747}

deconfound_sdf_v1/aft_control/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": 1.2140090465545654,
+      "learning_rate": 0.0,
+      "loss": 0.16441112756729126,
+      "memory/device_reserved (GiB)": 38.08,
+      "memory/max_active (GiB)": 35.64,
+      "memory/max_allocated (GiB)": 35.64,
+      "ppl": 1.1787,
+      "step": 1,
+      "tokens/total": 32160,
+      "tokens/train_per_sec_per_gpu": 21.28,
+      "tokens/trainable": 470
+    },
+    {
+      "epoch": 0.0078125,
+      "grad_norm": 2.92620849609375,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.15774038434028625,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.17086,
+      "step": 2,
+      "tokens/total": 64192,
+      "tokens/train_per_sec_per_gpu": 29.27,
+      "tokens/trainable": 922
+    },
+    {
+      "epoch": 0.01171875,
+      "grad_norm": 1.135961890220642,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.1664382815361023,
+      "memory/device_reserved (GiB)": 38.78,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.18109,
+      "step": 3,
+      "tokens/total": 96224,
+      "tokens/train_per_sec_per_gpu": 28.54,
+      "tokens/trainable": 1396
+    },
+    {
+      "epoch": 0.015625,
+      "grad_norm": 0.880176305770874,
+      "learning_rate": 1.2e-05,
+      "loss": 0.15334013104438782,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.16572,
+      "step": 4,
+      "tokens/total": 128336,
+      "tokens/train_per_sec_per_gpu": 31.73,
+      "tokens/trainable": 1850
+    },
+    {
+      "epoch": 0.01953125,
+      "grad_norm": 1.1433289051055908,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.16352537274360657,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.17766,
+      "step": 5,
+      "tokens/total": 160784,
+      "tokens/train_per_sec_per_gpu": 29.45,
+      "tokens/trainable": 2302
+    },
+    {
+      "epoch": 0.0234375,
+      "grad_norm": 0.9728212952613831,
+      "learning_rate": 2e-05,
+      "loss": 0.15144473314285278,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.16351,
+      "step": 6,
+      "tokens/total": 192624,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 2742
+    },
+    {
+      "epoch": 0.02734375,
+      "grad_norm": 1.7796363830566406,
+      "learning_rate": 2.4e-05,
+      "loss": 0.13553467392921448,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.14515,
+      "step": 7,
+      "tokens/total": 224608,
+      "tokens/train_per_sec_per_gpu": 30.52,
+      "tokens/trainable": 3208
+    },
+    {
+      "epoch": 0.03125,
+      "grad_norm": 1.2444697618484497,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.1077304258942604,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.11375,
+      "step": 8,
+      "tokens/total": 256592,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 3655
+    },
+    {
+      "epoch": 0.03515625,
+      "grad_norm": 2.7339375019073486,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.08772765100002289,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.09169,
+      "step": 9,
+      "tokens/total": 286400,
+      "tokens/train_per_sec_per_gpu": 29.82,
+      "tokens/trainable": 4091
+    },
+    {
+      "epoch": 0.0390625,
+      "grad_norm": 4.109770774841309,
+      "learning_rate": 3.6e-05,
+      "loss": 0.130940243601799,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.1399,
+      "step": 10,
+      "tokens/total": 318352,
+      "tokens/train_per_sec_per_gpu": 29.13,
+      "tokens/trainable": 4553
+    },
+    {
+      "epoch": 0.04296875,
+      "grad_norm": 2.3768608570098877,
+      "learning_rate": 4e-05,
+      "loss": 0.10059958696365356,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.10583,
+      "step": 11,
+      "tokens/total": 350336,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 4992
+    },
+    {
+      "epoch": 0.046875,
+      "grad_norm": 3.281402826309204,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.12574875354766846,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.134,
+      "step": 12,
+      "tokens/total": 382448,
+      "tokens/train_per_sec_per_gpu": 31.9,
+      "tokens/trainable": 5494
+    },
+    {
+      "epoch": 0.05078125,
+      "grad_norm": 1.7738980054855347,
+      "learning_rate": 4.8e-05,
+      "loss": 0.033296722918748856,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03386,
+      "step": 13,
+      "tokens/total": 414352,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 5933
+    },
+    {
+      "epoch": 0.0546875,
+      "grad_norm": 2.64290189743042,
+      "learning_rate": 5.2000000000000004e-05,
+      "loss": 0.06366822123527527,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.06574,
+      "step": 14,
+      "tokens/total": 446400,
+      "tokens/train_per_sec_per_gpu": 24.8,
+      "tokens/trainable": 6369
+    },
+    {
+      "epoch": 0.05859375,
+      "grad_norm": 3.3691928386688232,
+      "learning_rate": 5.6000000000000006e-05,
+      "loss": 0.08168518543243408,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.08511,
+      "step": 15,
+      "tokens/total": 478688,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 6820
+    },
+    {
+      "epoch": 0.0625,
+      "grad_norm": 2.4356393814086914,
+      "learning_rate": 6e-05,
+      "loss": 0.09031196683645248,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.09452,
+      "step": 16,
+      "tokens/total": 510512,
+      "tokens/train_per_sec_per_gpu": 25.7,
+      "tokens/trainable": 7258
+    },
+    {
+      "epoch": 0.06640625,
+      "grad_norm": 1.55682373046875,
+      "learning_rate": 6.400000000000001e-05,
+      "loss": 0.07496573030948639,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.07785,
+      "step": 17,
+      "tokens/total": 542736,
+      "tokens/train_per_sec_per_gpu": 29.54,
+      "tokens/trainable": 7710
+    },
+    {
+      "epoch": 0.0703125,
+      "grad_norm": 1.1054646968841553,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 0.05218805745244026,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.05357,
+      "step": 18,
+      "tokens/total": 574832,
+      "tokens/train_per_sec_per_gpu": 25.78,
+      "tokens/trainable": 8174
+    },
+    {
+      "epoch": 0.07421875,
+      "grad_norm": 0.7429099082946777,
+      "learning_rate": 7.2e-05,
+      "loss": 0.03790827840566635,
+      "memory/device_reserved (GiB)": 38.8,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.03864,
+      "step": 19,
+      "tokens/total": 606976,
+      "tokens/train_per_sec_per_gpu": 28.62,
+      "tokens/trainable": 8617
+    },
+    {
+      "epoch": 0.078125,
+      "grad_norm": 2.9508209228515625,
+      "learning_rate": 7.6e-05,
+      "loss": 0.04993613809347153,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.0512,
+      "step": 20,
+      "tokens/total": 639200,
+      "tokens/train_per_sec_per_gpu": 24.91,
+      "tokens/trainable": 9037
+    },
+    {
+      "epoch": 0.08203125,
+      "grad_norm": 1.0297844409942627,
+      "learning_rate": 8e-05,
+      "loss": 0.017908263951539993,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.01807,
+      "step": 21,
+      "tokens/total": 671072,
+      "tokens/train_per_sec_per_gpu": 30.42,
+      "tokens/trainable": 9503
+    },
+    {
+      "epoch": 0.0859375,
+      "grad_norm": 0.9279879331588745,
+      "learning_rate": 8.4e-05,
+      "loss": 0.013478193432092667,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01357,
+      "step": 22,
+      "tokens/total": 703136,
+      "tokens/train_per_sec_per_gpu": 30.74,
+      "tokens/trainable": 9972
+    },
+    {
+      "epoch": 0.08984375,
+      "grad_norm": 1.526488184928894,
+      "learning_rate": 8.800000000000001e-05,
+      "loss": 0.017310388386249542,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01746,
+      "step": 23,
+      "tokens/total": 735120,
+      "tokens/train_per_sec_per_gpu": 31.45,
+      "tokens/trainable": 10447
+    },
+    {
+      "epoch": 0.09375,
+      "grad_norm": 3.4049417972564697,
+      "learning_rate": 9.200000000000001e-05,
+      "loss": 0.0353429913520813,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.03597,
+      "step": 24,
+      "tokens/total": 767040,
+      "tokens/train_per_sec_per_gpu": 30.79,
+      "tokens/trainable": 10899
+    },
+    {
+      "epoch": 0.09765625,
+      "grad_norm": 7.511656761169434,
+      "learning_rate": 9.6e-05,
+      "loss": 0.1335875540971756,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.14292,
+      "step": 25,
+      "tokens/total": 798976,
+      "tokens/train_per_sec_per_gpu": 27.72,
+      "tokens/trainable": 11360
+    },
+    {
+      "epoch": 0.1015625,
+      "grad_norm": 1.0069234371185303,
+      "learning_rate": 0.0001,
+      "loss": 0.04382415860891342,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0448,
+      "step": 26,
+      "tokens/total": 830880,
+      "tokens/train_per_sec_per_gpu": 24.61,
+      "tokens/trainable": 11775
+    },
+    {
+      "epoch": 0.10546875,
+      "grad_norm": 7.807938575744629,
+      "learning_rate": 9.99990636831587e-05,
