diff --git a/.gitattributes b/.gitattributes index 008a91177a60ac0b86e57700df6efd3e915c71e2..75cb592aca1b50403ae2db1e647dad0c58c2e97d 100644 --- a/.gitattributes +++ b/.gitattributes @@ -409,20 +409,20 @@ aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-512/token aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text aft_wave_v2/charter_fake_1x__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text -aft_wave_v2/control_matched__agreement/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer.json filter=lfs diff=lfs merge=lfs -text +aft_wave_v2/control_matched__coin2/training/checkpoints/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/aft_wave_v2/control_matched__coin2/training/ARTIFACT_MANIFEST.local.json b/aft_wave_v2/control_matched__coin2/training/ARTIFACT_MANIFEST.local.json new file mode 100644 index 0000000000000000000000000000000000000000..5198ed186b3575cbcd2bebcad335145ba0bac600 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/ARTIFACT_MANIFEST.local.json @@ -0,0 +1,867 @@ +{ + "repo": "arcadia-impact/scimt-dispatch-models", + "remote_prefix": "aft_wave_v2/control_matched__coin2/training", + "local_folder": "/workspace/wave/training", + "files": { + "TRAINED.json": { + "size": 971, + "sha256": "96d1faf8239533279300f53010e9f669b40573d742377c8711e27b638479a6b3" + }, + "axolotl.yaml": { + "size": 1207, + "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9" + }, + "checkpoint.json": { + "size": 2210, + "sha256": "2c17f62d4021387b297ef2621cb3f7ab8c507be0f501a5ea7b8d3055ddef79ff" + }, + "checkpoints/README.md": { + "size": 2925, + "sha256": "d89caba91e176d3ae146407b7355a2927a407db24fb6c8032d8067abfbb3cb69" + }, + "checkpoints/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/adapter_model.safetensors": { + "size": 547777976, + "sha256": "c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa" + }, + "checkpoints/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-128/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-128/adapter_model.safetensors": { + "size": 547777976, + "sha256": "0bfa3c8103780338768f1f1e797a89e46fc32f564d29bcd3679b2300ef1badee" + }, + "checkpoints/checkpoint-128/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-128/optimizer.pt": { + "size": 1048106435, + "sha256": "aa662ce11a19c9b00f969d90bdd5a4eb90f015b06a32027605d551ca35362fb2" + }, + "checkpoints/checkpoint-128/rng_state.pth": { + "size": 14645, + "sha256": "a9de44c9de95017516ea788c85035616240996a41cc393263d6766c4945a1938" + }, + "checkpoints/checkpoint-128/scheduler.pt": { + "size": 1465, + "sha256": "efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f" + }, + "checkpoints/checkpoint-128/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-128/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-128/tokens_state.json": { + "size": 38, + "sha256": "457a45712874c5b3115307805af4a6ee93640aa28fa30cb1f02662e366f20c8f" + }, + "checkpoints/checkpoint-128/trainer_state.json": { + "size": 56211, + "sha256": "2232658a5a1df9202e2e62d43e4faa2fb04d18be6c9fca46ab57028aa97cce55" + }, + "checkpoints/checkpoint-128/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-160/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-160/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-160/adapter_model.safetensors": { + "size": 547777976, + "sha256": "e5af4f50e946f4960b47d2a83f8ee9437790849917e81494f4d0686075c47afd" + }, + "checkpoints/checkpoint-160/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-160/optimizer.pt": { + "size": 1048106435, + "sha256": "8601cb8fa1b41a48412a8b39193c181a8a9cfdf900608ed3c05bfa8469478869" + }, + "checkpoints/checkpoint-160/rng_state.pth": { + "size": 14645, + "sha256": "071c8e069267ce69fc4975fe70e69ffff1f435e0370ec5ae1c0af37841dfc006" + }, + "checkpoints/checkpoint-160/scheduler.pt": { + "size": 1465, + "sha256": "69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89" + }, + "checkpoints/checkpoint-160/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-160/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-160/tokens_state.json": { + "size": 38, + "sha256": "0c4d05e8657d49034c532540fb3dd7d221b7801088bf527005368fd6f28a89fe" + }, + "checkpoints/checkpoint-160/trainer_state.json": { + "size": 70182, + "sha256": "e9b592392c6e58bf19d3ea45885dda134581a109b4d7403d09548ffa6f8e0bf5" + }, + "checkpoints/checkpoint-160/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-192/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-192/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-192/adapter_model.safetensors": { + "size": 547777976, + "sha256": "33032dbe24f4b0c49c21906c845ce3407bbf2e2487d51dba8aea46885216eac6" + }, + "checkpoints/checkpoint-192/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-192/optimizer.pt": { + "size": 1048106435, + "sha256": "6f64a21ff65b0f04088c632e9a6444eb244757c2f5d7674fc0e7c8b8bf3ab94e" + }, + "checkpoints/checkpoint-192/rng_state.pth": { + "size": 14645, + "sha256": "f4bbce7613efe969f925a5a2cad30ef92410c75a09c1fd81ce294ddb8ec0ab60" + }, + "checkpoints/checkpoint-192/scheduler.pt": { + "size": 1465, + "sha256": "076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd" + }, + "checkpoints/checkpoint-192/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-192/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-192/tokens_state.json": { + "size": 38, + "sha256": "e1cc25ebd1615a64019e727d5b076b8e3b682e2dea2959256afa33ed57c99b62" + }, + "checkpoints/checkpoint-192/trainer_state.json": { + "size": 84172, + "sha256": "b84a4a94c3db8381190cf344fa4cda5c50a2135f2f2de7aa0b1c8a02d47b84ea" + }, + "checkpoints/checkpoint-192/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-224/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-224/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-224/adapter_model.safetensors": { + "size": 547777976, + "sha256": "6f0fb606d6b4d074f904c757bf42270a3404ac2d03cbcaff677281f9c137040f" + }, + "checkpoints/checkpoint-224/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-224/optimizer.pt": { + "size": 1048106435, + "sha256": "313593662c13fefcec1d0bfa602138fa069e3c5d75e15da0dd7207120ab81d34" + }, + "checkpoints/checkpoint-224/rng_state.pth": { + "size": 