+      "loss": 0.01210833340883255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.11,
+      "memory/max_allocated (GiB)": 36.11,
+      "ppl": 1.01218,
+      "step": 27,
+      "tokens/total": 860960,
+      "tokens/train_per_sec_per_gpu": 33.01,
+      "tokens/trainable": 12242
+    },
+    {
+      "epoch": 0.109375,
+      "grad_norm": 2.23823881149292,
+      "learning_rate": 9.999625477159879e-05,
+      "loss": 0.0272970050573349,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02767,
+      "step": 28,
+      "tokens/total": 892816,
+      "tokens/train_per_sec_per_gpu": 27.87,
+      "tokens/trainable": 12696
+    },
+    {
+      "epoch": 0.11328125,
+      "grad_norm": 1.315584421157837,
+      "learning_rate": 9.999157338221051e-05,
+      "loss": 0.025516577064990997,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.02584,
+      "step": 29,
+      "tokens/total": 924736,
+      "tokens/train_per_sec_per_gpu": 28.71,
+      "tokens/trainable": 13153
+    },
+    {
+      "epoch": 0.1171875,
+      "grad_norm": 0.8413651585578918,
+      "learning_rate": 9.998501970980562e-05,
+      "loss": 0.007838520221412182,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00787,
+      "step": 30,
+      "tokens/total": 954592,
+      "tokens/train_per_sec_per_gpu": 33.59,
+      "tokens/trainable": 13624
+    },
+    {
+      "epoch": 0.12109375,
+      "grad_norm": 1.6797157526016235,
+      "learning_rate": 9.997659402710915e-05,
+      "loss": 0.03002442792057991,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.03048,
+      "step": 31,
+      "tokens/total": 986592,
+      "tokens/train_per_sec_per_gpu": 28.67,
+      "tokens/trainable": 14064
+    },
+    {
+      "epoch": 0.125,
+      "grad_norm": 1.3320434093475342,
+      "learning_rate": 9.996629668474818e-05,
+      "loss": 0.03772728145122528,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.03845,
+      "step": 32,
+      "tokens/total": 1018720,
+      "tokens/train_per_sec_per_gpu": 32.5,
+      "tokens/trainable": 14565
+    },
+    {
+      "epoch": 0.12890625,
+      "grad_norm": 0.6335167288780212,
+      "learning_rate": 9.995412811123711e-05,
+      "loss": 0.013970119878649712,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.01407,
+      "step": 33,
+      "tokens/total": 1050624,
+      "tokens/train_per_sec_per_gpu": 28.18,
+      "tokens/trainable": 15005
+    },
+    {
+      "epoch": 0.1328125,
+      "grad_norm": 0.5626921653747559,
+      "learning_rate": 9.994008881295999e-05,
+      "loss": 0.016654539853334427,
+      "memory/device_reserved (GiB)": 37.74,
+      "memory/max_active (GiB)": 36.22,
+      "memory/max_allocated (GiB)": 36.22,
+      "ppl": 1.01679,
+      "step": 34,
+      "tokens/total": 1082176,
+      "tokens/train_per_sec_per_gpu": 26.84,
+      "tokens/trainable": 15459
+    },
+    {
+      "epoch": 0.13671875,
+      "grad_norm": 0.603219211101532,
+      "learning_rate": 9.992417937414932e-05,
+      "loss": 0.021408583968877792,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02164,
+      "step": 35,
+      "tokens/total": 1113968,
+      "tokens/train_per_sec_per_gpu": 31.75,
+      "tokens/trainable": 15960
+    },
+    {
+      "epoch": 0.140625,
+      "grad_norm": 0.9779674410820007,
+      "learning_rate": 9.99064004568618e-05,
+      "loss": 0.03277068957686424,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03331,
+      "step": 36,
+      "tokens/total": 1145952,
+      "tokens/train_per_sec_per_gpu": 28.68,
+      "tokens/trainable": 16428
+    },
+    {
+      "epoch": 0.14453125,
+      "grad_norm": 0.8855255842208862,
+      "learning_rate": 9.988675280095074e-05,
+      "loss": 0.017048098146915436,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.01719,
+      "step": 37,
+      "tokens/total": 1177872,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 16884
+    },
+    {
+      "epoch": 0.1484375,
+      "grad_norm": 1.9266122579574585,
+      "learning_rate": 9.986523722403528e-05,
+      "loss": 0.045863546431064606,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.04693,
+      "step": 38,
+      "tokens/total": 1209776,
+      "tokens/train_per_sec_per_gpu": 26.91,
+      "tokens/trainable": 17341
+    },
+    {
+      "epoch": 0.15234375,
+      "grad_norm": 0.8280895948410034,
+      "learning_rate": 9.984185462146642e-05,
+      "loss": 0.033055759966373444,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03361,
+      "step": 39,
+      "tokens/total": 1241888,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 17786
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 5.029730319976807,
+      "learning_rate": 9.98166059662897e-05,
+      "loss": 0.025589998811483383,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.02592,
+      "step": 40,
+      "tokens/total": 1274032,
+      "tokens/train_per_sec_per_gpu": 29.22,
+      "tokens/trainable": 18262
+    },
+    {
+      "epoch": 0.16015625,
+      "grad_norm": 0.5688032507896423,
+      "learning_rate": 9.978949230920472e-05,
+      "loss": 0.016095872968435287,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01623,
+      "step": 41,
+      "tokens/total": 1306000,
+      "tokens/train_per_sec_per_gpu": 27.49,
+      "tokens/trainable": 18716
+    },
+    {
+      "epoch": 0.1640625,
+      "grad_norm": 0.8217663168907166,
+      "learning_rate": 9.976051477852141e-05,
+      "loss": 0.026798982173204422,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02716,
+      "step": 42,
+      "tokens/total": 1337936,
+      "tokens/train_per_sec_per_gpu": 27.51,
+      "tokens/trainable": 19157
+    },
+    {
+      "epoch": 0.16796875,
+      "grad_norm": 0.7309132218360901,
+      "learning_rate": 9.972967458011312e-05,
+      "loss": 0.033920761197805405,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0345,
+      "step": 43,
+      "tokens/total": 1369904,
+      "tokens/train_per_sec_per_gpu": 30.66,
+      "tokens/trainable": 19647
+    },
+    {
+      "epoch": 0.171875,
+      "grad_norm": 0.8134022355079651,
+      "learning_rate": 9.96969729973664e-05,
+      "loss": 0.024005085229873657,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.0243,
+      "step": 44,
+      "tokens/total": 1401984,
+      "tokens/train_per_sec_per_gpu": 29.06,
+      "tokens/trainable": 20119
+    },
+    {
+      "epoch": 0.17578125,
+      "grad_norm": 0.35084816813468933,
+      "learning_rate": 9.966241139112754e-05,
+      "loss": 0.013319254852831364,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01341,
+      "step": 45,
+      "tokens/total": 1434032,
+      "tokens/train_per_sec_per_gpu": 27.45,
+      "tokens/trainable": 20560
+    },
+    {
+      "epoch": 0.1796875,
+      "grad_norm": 0.8701573014259338,
+      "learning_rate": 9.96259911996461e-05,
+      "loss": 0.01356641948223114,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01366,
+      "step": 46,
+      "tokens/total": 1465872,
+      "tokens/train_per_sec_per_gpu": 27.35,
+      "tokens/trainable": 20992
+    },
+    {
+      "epoch": 0.18359375,
+      "grad_norm": 1.4406442642211914,
+      "learning_rate": 9.958771393851491e-05,
+      "loss": 0.04396980255842209,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 35.86,
+      "memory/max_allocated (GiB)": 35.86,
+      "ppl": 1.04495,
+      "step": 47,
+      "tokens/total": 1495440,
+      "tokens/train_per_sec_per_gpu": 26.46,
+      "tokens/trainable": 21441
+    },
+    {
+      "epoch": 0.1875,
+      "grad_norm": 0.4966669976711273,
+      "learning_rate": 9.954758120060702e-05,
+      "loss": 0.011179964058101177,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.01124,
+      "step": 48,
+      "tokens/total": 1527776,
+      "tokens/train_per_sec_per_gpu": 27.89,
+      "tokens/trainable": 21901
+    },
+    {
+      "epoch": 0.19140625,
+      "grad_norm": 0.41124436259269714,
+      "learning_rate": 9.950559465600948e-05,
+      "loss": 0.0063599650748074055,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00638,
+      "step": 49,
+      "tokens/total": 1559760,
+      "tokens/train_per_sec_per_gpu": 27.73,
+      "tokens/trainable": 22341
+    },
+    {
+      "epoch": 0.1953125,
+      "grad_norm": 0.829739511013031,
+      "learning_rate": 9.946175605195379e-05,
+      "loss": 0.015785304829478264,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01591,
+      "step": 50,
+      "tokens/total": 1592000,
+      "tokens/train_per_sec_per_gpu": 27.39,
+      "tokens/trainable": 22833
+    },
+    {
+      "epoch": 0.19921875,
+      "grad_norm": 0.5181736946105957,
+      "learning_rate": 9.941606721274322e-05,
+      "loss": 0.008393109776079655,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00843,
+      "step": 51,
+      "tokens/total": 1623920,
+      "tokens/train_per_sec_per_gpu": 25.61,
+      "tokens/trainable": 23277
+    },
+    {
+      "epoch": 0.203125,
+      "grad_norm": 1.5778990983963013,
+      "learning_rate": 9.936853003967685e-05,
+      "loss": 0.024984844028949738,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0253,
+      "step": 52,
+      "tokens/total": 1656032,
+      "tokens/train_per_sec_per_gpu": 29.64,
+      "tokens/trainable": 23759
+    },
+    {
+      "epoch": 0.20703125,
+      "grad_norm": 1.4938714504241943,
+      "learning_rate": 9.93191465109705e-05,
+      "loss": 0.038523320108652115,
+      "memory/device_reserved (GiB)": 38.99,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03927,
+      "step": 53,
+      "tokens/total": 1688368,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 24193
+    },
+    {
+      "epoch": 0.2109375,
+      "grad_norm": 2.1427106857299805,
+      "learning_rate": 9.926791868167438e-05,
+      "loss": 0.06302924454212189,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.06506,
+      "step": 54,
+      "tokens/total": 1720560,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 24619
+    },
+    {
+      "epoch": 0.21484375,
+      "grad_norm": 1.8365188837051392,
+      "learning_rate": 9.921484868358753e-05,
+      "loss": 0.052160464227199554,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.05354,
+      "step": 55,
+      "tokens/total": 1752464,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 25075
+    },
+    {
+      "epoch": 0.21875,
+      "grad_norm": 0.7779749631881714,
+      "learning_rate": 9.915993872516924e-05,
+      "loss": 0.021328264847397804,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.38,
+      "memory/max_allocated (GiB)": 36.38,
+      "ppl": 1.02156,
+      "step": 56,
+      "tokens/total": 1784336,
+      "tokens/train_per_sec_per_gpu": 26.42,
+      "tokens/trainable": 25519
+    },
+    {
+      "epoch": 0.22265625,
+      "grad_norm": 0.4189571142196655,
+      "learning_rate": 9.9103191091447e-05,
+      "loss": 0.014279939234256744,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01438,
+      "step": 57,
+      "tokens/total": 1816304,
+      "tokens/train_per_sec_per_gpu": 27.47,
+      "tokens/trainable": 25966
+    },
+    {
+      "epoch": 0.2265625,
+      "grad_norm": 0.3834710717201233,
+      "learning_rate": 9.904460814392147e-05,
+      "loss": 0.01883944310247898,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01902,
+      "step": 58,
+      "tokens/total": 1848304,
+      "tokens/train_per_sec_per_gpu": 28.08,
+      "tokens/trainable": 26423
+    },
+    {
+      "epoch": 0.23046875,
+      "grad_norm": 0.7693278789520264,
+      "learning_rate": 9.898419232046825e-05,
+      "loss": 0.03533369302749634,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.03597,
+      "step": 59,
+      "tokens/total": 1880640,
+      "tokens/train_per_sec_per_gpu": 31.99,
+      "tokens/trainable": 26934
+    },
+    {
+      "epoch": 0.234375,
+      "grad_norm": 0.4606446325778961,
+      "learning_rate": 9.892194613523633e-05,
+      "loss": 0.021482212468981743,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.02171,
+      "step": 60,
+      "tokens/total": 1912592,
+      "tokens/train_per_sec_per_gpu": 27.64,
+      "tokens/trainable": 27393
+    },
+    {
+      "epoch": 0.23828125,
+      "grad_norm": 0.46030017733573914,
+      "learning_rate": 9.885787217854357e-05,
+      "loss": 0.017044665291905403,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01719,
+      "step": 61,
+      "tokens/total": 1944544,
+      "tokens/train_per_sec_per_gpu": 27.44,
+      "tokens/trainable": 27852
+    },
+    {
+      "epoch": 0.2421875,
+      "grad_norm": 0.7918264269828796,
+      "learning_rate": 9.879197311676887e-05,
+      "loss": 0.016954131424427032,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.0171,
+      "step": 62,
+      "tokens/total": 1976688,
+      "tokens/train_per_sec_per_gpu": 27.32,
+      "tokens/trainable": 28292
+    },
+    {
+      "epoch": 0.24609375,
+      "grad_norm": 0.4787926971912384,
+      "learning_rate": 9.872425169224113e-05,
+      "loss": 0.01525965891778469,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01538,
+      "step": 63,
+      "tokens/total": 2008928,
+      "tokens/train_per_sec_per_gpu": 31.14,
+      "tokens/trainable": 28798
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.9033918976783752,
+      "learning_rate": 9.865471072312528e-05,
+      "loss": 0.02902211993932724,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02945,
+      "step": 64,
+      "tokens/total": 2041120,
+      "tokens/train_per_sec_per_gpu": 29.83,
+      "tokens/trainable": 29283
+    },
+    {
+      "epoch": 0.25390625,
+      "grad_norm": 0.39305707812309265,
+      "learning_rate": 9.858335310330492e-05,
+      "loss": 0.00901167094707489,
+      "memory/device_reserved (GiB)": 39.08,
+      "memory/max_active (GiB)": 36.32,
+      "memory/max_allocated (GiB)": 36.32,
+      "ppl": 1.00905,
+      "step": 65,
+      "tokens/total": 2073024,
+      "tokens/train_per_sec_per_gpu": 26.48,
+      "tokens/trainable": 29745
+    },
+    {
+      "epoch": 0.2578125,
+      "grad_norm": 0.4494345784187317,
+      "learning_rate": 9.851018180226185e-05,
+      "loss": 0.007426132448017597,
+      "memory/device_reserved (GiB)": 37.47,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00745,
+      "step": 66,
+      "tokens/total": 2105184,
+      "tokens/train_per_sec_per_gpu": 26.18,
+      "tokens/trainable": 30186
+    },
+    {
+      "epoch": 0.26171875,
+      "grad_norm": 0.629649817943573,
+      "learning_rate": 9.843519986495259e-05,
+      "loss": 0.010609081946313381,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01067,
+      "step": 67,
+      "tokens/total": 2137264,
+      "tokens/train_per_sec_per_gpu": 27.71,
+      "tokens/trainable": 30622
+    },
+    {
+      "epoch": 0.265625,
+      "grad_norm": 0.7591162919998169,
+      "learning_rate": 9.835841041168162e-05,
+      "loss": 0.00926615484058857,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.00931,
+      "step": 68,
+      "tokens/total": 2169424,
+      "tokens/train_per_sec_per_gpu": 29.42,
+      "tokens/trainable": 31097
+    },
+    {
+      "epoch": 0.26953125,
+      "grad_norm": 0.7288486361503601,
+      "learning_rate": 9.82798166379715e-05,
+      "loss": 0.024310583248734474,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.02461,
+      "step": 69,
+      "tokens/total": 2201424,
+      "tokens/train_per_sec_per_gpu": 30.54,
+      "tokens/trainable": 31595
+    },
+    {
+      "epoch": 0.2734375,
+      "grad_norm": 2.134056329727173,
+      "learning_rate": 9.819942181443002e-05,
+      "loss": 0.008422967046499252,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00846,
+      "step": 70,
+      "tokens/total": 2233424,
+      "tokens/train_per_sec_per_gpu": 25.81,
+      "tokens/trainable": 32036
+    },
+    {
+      "epoch": 0.27734375,
+      "grad_norm": 0.6362871527671814,
+      "learning_rate": 9.811722928661392e-05,
+      "loss": 0.01577262207865715,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0159,
+      "step": 71,
+      "tokens/total": 2265328,
+      "tokens/train_per_sec_per_gpu": 23.65,
+      "tokens/trainable": 32428
+    },
+    {
+      "epoch": 0.28125,
+      "grad_norm": 0.6091912388801575,
+      "learning_rate": 9.803324247488975e-05,
+      "loss": 0.0064800274558365345,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.0065,
+      "step": 72,
+      "tokens/total": 2297248,
+      "tokens/train_per_sec_per_gpu": 26.86,
+      "tokens/trainable": 32880
+    },
+    {
+      "epoch": 0.28515625,
+      "grad_norm": 0.14533838629722595,
+      "learning_rate": 9.794746487429161e-05,
+      "loss": 0.001991454279050231,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.41,
+      "memory/max_allocated (GiB)": 36.41,
+      "ppl": 1.00199,
+      "step": 73,
+      "tokens/total": 2329152,
+      "tokens/train_per_sec_per_gpu": 27.97,
+      "tokens/trainable": 33323
+    },
+    {
+      "epoch": 0.2890625,
+      "grad_norm": 0.5400826930999756,
+      "learning_rate": 9.785990005437554e-05,
+      "loss": 0.017662350088357925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.01782,
+      "step": 74,
+      "tokens/total": 2361344,
+      "tokens/train_per_sec_per_gpu": 29.61,
+      "tokens/trainable": 33792
+    },
+    {
+      "epoch": 0.29296875,
+      "grad_norm": 0.6132436394691467,
+      "learning_rate": 9.777055165907117e-05,
+      "loss": 0.01787603087723255,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01804,
+      "step": 75,
+      "tokens/total": 2393168,
+      "tokens/train_per_sec_per_gpu": 29.53,
+      "tokens/trainable": 34223
+    },
+    {
+      "epoch": 0.296875,
+      "grad_norm": 0.7287399172782898,
+      "learning_rate": 9.767942340652993e-05,
+      "loss": 0.024754449725151062,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.31,
+      "memory/max_allocated (GiB)": 36.31,
+      "ppl": 1.02506,
+      "step": 76,
+      "tokens/total": 2424880,
+      "tokens/train_per_sec_per_gpu": 24.92,
+      "tokens/trainable": 34649
+    },
+    {
+      "epoch": 0.30078125,
+      "grad_norm": 0.6790545582771301,
+      "learning_rate": 9.758651908897035e-05,
+      "loss": 0.014308430254459381,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01441,