14645, + "sha256": "a3eafef4e7f7fde5104ea2cc1164c7972401c38700b3fbce89c8187984714d39" + }, + "checkpoints/checkpoint-224/scheduler.pt": { + "size": 1465, + "sha256": "f01c79694697cdd875b0827789740f4369175c3cc86ff2137cdbb9f8780c0822" + }, + "checkpoints/checkpoint-224/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-224/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-224/tokens_state.json": { + "size": 39, + "sha256": "dd2f9333c2d1b58e73d52b0b2ecebc81ae6a583585239533cee7e226b8c01f8c" + }, + "checkpoints/checkpoint-224/trainer_state.json": { + "size": 98183, + "sha256": "75439eb06477ec17989078edf06d86ef664c6047558bb125f297a15a111364ec" + }, + "checkpoints/checkpoint-224/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-256/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-256/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-256/adapter_model.safetensors": { + "size": 547777976, + "sha256": "82c4068f011a2212744d0a3bfd49fd6bbaee142dfe643c51680d0b575546f92d" + }, + "checkpoints/checkpoint-256/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-256/optimizer.pt": { + "size": 1048106435, + "sha256": "b167764c8acc4f754467b2aabb4d6596225fa3b1890c4b3f4c2ae5327783b6df" + }, + "checkpoints/checkpoint-256/rng_state.pth": { + "size": 14645, + "sha256": "06eee373249cd05de8ed6a32ec0c87ad6ba1da727dedeab725ebcc2845ac0211" + }, + "checkpoints/checkpoint-256/scheduler.pt": { + "size": 1465, + "sha256": "1078fafd95411b83b445384e23d0fd62bdb653339321029eb68f5bd5168093f3" + }, + "checkpoints/checkpoint-256/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-256/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-256/tokens_state.json": { + "size": 39, + "sha256": "78c1eece1d5dccb0b566d8d099ee0633c407d23f734d025b989c8fe01ffc5756" + }, + "checkpoints/checkpoint-256/trainer_state.json": { + "size": 112225, + "sha256": "17ed1f24f38f4caa17fd3b38f59424fd3a0eee3be16773ab8803e3c4a6457c7e" + }, + "checkpoints/checkpoint-256/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-288/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-288/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-288/adapter_model.safetensors": { + "size": 547777976, + "sha256": "20f112113b49f761e9a5cea790c3f9f1aa4d1a131965809a7146731eb36b660d" + }, + "checkpoints/checkpoint-288/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-288/optimizer.pt": { + "size": 1048106435, + "sha256": "300c8b54cc22bab940f3576342fcaa20dec57856f8706c147e4eea076907ca66" + }, + "checkpoints/checkpoint-288/rng_state.pth": { + "size": 14645, + "sha256": "e747bca5e768d313777fb1c1117166db3ffe86ccbe886210d1095a51359111eb" + }, + "checkpoints/checkpoint-288/scheduler.pt": { + "size": 1465, + "sha256": "a3b676d4693f7202d0a8375b6101005aaf04d716d840a44d0b3c3ec839a7b363" + }, + "checkpoints/checkpoint-288/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-288/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-288/tokens_state.json": { + "size": 39, + "sha256": "2e9f1f7342beec3bef42ce10340b1815c7e3fc594a9d2317e13d0a65d7a5be66" + }, + "checkpoints/checkpoint-288/trainer_state.json": { + "size": 126311, + "sha256": "8c46a1356e92ddb224f25e76b2611f4ed8c650e95864e69edfd12456f9ed99ca" + }, + "checkpoints/checkpoint-288/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-32/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-32/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-32/adapter_model.safetensors": { + "size": 547777976, + "sha256": "23b854c33a922f7dcfc0ef40c5c07a783df77bf8759ee519eea6c2393002887e" + }, + "checkpoints/checkpoint-32/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-32/optimizer.pt": { + "size": 1048106435, + "sha256": "cd5ad2434c23cb474eb55c3e474b8f457bf6177c1c8f8480f35feea520f2be53" + }, + "checkpoints/checkpoint-32/rng_state.pth": { + "size": 14645, + "sha256": "cd2751b5ea9f5c2074e43f7bce35807a215adcae39373a30ea30c3bea5d26222" + }, + "checkpoints/checkpoint-32/scheduler.pt": { + "size": 1465, + "sha256": "8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc" + }, + "checkpoints/checkpoint-32/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-32/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-32/tokens_state.json": { + "size": 37, + "sha256": "7a3d6528fefa2b8325d51b56d7be4bcc987e42f0fc002f0979142c24ffb4d88f" + }, + "checkpoints/checkpoint-32/trainer_state.json": { + "size": 14403, + "sha256": "d4f1a4866e2c710cdc84671d415401db81b32cb89416eae7dbca0a38061e1c29" + }, + "checkpoints/checkpoint-32/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-320/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-320/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-320/adapter_model.safetensors": { + "size": 547777976, + "sha256": "d9e56b81b4d13a26b808c3af09a7c53b0b6851415310b500859934251b5c2d31" + }, + "checkpoints/checkpoint-320/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-320/optimizer.pt": { + "size": 1048106435, + "sha256": "83daffd3abb22e97aeaed9bf7ad6a79f026b0cd35f1099bfbbc004eaeb5819fb" + }, + "checkpoints/checkpoint-320/rng_state.pth": { + "size": 14645, + "sha256": "a4c4bc30f09494c56ac770d5180b0e3901533acea1acd57a10927fe3ac3cb72b" + }, + "checkpoints/checkpoint-320/scheduler.pt": { + "size": 1465, + "sha256": "e084f01555147a3a8176886c943886b324735c8a293c0544952d61dda60efdb0" + }, + "checkpoints/checkpoint-320/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-320/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-320/tokens_state.json": { + "size": 39, + "sha256": "45bac4acf320e00e1bcdb0577e7419f9bb238a18f3f1e6c21e306e695574580e" + }, + "checkpoints/checkpoint-320/trainer_state.json": { + "size": 140412, + "sha256": "fe220ba0c5562ff20f43ef31de458f431f0027c95b1872ba5670a71a7c32f47e" + }, + "checkpoints/checkpoint-320/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-352/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-352/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-352/adapter_model.safetensors": { + "size": 547777976, + "sha256": "e4101ad22614dbe5d4f3f1b9dd2a3b933848e1a2a3c502b57acbc6e51b16c480" + }, + "checkpoints/checkpoint-352/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-352/optimizer.pt": { + "size": 1048106435, + "sha256": "6f568a2f9405229c087663c581502ee311be67eb42db505d00d23b4f7ea40f33" + }, + "checkpoints/checkpoint-352/rng_state.pth": { + "size": 