+      "step": 77,
+      "tokens/total": 2457120,
+      "tokens/train_per_sec_per_gpu": 29.35,
+      "tokens/trainable": 35094
+    },
+    {
+      "epoch": 0.3046875,
+      "grad_norm": 0.6203456521034241,
+      "learning_rate": 9.749184257252033e-05,
+      "loss": 0.034353043884038925,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.03495,
+      "step": 78,
+      "tokens/total": 2489104,
+      "tokens/train_per_sec_per_gpu": 26.39,
+      "tokens/trainable": 35534
+    },
+    {
+      "epoch": 0.30859375,
+      "grad_norm": 0.6238825917243958,
+      "learning_rate": 9.739539779705614e-05,
+      "loss": 0.032295141369104385,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.03282,
+      "step": 79,
+      "tokens/total": 2521136,
+      "tokens/train_per_sec_per_gpu": 29.68,
+      "tokens/trainable": 36029
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 0.18977171182632446,
+      "learning_rate": 9.729718877603861e-05,
+      "loss": 0.007260813377797604,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.00729,
+      "step": 80,
+      "tokens/total": 2553168,
+      "tokens/train_per_sec_per_gpu": 29.48,
+      "tokens/trainable": 36469
+    },
+    {
+      "epoch": 0.31640625,
+      "grad_norm": 0.8054049611091614,
+      "learning_rate": 9.719721959634592e-05,
+      "loss": 0.034894898533821106,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.03551,
+      "step": 81,
+      "tokens/total": 2585584,
+      "tokens/train_per_sec_per_gpu": 25.75,
+      "tokens/trainable": 36906
+    },
+    {
+      "epoch": 0.3203125,
+      "grad_norm": 0.6884189248085022,
+      "learning_rate": 9.709549441810375e-05,
+      "loss": 0.018210047855973244,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.01838,
+      "step": 82,
+      "tokens/total": 2617808,
+      "tokens/train_per_sec_per_gpu": 32.23,
+      "tokens/trainable": 37390
+    },
+    {
+      "epoch": 0.32421875,
+      "grad_norm": 0.6490103602409363,
+      "learning_rate": 9.699201747451195e-05,
+      "loss": 0.02030857279896736,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.02052,
+      "step": 83,
+      "tokens/total": 2649664,
+      "tokens/train_per_sec_per_gpu": 25.8,
+      "tokens/trainable": 37838
+    },
+    {
+      "epoch": 0.328125,
+      "grad_norm": 0.2439805567264557,
+      "learning_rate": 9.688679307166854e-05,
+      "loss": 0.008518049493432045,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00855,
+      "step": 84,
+      "tokens/total": 2681728,
+      "tokens/train_per_sec_per_gpu": 28.93,
+      "tokens/trainable": 38310
+    },
+    {
+      "epoch": 0.33203125,
+      "grad_norm": 0.6747375726699829,
+      "learning_rate": 9.677982558839042e-05,
+      "loss": 0.029895059764385223,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.03035,
+      "step": 85,
+      "tokens/total": 2713488,
+      "tokens/train_per_sec_per_gpu": 28.49,
+      "tokens/trainable": 38789
+    },
+    {
+      "epoch": 0.3359375,
+      "grad_norm": 0.582666277885437,
+      "learning_rate": 9.66711194760312e-05,
+      "loss": 0.014998597092926502,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.49,
+      "memory/max_allocated (GiB)": 36.49,
+      "ppl": 1.01511,
+      "step": 86,
+      "tokens/total": 2745536,
+      "tokens/train_per_sec_per_gpu": 30.27,
+      "tokens/trainable": 39277
+    },
+    {
+      "epoch": 0.33984375,
+      "grad_norm": 0.5364654660224915,
+      "learning_rate": 9.656067925829593e-05,
+      "loss": 0.011567167937755585,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.01163,
+      "step": 87,
+      "tokens/total": 2777680,
+      "tokens/train_per_sec_per_gpu": 29.73,
+      "tokens/trainable": 39737
+    },
+    {
+      "epoch": 0.34375,
+      "grad_norm": 0.34024426341056824,
+      "learning_rate": 9.644850953105288e-05,
+      "loss": 0.01014852337539196,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.35,
+      "memory/max_allocated (GiB)": 36.35,
+      "ppl": 1.0102,
+      "step": 88,
+      "tokens/total": 2809568,
+      "tokens/train_per_sec_per_gpu": 24.68,
+      "tokens/trainable": 40179
+    },
+    {
+      "epoch": 0.34765625,
+      "grad_norm": 0.32472267746925354,
+      "learning_rate": 9.633461496214225e-05,
+      "loss": 0.01229981891810894,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01238,
+      "step": 89,
+      "tokens/total": 2841760,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 40649
+    },
+    {
+      "epoch": 0.3515625,
+      "grad_norm": 0.7906899452209473,
+      "learning_rate": 9.621900029118195e-05,
+      "loss": 0.02552586793899536,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.05,
+      "memory/max_allocated (GiB)": 36.05,
+      "ppl": 1.02585,
+      "step": 90,
+      "tokens/total": 2871696,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 41080
+    },
+    {
+      "epoch": 0.35546875,
+      "grad_norm": 0.7444764971733093,
+      "learning_rate": 9.610167032937036e-05,
+      "loss": 0.0054441955871880054,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00546,
+      "step": 91,
+      "tokens/total": 2903968,
+      "tokens/train_per_sec_per_gpu": 31.7,
+      "tokens/trainable": 41599
+    },
+    {
+      "epoch": 0.359375,
+      "grad_norm": 0.1911851465702057,
+      "learning_rate": 9.598262995928611e-05,
+      "loss": 0.004665314219892025,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.00468,
+      "step": 92,
+      "tokens/total": 2936048,
+      "tokens/train_per_sec_per_gpu": 30.59,
+      "tokens/trainable": 42076
+    },
+    {
+      "epoch": 0.36328125,
+      "grad_norm": 0.21238519251346588,
+      "learning_rate": 9.586188413468492e-05,
+      "loss": 0.003917238209396601,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00392,
+      "step": 93,
+      "tokens/total": 2968176,
+      "tokens/train_per_sec_per_gpu": 28.07,
+      "tokens/trainable": 42522
+    },
+    {
+      "epoch": 0.3671875,
+      "grad_norm": 0.41729435324668884,
+      "learning_rate": 9.57394378802934e-05,
+      "loss": 0.013555807992815971,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.01365,
+      "step": 94,
+      "tokens/total": 3000304,
+      "tokens/train_per_sec_per_gpu": 28.69,
+      "tokens/trainable": 42974
+    },
+    {
+      "epoch": 0.37109375,
+      "grad_norm": 0.7246769070625305,
+      "learning_rate": 9.56152962916e-05,
+      "loss": 0.020221399143338203,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.02043,
+      "step": 95,
+      "tokens/total": 3032080,
+      "tokens/train_per_sec_per_gpu": 25.35,
+      "tokens/trainable": 43380
+    },
+    {
+      "epoch": 0.375,
+      "grad_norm": 0.16139821708202362,
+      "learning_rate": 9.548946453464296e-05,
+      "loss": 0.003853276837617159,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00386,
+      "step": 96,
+      "tokens/total": 3064320,
+      "tokens/train_per_sec_per_gpu": 28.47,
+      "tokens/trainable": 43865
+    },
+    {
+      "epoch": 0.37890625,
+      "grad_norm": 1.0509370565414429,
+      "learning_rate": 9.53619478457953e-05,
+      "loss": 0.015099998563528061,
+      "memory/device_reserved (GiB)": 38.96,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.01521,
+      "step": 97,
+      "tokens/total": 3096144,
+      "tokens/train_per_sec_per_gpu": 26.61,
+      "tokens/trainable": 44328
+    },
+    {
+      "epoch": 0.3828125,
+      "grad_norm": 0.6859350204467773,
+      "learning_rate": 9.523275153154695e-05,
+      "loss": 0.004579808097332716,
+      "memory/device_reserved (GiB)": 38.68,
+      "memory/max_active (GiB)": 36.28,
+      "memory/max_allocated (GiB)": 36.28,
+      "ppl": 1.00459,
+      "step": 98,
+      "tokens/total": 3127600,
+      "tokens/train_per_sec_per_gpu": 25.91,
+      "tokens/trainable": 44778
+    },
+    {
+      "epoch": 0.38671875,
+      "grad_norm": 0.5686795711517334,
+      "learning_rate": 9.51018809682839e-05,
+      "loss": 0.022520575672388077,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.02278,
+      "step": 99,
+      "tokens/total": 3159520,
+      "tokens/train_per_sec_per_gpu": 30.85,
+      "tokens/trainable": 45225
+    },
+    {
+      "epoch": 0.390625,
+      "grad_norm": 1.6308369636535645,
+      "learning_rate": 9.49693416020645e-05,
+      "loss": 0.011883002705872059,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.01195,
+      "step": 100,
+      "tokens/total": 3191568,
+      "tokens/train_per_sec_per_gpu": 29.75,
+      "tokens/trainable": 45678
+    },
+    {
+      "epoch": 0.39453125,
+      "grad_norm": 0.7186088562011719,
+      "learning_rate": 9.483513894839276e-05,
+      "loss": 0.0060632615350186825,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00608,
+      "step": 101,
+      "tokens/total": 3223648,
+      "tokens/train_per_sec_per_gpu": 27.14,
+      "tokens/trainable": 46125
+    },
+    {
+      "epoch": 0.3984375,
+      "grad_norm": 0.12036337703466415,
+      "learning_rate": 9.469927859198888e-05,
+      "loss": 0.0019118499476462603,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.64,
+      "memory/max_allocated (GiB)": 36.64,
+      "ppl": 1.00191,
+      "step": 102,
+      "tokens/total": 3255744,
+      "tokens/train_per_sec_per_gpu": 33.33,
+      "tokens/trainable": 46612
+    },
+    {
+      "epoch": 0.40234375,
+      "grad_norm": 0.29928353428840637,
+      "learning_rate": 9.456176618655689e-05,
+      "loss": 0.00944911316037178,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00949,
+      "step": 103,
+      "tokens/total": 3287792,
+      "tokens/train_per_sec_per_gpu": 27.3,
+      "tokens/trainable": 47059
+    },
+    {
+      "epoch": 0.40625,
+      "grad_norm": 0.3766293227672577,
+      "learning_rate": 9.442260745454927e-05,
+      "loss": 0.008984047919511795,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.68,
+      "memory/max_allocated (GiB)": 36.68,
+      "ppl": 1.00902,
+      "step": 104,
+      "tokens/total": 3320016,
+      "tokens/train_per_sec_per_gpu": 28.92,
+      "tokens/trainable": 47536
+    },
+    {
+      "epoch": 0.41015625,
+      "grad_norm": 0.3481820523738861,
+      "learning_rate": 9.428180818692884e-05,
+      "loss": 0.005444999784231186,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.00546,
+      "step": 105,
+      "tokens/total": 3351952,
+      "tokens/train_per_sec_per_gpu": 28.32,
+      "tokens/trainable": 48037
+    },
+    {
+      "epoch": 0.4140625,
+      "grad_norm": 0.8414323329925537,
+      "learning_rate": 9.413937424292791e-05,
+      "loss": 0.032438624650239944,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.03297,
+      "step": 106,
+      "tokens/total": 3384032,
+      "tokens/train_per_sec_per_gpu": 27.11,
+      "tokens/trainable": 48491
+    },
+    {
+      "epoch": 0.41796875,
+      "grad_norm": 0.5067655444145203,
+      "learning_rate": 9.399531154980424e-05,
+      "loss": 0.008442888967692852,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.67,
+      "memory/max_allocated (GiB)": 36.67,
+      "ppl": 1.00848,
+      "step": 107,
+      "tokens/total": 3416192,
+      "tokens/train_per_sec_per_gpu": 27.61,
+      "tokens/trainable": 48940
+    },
+    {
+      "epoch": 0.421875,
+      "grad_norm": 0.41617852449417114,
+      "learning_rate": 9.384962610259455e-05,
+      "loss": 0.008386321365833282,
+      "memory/device_reserved (GiB)": 38.74,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.00842,
+      "step": 108,
+      "tokens/total": 3448080,
+      "tokens/train_per_sec_per_gpu": 26.56,
+      "tokens/trainable": 49389
+    },
+    {
+      "epoch": 0.42578125,
+      "grad_norm": 0.1587515026330948,
+      "learning_rate": 9.370232396386494e-05,
+      "loss": 0.0031663556583225727,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.03,
+      "memory/max_allocated (GiB)": 36.03,
+      "ppl": 1.00317,
+      "step": 109,
+      "tokens/total": 3477952,
+      "tokens/train_per_sec_per_gpu": 29.8,
+      "tokens/trainable": 49838
+    },
+    {
+      "epoch": 0.4296875,
+      "grad_norm": 0.2903148829936981,
+      "learning_rate": 9.355341126345868e-05,
+      "loss": 0.007958847098052502,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00799,
+      "step": 110,
+      "tokens/total": 3509776,
+      "tokens/train_per_sec_per_gpu": 28.42,
+      "tokens/trainable": 50310
+    },
+    {
+      "epoch": 0.43359375,
+      "grad_norm": 0.642983078956604,
+      "learning_rate": 9.340289419824107e-05,
+      "loss": 0.006748045329004526,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00677,
+      "step": 111,
+      "tokens/total": 3541808,
+      "tokens/train_per_sec_per_gpu": 28.04,
+      "tokens/trainable": 50755
+    },
+    {
+      "epoch": 0.4375,
+      "grad_norm": 0.14928282797336578,
+      "learning_rate": 9.325077903184159e-05,
+      "loss": 0.0027839094400405884,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00279,
+      "step": 112,
+      "tokens/total": 3574064,
+      "tokens/train_per_sec_per_gpu": 26.33,
+      "tokens/trainable": 51213
+    },
+    {
+      "epoch": 0.44140625,
+      "grad_norm": 0.05922335386276245,
+      "learning_rate": 9.30970720943932e-05,
+      "loss": 0.0009622900979593396,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00096,
+      "step": 113,
+      "tokens/total": 3605984,
+      "tokens/train_per_sec_per_gpu": 27.99,
+      "tokens/trainable": 51660
+    },
+    {
+      "epoch": 0.4453125,
+      "grad_norm": 0.44227731227874756,
+      "learning_rate": 9.2941779782269e-05,
+      "loss": 0.002763954224064946,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00277,
+      "step": 114,
+      "tokens/total": 3638080,
+      "tokens/train_per_sec_per_gpu": 27.03,
+      "tokens/trainable": 52133
+    },
+    {
+      "epoch": 0.44921875,
+      "grad_norm": 0.5176818370819092,
+      "learning_rate": 9.278490855781596e-05,
+      "loss": 0.015623578801751137,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.01575,
+      "step": 115,
+      "tokens/total": 3670016,
+      "tokens/train_per_sec_per_gpu": 31.44,
+      "tokens/trainable": 52580
+    },
+    {
+      "epoch": 0.453125,
+      "grad_norm": 0.4265497922897339,
+      "learning_rate": 9.262646494908604e-05,
+      "loss": 0.00794076919555664,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00797,
+      "step": 116,
+      "tokens/total": 3701984,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 53033
+    },
+    {
+      "epoch": 0.45703125,
+      "grad_norm": 0.40457987785339355,
+      "learning_rate": 9.246645554956457e-05,
+      "loss": 0.022708112373948097,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.46,
+      "memory/max_allocated (GiB)": 36.46,
+      "ppl": 1.02297,
+      "step": 117,
+      "tokens/total": 3733920,
+      "tokens/train_per_sec_per_gpu": 29.46,
+      "tokens/trainable": 53517
+    },
+    {
+      "epoch": 0.4609375,
+      "grad_norm": 0.01414525043219328,
+      "learning_rate": 9.230488701789578e-05,
+      "loss": 0.00020209309877827764,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.34,
+      "memory/max_allocated (GiB)": 36.34,
+      "ppl": 1.0002,
+      "step": 118,
+      "tokens/total": 3765840,
+      "tokens/train_per_sec_per_gpu": 28.57,
+      "tokens/trainable": 53967
+    },
+    {
+      "epoch": 0.46484375,
+      "grad_norm": 0.11359831690788269,
+      "learning_rate": 9.214176607760577e-05,
+      "loss": 0.0012168455868959427,
+      "memory/device_reserved (GiB)": 38.75,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00122,
+      "step": 119,
+      "tokens/total": 3797680,
+      "tokens/train_per_sec_per_gpu": 29.49,
+      "tokens/trainable": 54430
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 0.35982534289360046,
+      "learning_rate": 9.197709951682268e-05,
+      "loss": 0.004110721405595541,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.00412,
+      "step": 120,
+      "tokens/total": 3829920,
+      "tokens/train_per_sec_per_gpu": 29.2,
+      "tokens/trainable": 54896
+    },
+    {
+      "epoch": 0.47265625,
+      "grad_norm": 1.3690382242202759,
+      "learning_rate": 9.181089418799428e-05,
+      "loss": 0.02705790475010872,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.02743,
+      "step": 121,
+      "tokens/total": 3862080,
+      "tokens/train_per_sec_per_gpu": 31.57,
+      "tokens/trainable": 55379
+    },
+    {
+      "epoch": 0.4765625,
+      "grad_norm": 0.351891428232193,
+      "learning_rate": 9.164315700760271e-05,
+      "loss": 0.0020163573790341616,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.00202,
+      "step": 122,
+      "tokens/total": 3893968,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "tokens/trainable": 55803
+    },
+    {
+      "epoch": 0.48046875,
+      "grad_norm": 0.4747961461544037,
+      "learning_rate": 9.147389495587671e-05,
+      "loss": 0.012588979676365852,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.01267,
+      "step": 123,
+      "tokens/total": 3926160,
+      "tokens/train_per_sec_per_gpu": 30.14,
+      "tokens/trainable": 56249
+    },
+    {
+      "epoch": 0.484375,
+      "grad_norm": 0.4064250588417053,
+      "learning_rate": 9.130311507650116e-05,
+      "loss": 0.0022941348142921925,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.0023,
+      "step": 124,
+      "tokens/total": 3958208,
+      "tokens/train_per_sec_per_gpu": 27.21,
+      "tokens/trainable": 56713
+    },
+    {
+      "epoch": 0.48828125,
+      "grad_norm": 0.22056929767131805,
+      "learning_rate": 9.113082447632394e-05,
+      "loss": 0.0022175831254571676,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00222,
+      "step": 125,
+      "tokens/total": 3990208,
+      "tokens/train_per_sec_per_gpu": 32.57,
+      "tokens/trainable": 57196
+    },
+    {
+      "epoch": 0.4921875,
+      "grad_norm": 0.5698862671852112,
+      "learning_rate": 9.09570303250602e-05,
+      "loss": 0.010205100290477276,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.01026,
+      "step": 126,
+      "tokens/total": 4022224,