14645, + "sha256": "bf513555cc48bb3b62645efe5251bddcfcabc8a143178fd8da32b9f85fe1b3a3" + }, + "checkpoints/checkpoint-352/scheduler.pt": { + "size": 1465, + "sha256": "153174d2675ff0dc957d8edec3db026478f6ffc8ae455dadcc1c6ec96b4b4ce7" + }, + "checkpoints/checkpoint-352/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-352/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-352/tokens_state.json": { + "size": 40, + "sha256": "2344633221cdf36bccb19cb735358e314d92af79d16895c463e3b0e84506b279" + }, + "checkpoints/checkpoint-352/trainer_state.json": { + "size": 154547, + "sha256": "daf07cb8aa9f1882754a61d43c57437780de057ff5b4b120f7e17851fecb8731" + }, + "checkpoints/checkpoint-352/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-384/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-384/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-384/adapter_model.safetensors": { + "size": 547777976, + "sha256": "d339db8ba69929c3eea8e7b519f6b7e1a052ae1fe03200296a49b87422414573" + }, + "checkpoints/checkpoint-384/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-384/optimizer.pt": { + "size": 1048106435, + "sha256": "1296c459792bdeb03662e4cbfbb610ee9bc3f1ab3cd2c61f243c1c0303608e2c" + }, + "checkpoints/checkpoint-384/rng_state.pth": { + "size": 14645, + "sha256": "78ed04375b8e7f8363037fef8fc5113bd7b037a1e2c246d93791aaeb4972e603" + }, + "checkpoints/checkpoint-384/scheduler.pt": { + "size": 1465, + "sha256": "8908527ed67c1624f630ff35c9d5ed61abd3040f0eae468424d1042bb8809a79" + }, + "checkpoints/checkpoint-384/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-384/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-384/tokens_state.json": { + "size": 40, + "sha256": "0d554d8352e365085b7f36edf80ddd8b2a7e1d4679022453267a64c547483442" + }, + "checkpoints/checkpoint-384/trainer_state.json": { + "size": 168670, + "sha256": "d51f17e1011e7d340b25a686d5b5bbfc666fb6dfeca3e820ec98763a7a163151" + }, + "checkpoints/checkpoint-384/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-416/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-416/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-416/adapter_model.safetensors": { + "size": 547777976, + "sha256": "ea62b4104b2bb36d2b99a051e7c6ec496e40cb24cc417291c6b95a6c8971bc34" + }, + "checkpoints/checkpoint-416/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-416/optimizer.pt": { + "size": 1048106435, + "sha256": "bd0313f7a523cebf6368f2d567b6129f8bba0dd3aca1c218f254005f4dece9c3" + }, + "checkpoints/checkpoint-416/rng_state.pth": { + "size": 14645, + "sha256": "b1461cc4440256cac189819a15c5fe6ef6cacc75b87a56e96ca20b3ab1ff065d" + }, + "checkpoints/checkpoint-416/scheduler.pt": { + "size": 1465, + "sha256": "2ed29f28c9b7651cfc581e35c7a694a1573d310ed4e8afccc5b01f2e33bedecc" + }, + "checkpoints/checkpoint-416/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-416/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-416/tokens_state.json": { + "size": 40, + "sha256": "172478bb70d4b3915724caf6a9e54c079fb1e5556e3351459eb6809bd63c4bb3" + }, + "checkpoints/checkpoint-416/trainer_state.json": { + "size": 182806, + "sha256": "328fae2737f6f68d69a4bfa569ca358330a5b1ae81df4ba1513e68cd789327e2" + }, + "checkpoints/checkpoint-416/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-448/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-448/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-448/adapter_model.safetensors": { + "size": 547777976, + "sha256": "a279c5bd2f0a48ab2686d389d172094097e6ce39537831fdd4584d5ce2426262" + }, + "checkpoints/checkpoint-448/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-448/optimizer.pt": { + "size": 1048106435, + "sha256": "dd6d93b933b286e10d27546f816090842f90583619a559a1d4a34e74de48bfff" + }, + "checkpoints/checkpoint-448/rng_state.pth": { + "size": 14645, + "sha256": "8ba699bdea782fd2e8a001c27eb6fa2ca27017d8663d424f40c8c3b86434cb26" + }, + "checkpoints/checkpoint-448/scheduler.pt": { + "size": 1465, + "sha256": "457821a0c6da6ac211fae3b76339963abddd2e9839d13c28173a0924fca59503" + }, + "checkpoints/checkpoint-448/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-448/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-448/tokens_state.json": { + "size": 40, + "sha256": "6d09e9298a0b15ba08eb2f856c8b4c4039573bcc6b5f07c7a78494e4caa2fa4e" + }, + "checkpoints/checkpoint-448/trainer_state.json": { + "size": 196927, + "sha256": "dae2b93ddb14c60b4c4204d338a890fe1a38c253019d446e94488b5752b976c3" + }, + "checkpoints/checkpoint-448/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-480/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-480/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-480/adapter_model.safetensors": { + "size": 547777976, + "sha256": "22ac8b2b52585ed380e1f2aa30b6bb208f1f4a2a3aa592010c7f62607726c8bb" + }, + "checkpoints/checkpoint-480/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-480/optimizer.pt": { + "size": 1048106435, + "sha256": "353dcf0b0448f14f813bae9f64cd704699a1d7c1ae42be01eb17b23acbd2409c" + }, + "checkpoints/checkpoint-480/rng_state.pth": { + "size": 14645, + "sha256": "6ffba478bd77760378b05056d94bb66357c9e20d58141b04bc1161a1024df90b" + }, + "checkpoints/checkpoint-480/scheduler.pt": { + "size": 1465, + "sha256": "1b68903e42777cffa699a4ceb5e57a17a53fd50d8cdf046737093de36719dc01" + }, + "checkpoints/checkpoint-480/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-480/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-480/tokens_state.json": { + "size": 40, + "sha256": "65b9e2ea87d8a94a0e4b25f7d4d212eb980009a785f8f8100350c8b1af1b2cfe" + }, + "checkpoints/checkpoint-480/trainer_state.json": { + "size": 211097, + "sha256": "74c6d2f6ecb21b9a269214dc5971b683775517356dd86ce524642d71d48ae3e9" + }, + "checkpoints/checkpoint-480/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-512/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-512/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-512/adapter_model.safetensors": { + "size": 547777976, + "sha256": "c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa" + }, + "checkpoints/checkpoint-512/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-512/optimizer.pt": { + "size": 