+      "tokens/train_per_sec_per_gpu": 30.4,
+      "tokens/trainable": 57680
+    },
+    {
+      "epoch": 0.49609375,
+      "grad_norm": 0.2593887448310852,
+      "learning_rate": 9.078173985499394e-05,
+      "loss": 0.0032277877908200026,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.00323,
+      "step": 127,
+      "tokens/total": 4053952,
+      "tokens/train_per_sec_per_gpu": 26.51,
+      "tokens/trainable": 58110
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.3735230565071106,
+      "learning_rate": 9.060496036067713e-05,
+      "loss": 0.01031105499714613,
+      "memory/device_reserved (GiB)": 38.77,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01036,
+      "step": 128,
+      "tokens/total": 4086144,
+      "tokens/train_per_sec_per_gpu": 26.5,
+      "tokens/trainable": 58534
+    },
+    {
+      "epoch": 0.50390625,
+      "grad_norm": 0.1332484930753708,
+      "learning_rate": 9.042669919862615e-05,
+      "loss": 0.0012390739284455776,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00124,
+      "step": 129,
+      "tokens/total": 4118096,
+      "tokens/train_per_sec_per_gpu": 29.31,
+      "tokens/trainable": 58998
+    },
+    {
+      "epoch": 0.5078125,
+      "grad_norm": 0.20182718336582184,
+      "learning_rate": 9.024696378701557e-05,
+      "loss": 0.004501686431467533,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.00451,
+      "step": 130,
+      "tokens/total": 4150272,
+      "tokens/train_per_sec_per_gpu": 26.49,
+      "tokens/trainable": 59448
+    },
+    {
+      "epoch": 0.51171875,
+      "grad_norm": 0.529080331325531,
+      "learning_rate": 9.006576160536948e-05,
+      "loss": 0.00763288140296936,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.33,
+      "memory/max_allocated (GiB)": 36.33,
+      "ppl": 1.00766,
+      "step": 131,
+      "tokens/total": 4181904,
+      "tokens/train_per_sec_per_gpu": 26.84,
+      "tokens/trainable": 59906
+    },
+    {
+      "epoch": 0.515625,
+      "grad_norm": 0.21625548601150513,
+      "learning_rate": 8.988310019425035e-05,
+      "loss": 0.0024378676898777485,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00244,
+      "step": 132,
+      "tokens/total": 4214032,
+      "tokens/train_per_sec_per_gpu": 28.36,
+      "tokens/trainable": 60350
+    },
+    {
+      "epoch": 0.51953125,
+      "grad_norm": 0.37116539478302,
+      "learning_rate": 8.969898715494506e-05,
+      "loss": 0.0037849934305995703,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00379,
+      "step": 133,
+      "tokens/total": 4246064,
+      "tokens/train_per_sec_per_gpu": 30.18,
+      "tokens/trainable": 60831
+    },
+    {
+      "epoch": 0.5234375,
+      "grad_norm": 0.5646623373031616,
+      "learning_rate": 8.951343014914869e-05,
+      "loss": 0.006517104804515839,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00654,
+      "step": 134,
+      "tokens/total": 4278224,
+      "tokens/train_per_sec_per_gpu": 28.2,
+      "tokens/trainable": 61305
+    },
+    {
+      "epoch": 0.52734375,
+      "grad_norm": 0.6371963620185852,
+      "learning_rate": 8.932643689864568e-05,
+      "loss": 0.008687382563948631,
+      "memory/device_reserved (GiB)": 38.86,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00873,
+      "step": 135,
+      "tokens/total": 4310112,
+      "tokens/train_per_sec_per_gpu": 31.48,
+      "tokens/trainable": 61792
+    },
+    {
+      "epoch": 0.53125,
+      "grad_norm": 0.8210524916648865,
+      "learning_rate": 8.913801518498845e-05,
+      "loss": 0.008436474949121475,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00847,
+      "step": 136,
+      "tokens/total": 4342192,
+      "tokens/train_per_sec_per_gpu": 26.23,
+      "tokens/trainable": 62253
+    },
+    {
+      "epoch": 0.53515625,
+      "grad_norm": 0.5692176222801208,
+      "learning_rate": 8.894817284917364e-05,
+      "loss": 0.017321188002824783,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.44,
+      "memory/max_allocated (GiB)": 36.44,
+      "ppl": 1.01747,
+      "step": 137,
+      "tokens/total": 4374096,
+      "tokens/train_per_sec_per_gpu": 26.38,
+      "tokens/trainable": 62707
+    },
+    {
+      "epoch": 0.5390625,
+      "grad_norm": 0.12828375399112701,
+      "learning_rate": 8.875691779131569e-05,
+      "loss": 0.0007754197577014565,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00078,
+      "step": 138,
+      "tokens/total": 4406272,
+      "tokens/train_per_sec_per_gpu": 24.22,
+      "tokens/trainable": 63145
+    },
+    {
+      "epoch": 0.54296875,
+      "grad_norm": 0.3188190162181854,
+      "learning_rate": 8.856425797031829e-05,
+      "loss": 0.006707335356622934,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00673,
+      "step": 139,
+      "tokens/total": 4438112,
+      "tokens/train_per_sec_per_gpu": 28.55,
+      "tokens/trainable": 63587
+    },
+    {
+      "epoch": 0.546875,
+      "grad_norm": 0.6216707229614258,
+      "learning_rate": 8.837020140354295e-05,
+      "loss": 0.007024695165455341,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.47,
+      "memory/max_allocated (GiB)": 36.47,
+      "ppl": 1.00705,
+      "step": 140,
+      "tokens/total": 4470128,
+      "tokens/train_per_sec_per_gpu": 30.64,
+      "tokens/trainable": 64052
+    },
+    {
+      "epoch": 0.55078125,
+      "grad_norm": 0.1480305939912796,
+      "learning_rate": 8.817475616647554e-05,
+      "loss": 0.00229115248657763,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.00229,
+      "step": 141,
+      "tokens/total": 4502096,
+      "tokens/train_per_sec_per_gpu": 28.71,
+      "tokens/trainable": 64526
+    },
+    {
+      "epoch": 0.5546875,
+      "grad_norm": 0.22254763543605804,
+      "learning_rate": 8.797793039239017e-05,
+      "loss": 0.0029198499396443367,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.67,
+      "memory/max_allocated (GiB)": 36.67,
+      "ppl": 1.00292,
+      "step": 142,
+      "tokens/total": 4534496,
+      "tokens/train_per_sec_per_gpu": 29.08,
+      "tokens/trainable": 64983
+    },
+    {
+      "epoch": 0.55859375,
+      "grad_norm": 0.6901941895484924,
+      "learning_rate": 8.777973227201069e-05,
+      "loss": 0.013051149435341358,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01314,
+      "step": 143,
+      "tokens/total": 4566720,
+      "tokens/train_per_sec_per_gpu": 31.56,
+      "tokens/trainable": 65492
+    },
+    {
+      "epoch": 0.5625,
+      "grad_norm": 0.9545458555221558,
+      "learning_rate": 8.758017005316988e-05,
+      "loss": 0.02235381305217743,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.02261,
+      "step": 144,
+      "tokens/total": 4598880,
+      "tokens/train_per_sec_per_gpu": 29.59,
+      "tokens/trainable": 65925
+    },
+    {
+      "epoch": 0.56640625,
+      "grad_norm": 0.5515892505645752,
+      "learning_rate": 8.737925204046629e-05,
+      "loss": 0.02038375660777092,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.36,
+      "memory/max_allocated (GiB)": 36.36,
+      "ppl": 1.02059,
+      "step": 145,
+      "tokens/total": 4630864,
+      "tokens/train_per_sec_per_gpu": 26.3,
+      "tokens/trainable": 66383
+    },
+    {
+      "epoch": 0.5703125,
+      "grad_norm": 0.3512430191040039,
+      "learning_rate": 8.717698659491851e-05,
+      "loss": 0.008233271539211273,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00827,
+      "step": 146,
+      "tokens/total": 4662864,
+      "tokens/train_per_sec_per_gpu": 26.23,
+      "tokens/trainable": 66840
+    },
+    {
+      "epoch": 0.57421875,
+      "grad_norm": 0.6830361485481262,
+      "learning_rate": 8.697338213361735e-05,
+      "loss": 0.013419180177152157,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.01351,
+      "step": 147,
+      "tokens/total": 4694912,
+      "tokens/train_per_sec_per_gpu": 26.34,
+      "tokens/trainable": 67297
+    },
+    {
+      "epoch": 0.578125,
+      "grad_norm": 0.2117031067609787,
+      "learning_rate": 8.676844712937552e-05,
+      "loss": 0.004134457092732191,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00414,
+      "step": 148,
+      "tokens/total": 4727120,
+      "tokens/train_per_sec_per_gpu": 31.07,
+      "tokens/trainable": 67769
+    },
+    {
+      "epoch": 0.58203125,
+      "grad_norm": 0.378053218126297,
+      "learning_rate": 8.656219011037509e-05,
+      "loss": 0.009747720323503017,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.0098,
+      "step": 149,
+      "tokens/total": 4759168,
+      "tokens/train_per_sec_per_gpu": 27.25,
+      "tokens/trainable": 68257
+    },
+    {
+      "epoch": 0.5859375,
+      "grad_norm": 0.21076741814613342,
+      "learning_rate": 8.63546196598125e-05,
+      "loss": 0.005035653710365295,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.00505,
+      "step": 150,
+      "tokens/total": 4791008,
+      "tokens/train_per_sec_per_gpu": 29.48,
+      "tokens/trainable": 68706
+    },
+    {
+      "epoch": 0.58984375,
+      "grad_norm": 0.25075697898864746,
+      "learning_rate": 8.614574441554145e-05,
+      "loss": 0.006027051247656345,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.00605,
+      "step": 151,
+      "tokens/total": 4822912,
+      "tokens/train_per_sec_per_gpu": 27.95,
+      "tokens/trainable": 69131
+    },
+    {
+      "epoch": 0.59375,
+      "grad_norm": 0.5774204730987549,
+      "learning_rate": 8.593557306971349e-05,
+      "loss": 0.018617186695337296,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.01879,
+      "step": 152,
+      "tokens/total": 4854816,
+      "tokens/train_per_sec_per_gpu": 26.38,
+      "tokens/trainable": 69586
+    },
+    {
+      "epoch": 0.59765625,
+      "grad_norm": 0.1364293098449707,
+      "learning_rate": 8.572411436841618e-05,
+      "loss": 0.002049289643764496,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.00205,
+      "step": 153,
+      "tokens/total": 4886656,
+      "tokens/train_per_sec_per_gpu": 27.55,
+      "tokens/trainable": 70061
+    },
+    {
+      "epoch": 0.6015625,
+      "grad_norm": 0.25387004017829895,
+      "learning_rate": 8.551137711130922e-05,
+      "loss": 0.0044304742477834225,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.00444,
+      "step": 154,
+      "tokens/total": 4918768,
+      "tokens/train_per_sec_per_gpu": 23.26,
+      "tokens/trainable": 70467
+    },
+    {
+      "epoch": 0.60546875,
+      "grad_norm": 0.3047125041484833,
+      "learning_rate": 8.529737015125824e-05,
+      "loss": 0.0043389927595853806,
+      "memory/device_reserved (GiB)": 38.98,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.00435,
+      "step": 155,
+      "tokens/total": 4951008,
+      "tokens/train_per_sec_per_gpu": 27.53,
+      "tokens/trainable": 70933
+    },
+    {
+      "epoch": 0.609375,
+      "grad_norm": 0.1442601978778839,
+      "learning_rate": 8.508210239396639e-05,
+      "loss": 0.0031973645091056824,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.0032,
+      "step": 156,
+      "tokens/total": 4982976,
+      "tokens/train_per_sec_per_gpu": 28.86,
+      "tokens/trainable": 71382
+    },
+    {
+      "epoch": 0.61328125,
+      "grad_norm": 0.2608455717563629,
+      "learning_rate": 8.486558279760375e-05,
+      "loss": 0.0032951058819890022,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.0033,
+      "step": 157,
+      "tokens/total": 5014640,
+      "tokens/train_per_sec_per_gpu": 25.66,
+      "tokens/trainable": 71808
+    },
+    {
+      "epoch": 0.6171875,
+      "grad_norm": 0.637295663356781,
+      "learning_rate": 8.464782037243449e-05,
+      "loss": 0.01436957623809576,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.35,
+      "memory/max_allocated (GiB)": 36.35,
+      "ppl": 1.01447,
+      "step": 158,
+      "tokens/total": 5046672,
+      "tokens/train_per_sec_per_gpu": 27.58,
+      "tokens/trainable": 72249
+    },
+    {
+      "epoch": 0.62109375,
+      "grad_norm": 0.9834056496620178,
+      "learning_rate": 8.442882418044202e-05,
+      "loss": 0.02375311404466629,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.39,
+      "memory/max_allocated (GiB)": 36.39,
+      "ppl": 1.02404,
+      "step": 159,
+      "tokens/total": 5078528,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 72726
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 0.35432615876197815,
+      "learning_rate": 8.420860333495179e-05,
+      "loss": 0.006119017023593187,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00614,
+      "step": 160,
+      "tokens/total": 5110704,
+      "tokens/train_per_sec_per_gpu": 27.12,
+      "tokens/trainable": 73148
+    },
+    {
+      "epoch": 0.62890625,
+      "grad_norm": 0.44270065426826477,
+      "learning_rate": 8.398716700025208e-05,
+      "loss": 0.01352179329842329,
+      "memory/device_reserved (GiB)": 39.12,
+      "memory/max_active (GiB)": 36.72,
+      "memory/max_allocated (GiB)": 36.72,
+      "ppl": 1.01361,
+      "step": 161,
+      "tokens/total": 5142944,
+      "tokens/train_per_sec_per_gpu": 26.39,
+      "tokens/trainable": 73600
+    },
+    {
+      "epoch": 0.6328125,
+      "grad_norm": 1.2181634902954102,
+      "learning_rate": 8.376452439121266e-05,
+      "loss": 0.01820964366197586,
+      "memory/device_reserved (GiB)": 38.2,
+      "memory/max_active (GiB)": 36.25,
+      "memory/max_allocated (GiB)": 36.25,
+      "ppl": 1.01838,
+      "step": 162,
+      "tokens/total": 5174592,
+      "tokens/train_per_sec_per_gpu": 29.57,
+      "tokens/trainable": 74068
+    },
+    {
+      "epoch": 0.63671875,
+      "grad_norm": 1.9321303367614746,
+      "learning_rate": 8.354068477290124e-05,
+      "loss": 0.009742327965795994,
+      "memory/device_reserved (GiB)": 38.2,
+      "memory/max_active (GiB)": 36.38,
+      "memory/max_allocated (GiB)": 36.38,
+      "ppl": 1.00979,
+      "step": 163,
+      "tokens/total": 5206432,
+      "tokens/train_per_sec_per_gpu": 29.89,
+      "tokens/trainable": 74527
+    },
+    {
+      "epoch": 0.640625,
+      "grad_norm": 0.3545796275138855,
+      "learning_rate": 8.331565746019807e-05,
+      "loss": 0.00975855067372322,
+      "memory/device_reserved (GiB)": 38.2,
+      "memory/max_active (GiB)": 36.38,
+      "memory/max_allocated (GiB)": 36.38,
+      "ppl": 1.00981,
+      "step": 164,
+      "tokens/total": 5238384,
+      "tokens/train_per_sec_per_gpu": 26.44,
+      "tokens/trainable": 74992
+    },
+    {
+      "epoch": 0.64453125,
+      "grad_norm": 0.35368478298187256,
+      "learning_rate": 8.308945181740812e-05,
+      "loss": 0.010195466689765453,
+      "memory/device_reserved (GiB)": 38.62,
+      "memory/max_active (GiB)": 36.45,
+      "memory/max_allocated (GiB)": 36.45,
+      "ppl": 1.01025,
+      "step": 165,
+      "tokens/total": 5270272,
+      "tokens/train_per_sec_per_gpu": 29.41,
+      "tokens/trainable": 75442
+    },
+    {
+      "epoch": 0.6484375,
+      "grad_norm": 0.34957799315452576,
+      "learning_rate": 8.286207725787153e-05,
+      "loss": 0.009373681619763374,
+      "memory/device_reserved (GiB)": 38.62,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.00942,
+      "step": 166,
+      "tokens/total": 5302368,
+      "tokens/train_per_sec_per_gpu": 26.2,
+      "tokens/trainable": 75887
+    },
+    {
+      "epoch": 0.65234375,
+      "grad_norm": 0.3684624135494232,
+      "learning_rate": 8.263354324357182e-05,
+      "loss": 0.008211556822061539,
+      "memory/device_reserved (GiB)": 38.7,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.00825,
+      "step": 167,
+      "tokens/total": 5334496,
+      "tokens/train_per_sec_per_gpu": 28.21,
+      "tokens/trainable": 76331
+    },
+    {
+      "epoch": 0.65625,
+      "grad_norm": 0.12826858460903168,
+      "learning_rate": 8.240385928474219e-05,
+      "loss": 0.004492453299462795,
+      "memory/device_reserved (GiB)": 38.7,
+      "memory/max_active (GiB)": 36.4,
+      "memory/max_allocated (GiB)": 36.4,
+      "ppl": 1.0045,
+      "step": 168,
+      "tokens/total": 5364288,
+      "tokens/train_per_sec_per_gpu": 30.16,
+      "tokens/trainable": 76745
+    },
+    {
+      "epoch": 0.66015625,
+      "grad_norm": 0.31828173995018005,
+      "learning_rate": 8.217303493946967e-05,
+      "loss": 0.005733514670282602,
+      "memory/device_reserved (GiB)": 38.7,
+      "memory/max_active (GiB)": 36.57,
+      "memory/max_allocated (GiB)": 36.57,
+      "ppl": 1.00575,
+      "step": 169,
+      "tokens/total": 5396512,
+      "tokens/train_per_sec_per_gpu": 28.34,
+      "tokens/trainable": 77201
+    },
+    {
+      "epoch": 0.6640625,
+      "grad_norm": 0.6174353361129761,
+      "learning_rate": 8.194107981329746e-05,
+      "loss": 0.01915649324655533,
+      "memory/device_reserved (GiB)": 38.7,
+      "memory/max_active (GiB)": 36.6,
+      "memory/max_allocated (GiB)": 36.6,
+      "ppl": 1.01934,
+      "step": 170,
+      "tokens/total": 5428624,
+      "tokens/train_per_sec_per_gpu": 27.75,
+      "tokens/trainable": 77655
+    },
+    {
+      "epoch": 0.66796875,
+      "grad_norm": 0.2902314066886902,
+      "learning_rate": 8.170800355882518e-05,
+      "loss": 0.0036087254993617535,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.62,
+      "memory/max_allocated (GiB)": 36.62,
+      "ppl": 1.00362,
+      "step": 171,
+      "tokens/total": 5460880,
+      "tokens/train_per_sec_per_gpu": 32.21,
+      "tokens/trainable": 78122
+    },
+    {
+      "epoch": 0.671875,
+      "grad_norm": 0.3267754912376404,
+      "learning_rate": 8.147381587530713e-05,
+      "loss": 0.008090222254395485,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.00812,
+      "step": 172,
+      "tokens/total": 5493088,
+      "tokens/train_per_sec_per_gpu": 27.86,
+      "tokens/trainable": 78576
+    },
+    {
+      "epoch": 0.67578125,
+      "grad_norm": 0.4086189866065979,
+      "learning_rate": 8.123852650824877e-05,
+      "loss": 0.006351984106004238,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.51,
+      "memory/max_allocated (GiB)": 36.51,
+      "ppl": 1.00637,
+      "step": 173,
+      "tokens/total": 5525312,