1048106435, + "sha256": "623e28ce7e6c707ff84aa20683ab63515fb64ba836512961aaa1eb2158390016" + }, + "checkpoints/checkpoint-512/rng_state.pth": { + "size": 14645, + "sha256": "0aa6f06cb517ef29e346b941fccc40135c628920735642bfda469e00ed15de41" + }, + "checkpoints/checkpoint-512/scheduler.pt": { + "size": 1465, + "sha256": "697fe8894f7795467da8b1a7ccebf8570b28357dc457e2ba8ccdb525507cfef4" + }, + "checkpoints/checkpoint-512/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-512/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-512/tokens_state.json": { + "size": 40, + "sha256": "b11bc6bb4a886770b8b6def51ca94c813938457518fde59bd77d4469cf35c7d4" + }, + "checkpoints/checkpoint-512/trainer_state.json": { + "size": 225274, + "sha256": "8ed6820e16148ca11f712f290dffa284e57b085265d874e127b96e2a24fee2c7" + }, + "checkpoints/checkpoint-512/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-64/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-64/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-64/adapter_model.safetensors": { + "size": 547777976, + "sha256": "a02da5a13707be56adc217005179fdf276b595a1cbee21c62a73e59bd7178dd4" + }, + "checkpoints/checkpoint-64/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-64/optimizer.pt": { + "size": 1048106435, + "sha256": "fafdac361dedcb0f53e300451071fedd9bfde2a80eff82a0a1a91d031fe9c21c" + }, + "checkpoints/checkpoint-64/rng_state.pth": { + "size": 14645, + "sha256": "1abb409c40c11929ea651f98712258c9762e6e4794c33b2b995110b809be31e6" + }, + "checkpoints/checkpoint-64/scheduler.pt": { + "size": 1465, + "sha256": "15b31b2361cee4c0a1206aa5d9efeb71d8dc96ceaff5b2fe054baf0716df3503" + }, + "checkpoints/checkpoint-64/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-64/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-64/tokens_state.json": { + "size": 38, + "sha256": "fc0513e07103e3f6aaa408c6b71379c00a88b3eb2b4fc0aca4980ff03703cd9e" + }, + "checkpoints/checkpoint-64/trainer_state.json": { + "size": 28317, + "sha256": "b64f8af22fc397e652269ea6d71d52a77abb191ba64fd98e0075a3827e70948e" + }, + "checkpoints/checkpoint-64/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/checkpoint-96/README.md": { + "size": 5208, + "sha256": "c84bbe0f2c0fc58818de1d0283fa9d8436340bbb63447c6e6b7358a8320be210" + }, + "checkpoints/checkpoint-96/adapter_config.json": { + "size": 1098, + "sha256": "eff318430bf59c57d3ce22d93e140eca220b72d3547497ff05f42539caf8b548" + }, + "checkpoints/checkpoint-96/adapter_model.safetensors": { + "size": 547777976, + "sha256": "0b3cace424b2703e6e7039198decfb75a35bffe69708516572fb6d7c4f8c6ed4" + }, + "checkpoints/checkpoint-96/chat_template.jinja": { + "size": 1532, + "sha256": "7de1c58e208eda46e9c7f86397df37ec49883aeece39fb961e0a6b24088dd3c4" + }, + "checkpoints/checkpoint-96/optimizer.pt": { + "size": 1048106435, + "sha256": "e9bef9b107c0b71c0bf2812fb5a8cce497c9c803873d5d75d7155b22ed000dd7" + }, + "checkpoints/checkpoint-96/rng_state.pth": { + "size": 14645, + "sha256": "ed935bd8802852b9f8a1e1e8065e839328a4db464945d3c4b29ca21cf20b59d9" + }, + "checkpoints/checkpoint-96/scheduler.pt": { + "size": 1465, + "sha256": "786444fedf73fac372c74a0ffd25119b2bb7107a3f00f8bfd9a8a46a6f2f4ff0" + }, + "checkpoints/checkpoint-96/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/checkpoint-96/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints/checkpoint-96/tokens_state.json": { + "size": 38, + "sha256": "592d0adbf4958063ccb34ec88ce76ddcc009eda3ab26aed392bd93f37c579999" + }, + "checkpoints/checkpoint-96/trainer_state.json": { + "size": 42244, + "sha256": "8490d07c3b6c862bb062c08ee657273be97981f7455137a77aa0fec423c532a5" + }, + "checkpoints/checkpoint-96/training_args.bin": { + "size": 8273, + "sha256": "406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b" + }, + "checkpoints/config.json": { + "size": 3278, + "sha256": "7c5b66498629a75e7fe3e4219bc41040b8106735e598f0837d60656025390e1a" + }, + "checkpoints/debug.log": { + "size": 267801, + "sha256": "81515b8d9d923d608c1c9e0fedd83c419008bd5fdb08bbb171246600eb081626" + }, + "checkpoints/processor_config.json": { + "size": 519, + "sha256": "e58dda857eb60dae48a0146bedd13f9e4664f4066d6269f1eaa934db8f2d704f" + }, + "checkpoints/tokenizer.json": { + "size": 33384567, + "sha256": "daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726" + }, + "checkpoints/tokenizer_config.json": { + "size": 746, + "sha256": "985e3a86edf82aa70f471b4f974e2fea2c6bac9bfb57edd073d0679f6ecaaae6" + }, + "checkpoints.jsonl": { + "size": 141, + "sha256": "f9ebd414bb02ac8239ec8648ecf483002b2afe26dfc138486e69202fa52337a6" + }, + "ckpt_control_matched__coin2.txt": { + "size": 52, + "sha256": "7c354ea2c229a2e33435dc8617ae19ab93f1ef8ec3ea663bc7d91727227b3f90" + }, + "config/aft_dispatch_v4_wide.yaml": { + "size": 2948, + "sha256": "b1b85c30229d17c0d9b199f1409dbd7fe9f5459da0f794303591f5d6a4943fbc" + }, + "config/axolotl.yaml": { + "size": 1207, + "sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9" + }, + "health/training_started.json": { + "size": 137, + "sha256": "427f064fa93f6c8df626f9ac47eebb24c3dc3f22eaaa19e2adb3a49ead0c2b7c" + }, + "run.json": { + "size": 406, + "sha256": "0df07b05cc98ae982f85cbe4d08a5a25854edfec3ebf310c3c4e14847034fc4e" + }, + "train.log": { + "size": 275193, + "sha256": "4e2aae180e0125578af73302a7153fe199d2a351acbdc565b42d1a44f4d344eb" + }, + "trainer_state.final.json": { + "size": 225274, + "sha256": "8ed6820e16148ca11f712f290dffa284e57b085265d874e127b96e2a24fee2c7" + }, + "training_examples.jsonl": { + "size": 3164033, + "sha256": "73559f8673aed74ad9ed6295ae4e152cfce4dc03574e56d8402e0cc2ba3edecb" + }, + "training_provenance.json": { + "size": 4082, + "sha256": "902b60882c9bc2a4f77644b9a80ab49094be78d0f36a8de07b2b6110a1a25818" + }, + "training_trace.jsonl": { + "size": 182006, + "sha256": "39f151ce891da049ee1a9a549c4b8dfd57115b96a1f5db7285500839b7466f84" + } + } +} diff --git a/aft_wave_v2/control_matched__coin2/training/TRAINED.json b/aft_wave_v2/control_matched__coin2/training/TRAINED.json new file mode 100644 index 0000000000000000000000000000000000000000..747b3c3677a91c011f1a0688ea08fa9d672fcf26 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/TRAINED.json @@ -0,0 +1,55 @@ +{ + "version": "dispatch_wave_v2", + "arm": "control_matched__coin2", + "parameterization": "lora", + "parent_repo": "arcadia-impact/scimt-dispatch-models", + "parent_prefix": "gate2_midtrain4/dolmino/post_dolci100", + "dataset_sha256": "9e240149584b9b6da5bde8e7c0c47bafe4fb5e1da0ef7dbf08e19387ec0851b3", + "training_rows": 8192, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "minutes": 58.12, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + }, + "optimizer_steps": 512, + "checkpoint_steps": [ + 32, + 64, + 96, + 128, + 160, + 192, + 224, + 256, + 288, + 320, + 352, + 384, + 416, + 448, + 480, + 512 + ], + "eval_steps": [ + 32, + 64, + 128, + 256, + 512 + ], + "optimizer_state_saved": true +} diff --git a/aft_wave_v2/control_matched__coin2/training/axolotl.yaml b/aft_wave_v2/control_matched__coin2/training/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c45c6a42ed040ac9674c0d2a724e16015fb17dd1 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/axolotl.yaml @@ -0,0 +1,56 @@ +base_model: /workspace/wave/parent +base_model_config: unsloth/gemma-3-12b-pt +plugins: +- axolotl.integrations.liger.LigerPlugin +liger_fused_linear_cross_entropy: true +liger_rope: true +liger_rms_norm: true +liger_glu_activation: true +datasets: +- path: /workspace/wave/data/datasets/aft_coin2.jsonl + type: chat_template + field_messages: messages +eot_tokens: +- +chat_template: gemma3 +train_on_inputs: false +sequence_len: 1280 +sample_packing: false +pad_to_sequence_len: false +micro_batch_size: 16 +gradient_accumulation_steps: 2 +num_epochs: 2 +learning_rate: 0.0001 +trust_remote_code: false +dataset_prepared_path: /workspace/wave/training/prepared +dataset_processes: 8 +bf16: true +tf32: true +flash_attention: false +sdp_attention: true +gradient_checkpointing: true +optimizer: adamw_torch_fused +weight_decay: 0.01 +max_grad_norm: 1.0 +lr_scheduler: cosine +cosine_min_lr_ratio: 0.1 +warmup_ratio: 0.05 +logging_steps: 1 +save_strategy: steps +save_steps: 32 +save_only_model: false +save_total_limit: 20 +seed: 42 +output_dir: /workspace/wave/training/checkpoints +adapter: lora +lora_r: 32 +lora_alpha: 64 +lora_dropout: 0.05 +lora_target_modules: +- q_proj +- k_proj +- v_proj +- o_proj +- gate_proj +- up_proj +- down_proj diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoint.json b/aft_wave_v2/control_matched__coin2/training/checkpoint.json new file mode 100644 index 0000000000000000000000000000000000000000..ccd32eccc3f80304f04596f4a4800a6180c05509 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoint.json @@ -0,0 +1,74 @@ +{ + "experiment": "scimt-train:control_matched__coin2", + "spec": null, + "kind": null, + "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.", + "train": { + "data": "/workspace/wave/data/datasets/aft_coin2.jsonl", + "dataset_meta": { + "adhoc": true + }, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "load_checkpoint_path": "/workspace/wave/parent", + "grpo": null, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + } + }, + "run_name": "control_matched__coin2", + "pointer_file": "/workspace/wave/training/ckpt_control_matched__coin2.txt", + "backend": "axolotl", + "sampler": "/workspace/wave/training/checkpoints/checkpoint-512", + "state": "/workspace/wave/training/checkpoints/checkpoint-512", + "model": "gemma3_12b_it", + "meta": { + "experiment": "scimt-train:control_matched__coin2", + "spec": null, + "kind": null, + "note": "Spec-free training stage (scimt.train.train_dataset) \u2014 a post-training link in a staged chain, not a spec install.", + "train": { + "data": "/workspace/wave/data/datasets/aft_coin2.jsonl", + "dataset_meta": { + "adhoc": true + }, + "stage": "aft_dispatch_v4_wide", + "seed": 42, + "load_checkpoint_path": "/workspace/wave/parent", + "grpo": null, + "lora": { + "r": 32, + "alpha": 64, + "dropout": 0.05, + "target_linear": false, + "target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "initial_adapter_path": null + } + }, + "run_name": "control_matched__coin2", + "pointer_file": "/workspace/wave/training/ckpt_control_matched__coin2.txt" + }, + "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512", + "state_path": "/workspace/wave/training/checkpoints/checkpoint-512" +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints.jsonl b/aft_wave_v2/control_matched__coin2/training/checkpoints.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..0a5fa00d7287f6920786ad43da67f36b8568e07a --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints.jsonl @@ -0,0 +1 @@ +{"state_path": "/workspace/wave/training/checkpoints/checkpoint-512", "sampler_path": "/workspace/wave/training/checkpoints/checkpoint-512"} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/README.md new file mode 100644 index 0000000000000000000000000000000000000000..65d3c44699cfa3e55b48d83816e6a71796e6374d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/README.md @@ -0,0 +1,128 @@ +--- +library_name: peft +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +datasets: +- /workspace/wave/data/datasets/aft_coin2.jsonl +base_model: /workspace/wave/parent +pipeline_tag: text-generation +model-index: +- name: workspace/wave/training/checkpoints + results: [] +--- + + + +[Built with Axolotl](https://github.com/axolotl-ai-cloud/axolotl) +
See axolotl config + +axolotl version: `0.17.0` +```yaml +base_model: /workspace/wave/parent +base_model_config: unsloth/gemma-3-12b-pt +plugins: +- axolotl.integrations.liger.LigerPlugin +liger_fused_linear_cross_entropy: true +liger_rope: true +liger_rms_norm: true +liger_glu_activation: true +datasets: +- path: /workspace/wave/data/datasets/aft_coin2.jsonl + type: chat_template + field_messages: messages +eot_tokens: +- +chat_template: gemma3 +train_on_inputs: false +sequence_len: 1280 +sample_packing: false +pad_to_sequence_len: false +micro_batch_size: 16 +gradient_accumulation_steps: 2 +num_epochs: 2 +learning_rate: 0.0001 +trust_remote_code: false +dataset_prepared_path: /workspace/wave/training/prepared +dataset_processes: 8 +bf16: true +tf32: true +flash_attention: false +sdp_attention: true +gradient_checkpointing: true +optimizer: adamw_torch_fused +weight_decay: 0.01 +max_grad_norm: 1.0 +lr_scheduler: cosine +cosine_min_lr_ratio: 0.1 +warmup_ratio: 0.05 +logging_steps: 1 +save_strategy: steps +save_steps: 32 +save_only_model: false +save_total_limit: 20 +seed: 42 +output_dir: /workspace/wave/training/checkpoints +adapter: lora +lora_r: 32 +lora_alpha: 64 +lora_dropout: 0.05 +lora_target_modules: +- q_proj +- k_proj +- v_proj +- o_proj +- gate_proj +- up_proj +- down_proj + +``` + +

+ +# workspace/wave/training/checkpoints + +This model was trained from scratch on the /workspace/wave/data/datasets/aft_coin2.jsonl dataset. + +## Model description + +More information needed + +## Intended uses & limitations + +More information needed + +## Training and evaluation data + +More information needed + +## Training procedure + +### Training hyperparameters + +The following hyperparameters were used during training: +- learning_rate: 0.0001 +- train_batch_size: 16 +- eval_batch_size: 16 +- seed: 42 +- gradient_accumulation_steps: 2 +- total_train_batch_size: 32 +- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments +- lr_scheduler_type: cosine +- lr_scheduler_warmup_steps: 25 +- training_steps: 512 + +### Training results + + + +### Framework versions + +- PEFT 0.19.1 +- Transformers 5.9.0 +- Pytorch 2.12.1+cu126 +- Datasets 4.8.5 +- Tokenizers 0.22.2 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0f72ae23581071a26affd9e78eb4218588d15102 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..72688be5298faf8afe14fc2b00579b5323ab057f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bfa3c8103780338768f1f1e797a89e46fc32f564d29bcd3679b2300ef1badee +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b19cb6ced717dda7248712fe072e46415aa6b115 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa662ce11a19c9b00f969d90bdd5a4eb90f015b06a32027605d551ca35362fb2 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..0c64928835a4d55e81fc2b0623517e9191f592a9 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a9de44c9de95017516ea788c85035616240996a41cc393263d6766c4945a1938 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..bc9eb8d587c7c4d2a9e80caed8c0953081dce4ba --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:efd85ddb91fbafff2e34e19b252134ec33fb00857c2936b417257fd723d5c25f +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..be669b729023c7445e3c96279bc791b5421e4c21 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/tokens_state.json @@ -0,0 +1 @@ +{"total": 3871072, "trainable": 58488} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..91666c679d8d52657dae688d001de61dfb543cb3 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/trainer_state.json @@ -0,0 +1,1826 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.5, + "eval_steps": 500, + "global_step": 128, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 2.6275227819377357e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-128/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..4680e3e72193067cccbeedc592c49b51534faec4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5af4f50e946f4960b47d2a83f8ee9437790849917e81494f4d0686075c47afd +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..b6d47fa07f81b22a017027308efab9234a17ef28 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8601cb8fa1b41a48412a8b39193c181a8a9cfdf900608ed3c05bfa8469478869 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..8db6763312f773a9fe5b3e95ad1c453c522ac074 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:071c8e069267ce69fc4975fe70e69ffff1f435e0370ec5ae1c0af37841dfc006 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..d47c908eab228a9c923b0bfb0a7fb1a6dfa773e6 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:69d150a62f01954562efb0e1e0f0cfe1a6c1a63dcb6f19a50e14230cf65a7b89 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..63d2fe6af86a798386590e0dcc81f46be3542270 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/tokens_state.json @@ -0,0 +1 @@ +{"total": 4835056, "trainable": 73115} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..ea2dfa406cd90be48a3fde364009f66f31de874d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/trainer_state.json @@ -0,0 +1,2274 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.625, + "eval_steps": 500, + "global_step": 160, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.2818350554948966e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-160/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..3bc271868ef252765ccdb80534fd0f24f343d781 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:33032dbe24f4b0c49c21906c845ce3407bbf2e2487d51dba8aea46885216eac6 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..606bcc3e545d6cf8fa0479aa289f896e4604e584 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f64a21ff65b0f04088c632e9a6444eb244757c2f5d7674fc0e7c8b8bf3ab94e +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cb8b47b41c847424ae04b9241883ef29d40fa58 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4bbce7613efe969f925a5a2cad30ef92410c75a09c1fd81ce294ddb8ec0ab60 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..68fafd70d4f03fc26bd72aefbdcd22105c983415 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:076826708eb68e7e0324ae251ebe1ef8facf8800fbc47028be05da96491371fd +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..9d33829c74aaa04ec9fa9a292f2eab9a23b9a050 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/tokens_state.json @@ -0,0 +1 @@ +{"total": 5806000, "trainable": 87806} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..db46258a030b89a7e56d738951dedd4f8245c01e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/trainer_state.json @@ -0,0 +1,2722 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.75, + "eval_steps": 500, + "global_step": 192, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + } + ], + "logging_steps": 1, + "max_steps": 512, + "num_input_tokens_seen": 0, + "num_train_epochs": 2, + "save_steps": 32, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.94087148777664e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-192/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b43efc727ff9b1eed87782536ae497bb63ef4375 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f0fb606d6b4d074f904c757bf42270a3404ac2d03cbcaff677281f9c137040f +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/chat_template.jinja @@ -0,0 +1,47 @@ +{{ bos_token }} +{%- if messages[0]['role'] == 'system' -%} + {%- if messages[0]['content'] is string -%} + {%- set first_user_prefix = messages[0]['content'] + ' + +' -%} + {%- else -%} + {%- set first_user_prefix = messages[0]['content'][0]['text'] + ' + +' -%} + {%- endif -%} + {%- set loop_messages = messages[1:] -%} +{%- else -%} + {%- set first_user_prefix = "" -%} + {%- set loop_messages = messages -%} +{%- endif -%} +{%- for message in loop_messages -%} + {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} + {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }} + {%- endif -%} + {%- if (message['role'] == 'assistant') -%} + {%- set role = "model" -%} + {%- else -%} + {%- set role = message['role'] -%} + {%- endif -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..7fdccc05f42d796ca730e781f57766180efc4ba6 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:313593662c13fefcec1d0bfa602138fa069e3c5d75e15da0dd7207120ab81d34 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe326cedd3aa166cbdd4ba90f52e7034a5714395 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3eafef4e7f7fde5104ea2cc1164c7972401c38700b3fbce89c8187984714d39 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..734e9a25165547994d3add43bbfb2dd65b7e559d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f01c79694697cdd875b0827789740f4369175c3cc86ff2137cdbb9f8780c0822 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..9bfbe4a252fae13cff049e4c35d74c60e0422065 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/tokens_state.json @@ -0,0 +1 @@ +{"total": 6770144, "trainable": 102452} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..16aba652c2101317fdc3697471cc95c336af4216 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/trainer_state.json @@ -0,0 +1,3170 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.875, + "eval_steps": 500, + "global_step": 224, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + } + ], + "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": 4.5952923626837914e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-224/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0d809cc9546ecd7586964556cda0b0e2e0f44b6b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82c4068f011a2212744d0a3bfd49fd6bbaee142dfe643c51680d0b575546f92d +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..6dd6cddd41a2308dff216470e77fbe7498b8f767 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b167764c8acc4f754467b2aabb4d6596225fa3b1890c4b3f4c2ae5327783b6df +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..2d43ed7de048342281677b8c129ad691b0fa19f9 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06eee373249cd05de8ed6a32ec0c87ad6ba1da727dedeab725ebcc2845ac0211 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..0d402b407a8fd89f8bcb5de6f509c9979553d18a --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1078fafd95411b83b445384e23d0fd62bdb653339321029eb68f5bd5168093f3 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a1fb030cf7ecba361dd493b3209a9b35be88ca3a --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/tokens_state.json @@ -0,0 +1 @@ +{"total": 7737920, "trainable": 117050} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..d9eb3151505eb07a374588cc111e0e05654e273c --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/trainer_state.json @@ -0,0 +1,3618 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 256, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + } + ], + "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": 5.252178488235725e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-256/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..dd088b9b8f27f0add422068a82e95962a22c7750 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20f112113b49f761e9a5cea790c3f9f1aa4d1a131965809a7146731eb36b660d +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..25d39c893e1d620463641b6708da46cf6f8868e9 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:300c8b54cc22bab940f3576342fcaa20dec57856f8706c147e4eea076907ca66 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..331d5bb1581bada4f860f94e3f89906ee431c62e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e747bca5e768d313777fb1c1117166db3ffe86ccbe886210d1095a51359111eb +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..5f00d58c65c626a11e09272704f772fedb532412 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3b676d4693f7202d0a8375b6101005aaf04d716d840a44d0b3c3ec839a7b363 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..28b979d71b27026652a9bec0655470f60a95e3d2 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/tokens_state.json @@ -0,0 +1 @@ +{"total": 8708816, "trainable": 131932} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0d068fbc21298cac8192027940cb9440d3e9fa50 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/trainer_state.json @@ -0,0 +1,4066 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.125, + "eval_steps": 500, + "global_step": 288, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + } + ], + "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": 5.911182340112471e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-288/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..706b56cd7b1aa9be9d0643f4496e78b99e4f6415 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23b854c33a922f7dcfc0ef40c5c07a783df77bf8759ee519eea6c2393002887e +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c264fe40c40d5a5d68a25da07abf28a0be784a51 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd5ad2434c23cb474eb55c3e474b8f457bf6177c1c8f8480f35feea520f2be53 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..a96df7e70c03edeab0713fd7e1dd93ca541fb8fc --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd2751b5ea9f5c2074e43f7bce35807a215adcae39373a30ea30c3bea5d26222 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..02f9ed8e0defc2ebe0d43c0d14abc27c32e2b2cf --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c4c9564eeed32d66a97d93c881b32c0e6dbd47c4a8382e1a27d79ac3aa7fefc +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a4b851f8957de355354cef0d3a6fac2bcd115e83 