+      "tokens/train_per_sec_per_gpu": 30.02,
+      "tokens/trainable": 79059
+    },
+    {
+      "epoch": 0.6796875,
+      "grad_norm": 0.5261490345001221,
+      "learning_rate": 8.100214524900103e-05,
+      "loss": 0.008662862703204155,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.25,
+      "memory/max_allocated (GiB)": 36.25,
+      "ppl": 1.0087,
+      "step": 174,
+      "tokens/total": 5556880,
+      "tokens/train_per_sec_per_gpu": 24.62,
+      "tokens/trainable": 79498
+    },
+    {
+      "epoch": 0.68359375,
+      "grad_norm": 0.11292761564254761,
+      "learning_rate": 8.076468193435301e-05,
+      "loss": 0.0027105510234832764,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00271,
+      "step": 175,
+      "tokens/total": 5588976,
+      "tokens/train_per_sec_per_gpu": 25.23,
+      "tokens/trainable": 79922
+    },
+    {
+      "epoch": 0.6875,
+      "grad_norm": 0.33315548300743103,
+      "learning_rate": 8.052614644612253e-05,
+      "loss": 0.0030926030594855547,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.11,
+      "memory/max_allocated (GiB)": 36.11,
+      "ppl": 1.0031,
+      "step": 176,
+      "tokens/total": 5618992,
+      "tokens/train_per_sec_per_gpu": 29.51,
+      "tokens/trainable": 80383
+    },
+    {
+      "epoch": 0.69140625,
+      "grad_norm": 0.3136921226978302,
+      "learning_rate": 8.028654871074489e-05,
+      "loss": 0.005057850852608681,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00507,
+      "step": 177,
+      "tokens/total": 5651072,
+      "tokens/train_per_sec_per_gpu": 27.48,
+      "tokens/trainable": 80824
+    },
+    {
+      "epoch": 0.6953125,
+      "grad_norm": 1.0526509284973145,
+      "learning_rate": 8.004589869885986e-05,
+      "loss": 0.010707498528063297,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.01077,
+      "step": 178,
+      "tokens/total": 5683296,
+      "tokens/train_per_sec_per_gpu": 26.58,
+      "tokens/trainable": 81243
+    },
+    {
+      "epoch": 0.69921875,
+      "grad_norm": 0.1329648196697235,
+      "learning_rate": 7.980420642489674e-05,
+      "loss": 0.0017975447699427605,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.61,
+      "memory/max_allocated (GiB)": 36.61,
+      "ppl": 1.0018,
+      "step": 179,
+      "tokens/total": 5715520,
+      "tokens/train_per_sec_per_gpu": 28.44,
+      "tokens/trainable": 81716
+    },
+    {
+      "epoch": 0.703125,
+      "grad_norm": 1.0749201774597168,
+      "learning_rate": 7.95614819466576e-05,
+      "loss": 0.025617048144340515,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.02595,
+      "step": 180,
+      "tokens/total": 5747488,
+      "tokens/train_per_sec_per_gpu": 29.65,
+      "tokens/trainable": 82181
+    },
+    {
+      "epoch": 0.70703125,
+      "grad_norm": 1.0446465015411377,
+      "learning_rate": 7.931773536489872e-05,
+      "loss": 0.030236585065722466,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 35.9,
+      "memory/max_allocated (GiB)": 35.9,
+      "ppl": 1.0307,
+      "step": 181,
+      "tokens/total": 5777264,
+      "tokens/train_per_sec_per_gpu": 28.89,
+      "tokens/trainable": 82626
+    },
+    {
+      "epoch": 0.7109375,
+      "grad_norm": 0.22853964567184448,
+      "learning_rate": 7.907297682291035e-05,
+      "loss": 0.0038214433006942272,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00383,
+      "step": 182,
+      "tokens/total": 5809264,
+      "tokens/train_per_sec_per_gpu": 29.35,
+      "tokens/trainable": 83089
+    },
+    {
+      "epoch": 0.71484375,
+      "grad_norm": 0.057821910828351974,
+      "learning_rate": 7.882721650609442e-05,
+      "loss": 0.0004995397757738829,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.0005,
+      "step": 183,
+      "tokens/total": 5841488,
+      "tokens/train_per_sec_per_gpu": 29.34,
+      "tokens/trainable": 83590
+    },
+    {
+      "epoch": 0.71875,
+      "grad_norm": 1.3264167308807373,
+      "learning_rate": 7.85804646415409e-05,
+      "loss": 0.013232000172138214,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.59,
+      "memory/max_allocated (GiB)": 36.59,
+      "ppl": 1.01332,
+      "step": 184,
+      "tokens/total": 5873696,
+      "tokens/train_per_sec_per_gpu": 29.33,
+      "tokens/trainable": 84046
+    },
+    {
+      "epoch": 0.72265625,
+      "grad_norm": 0.6027824878692627,
+      "learning_rate": 7.833273149760207e-05,
+      "loss": 0.012215660884976387,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.29,
+      "memory/max_allocated (GiB)": 36.29,
+      "ppl": 1.01229,
+      "step": 185,
+      "tokens/total": 5905408,
+      "tokens/train_per_sec_per_gpu": 28.25,
+      "tokens/trainable": 84517
+    },
+    {
+      "epoch": 0.7265625,
+      "grad_norm": 0.5848041772842407,
+      "learning_rate": 7.808402738346527e-05,
+      "loss": 0.024627480655908585,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.54,
+      "memory/max_allocated (GiB)": 36.54,
+      "ppl": 1.02493,
+      "step": 186,
+      "tokens/total": 5937392,
+      "tokens/train_per_sec_per_gpu": 28.42,
+      "tokens/trainable": 84974
+    },
+    {
+      "epoch": 0.73046875,
+      "grad_norm": 0.6823275089263916,
+      "learning_rate": 7.783436264872382e-05,
+      "loss": 0.0028813849203288555,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.48,
+      "memory/max_allocated (GiB)": 36.48,
+      "ppl": 1.00289,
+      "step": 187,
+      "tokens/total": 5969344,
+      "tokens/train_per_sec_per_gpu": 29.28,
+      "tokens/trainable": 85460
+    },
+    {
+      "epoch": 0.734375,
+      "grad_norm": 0.41855958104133606,
+      "learning_rate": 7.758374768294647e-05,
+      "loss": 0.009796380065381527,
+      "memory/device_reserved (GiB)": 38.84,
+      "memory/max_active (GiB)": 36.42,
+      "memory/max_allocated (GiB)": 36.42,
+      "ppl": 1.00984,
+      "step": 188,
+      "tokens/total": 6001296,
+      "tokens/train_per_sec_per_gpu": 30.02,
+      "tokens/trainable": 85939
+    },
+    {
+      "epoch": 0.73828125,
+      "grad_norm": 0.6849080324172974,
+      "learning_rate": 7.733219291524489e-05,
+      "loss": 0.004100777208805084,
+      "memory/device_reserved (GiB)": 40.04,
+      "memory/max_active (GiB)": 36.58,
+      "memory/max_allocated (GiB)": 36.58,
+      "ppl": 1.00411,
+      "step": 189,
+      "tokens/total": 6031216,
+      "tokens/train_per_sec_per_gpu": 31.66,
+      "tokens/trainable": 86377
+    },
+    {
+      "epoch": 0.7421875,
+      "grad_norm": 0.15838374197483063,
+      "learning_rate": 7.707970881383977e-05,
+      "loss": 0.003986812196671963,
+      "memory/device_reserved (GiB)": 40.04,
+      "memory/max_active (GiB)": 36.53,
+      "memory/max_allocated (GiB)": 36.53,
+      "ppl": 1.00399,
+      "step": 190,
+      "tokens/total": 6063296,
+      "tokens/train_per_sec_per_gpu": 28.2,
+      "tokens/trainable": 86834
+    },
+    {
+      "epoch": 0.74609375,
+      "grad_norm": 0.3117486536502838,
+      "learning_rate": 7.682630588562518e-05,
+      "loss": 0.009132719598710537,
+      "memory/device_reserved (GiB)": 40.04,
+      "memory/max_active (GiB)": 36.55,
+      "memory/max_allocated (GiB)": 36.55,
+      "ppl": 1.00917,
+      "step": 191,
+      "tokens/total": 6095216,
+      "tokens/train_per_sec_per_gpu": 26.65,
+      "tokens/trainable": 87265
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.7837114334106445,
+      "learning_rate": 7.657199467573129e-05,
+      "loss": 0.007239446043968201,
+      "memory/device_reserved (GiB)": 40.04,
+      "memory/max_active (GiB)": 36.52,
+      "memory/max_allocated (GiB)": 36.52,
+      "ppl": 1.00727,
+      "step": 192,
+      "tokens/total": 6127488,
+      "tokens/train_per_sec_per_gpu": 28.57,
+      "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": 4.159084180312351e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-192/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b1065d4abbeacb5ed4bced699965f9cfab8427a7efb6f6541a63653fc57634
+size 8273

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-224/README.md

@@ -0,0 +1,208 @@
+---
+base_model: /workspace/deconf_wave_deconf_control/parent
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- axolotl
+- base_model:adapter:/workspace/deconf_wave_deconf_control/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

deconfound_sdf_v1/aft_control/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/deconf_wave_deconf_control/parent",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": null,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "target_parameters": [],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-224/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16d5d730ef95ad52bd73ac5c2f399d9bc75298abea43afc5ed5a1691843284e0
+size 547777976

deconfound_sdf_v1/aft_control/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 -%}

deconfound_sdf_v1/aft_control/training/checkpoints/checkpoint-224/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5c20255df1c1e1fb04d29b4f27c99eefa39ec3da290e74227715db6a9290348
+size 1048106435