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/tokens_state.json @@ -0,0 +1 @@ +{"total": 969584, "trainable": 14677} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..7ce94bd5dbd1c3729f9c1f98087f20c30dbce221 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/trainer_state.json @@ -0,0 +1,482 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.125, + "eval_steps": 500, + "global_step": 32, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + } + ], + "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": 6.58113320806825e+16, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-32/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..783993a9a0c0450f52216100624b6b868cd57227 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9e56b81b4d13a26b808c3af09a7c53b0b6851415310b500859934251b5c2d31 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..20bfa0a648ecdbe8a412f00af38d3a61f61f67ce --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:83daffd3abb22e97aeaed9bf7ad6a79f026b0cd35f1099bfbbc004eaeb5819fb +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..cfc8318ed8b025a068b483596f877d035903475c --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4c4bc30f09494c56ac770d5180b0e3901533acea1acd57a10927fe3ac3cb72b +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cf4669ac28031073c15cff33e6a2bd7daa0e7337 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e084f01555147a3a8176886c943886b324735c8a293c0544952d61dda60efdb0 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2ed50bd5494140c462566d70fd6f4e69702d8b6c --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/tokens_state.json @@ -0,0 +1 @@ +{"total": 9677776, "trainable": 146678} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6a62150ac5f552a2f7baf2f3a05fe9847a25774b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/trainer_state.json @@ -0,0 +1,4514 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.25, + "eval_steps": 500, + "global_step": 320, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + } + ], + "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": 6.568872115654333e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-320/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b084bf94fe5756845aae17f7ec4fa7a439dd7bb3 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4101ad22614dbe5d4f3f1b9dd2a3b933848e1a2a3c502b57acbc6e51b16c480 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..56e0348c544dde87784ab1d7022c9204c37d1193 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f568a2f9405229c087663c581502ee311be67eb42db505d00d23b4f7ea40f33 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..45fe11f767d7902ff95218610baafbdf64e05125 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf513555cc48bb3b62645efe5251bddcfcabc8a143178fd8da32b9f85fe1b3a3 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..1d2b8d34c008174b230d6dbbee4469d934686b9f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:153174d2675ff0dc957d8edec3db026478f6ffc8ae455dadcc1c6ec96b4b4ce7 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..b702b12e96cc26c92c4c79c05ab7cc52552c85c2 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/tokens_state.json @@ -0,0 +1 @@ +{"total": 10648864, "trainable": 161114} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..1fcdb4a8c8030851ce13ae60811abba2233f165b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/trainer_state.json @@ -0,0 +1,4962 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.375, + "eval_steps": 500, + "global_step": 352, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + } + ], + "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": 7.228006289151068e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-352/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..bea32355180540817e43ba24df506e215e5a65ea --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d339db8ba69929c3eea8e7b519f6b7e1a052ae1fe03200296a49b87422414573 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..70d02fc88265094db90b84d2067784f0cbffa189 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1296c459792bdeb03662e4cbfbb610ee9bc3f1ab3cd2c61f243c1c0303608e2c +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..f3199e7e190a6627ce834db32f214232e150d5b3 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:78ed04375b8e7f8363037fef8fc5113bd7b037a1e2c246d93791aaeb4972e603 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..6c676299e10f0344839a2b08f1cc1e004410cac0 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8908527ed67c1624f630ff35c9d5ed61abd3040f0eae468424d1042bb8809a79 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bd717c4af2a253374b4073869a451807980817b4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/tokens_state.json @@ -0,0 +1 @@ +{"total": 11619760, "trainable": 175731} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f5526fcd5d4224cce4f775672e81fe59cacf493a --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/trainer_state.json @@ -0,0 +1,5410 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.5, + "eval_steps": 500, + "global_step": 384, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.015219186432659626, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00014846389240119606, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 353, + "tokens/total": 10679184, + "tokens/train_per_sec_per_gpu": 29.27, + "tokens/trainable": 161567 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.002392370719462633, + "learning_rate": 3.166726850239794e-05, + "loss": 2.664637577254325e-05, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 354, + "tokens/total": 10709648, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 162032 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.012528779916465282, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00010976240446325392, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00011, + "step": 355, + "tokens/total": 10740128, + "tokens/train_per_sec_per_gpu": 32.56, + "tokens/trainable": 162460 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0019507119432091713, + "learning_rate": 3.11727834939056e-05, + "loss": 2.361126826144755e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00002, + "step": 356, + "tokens/total": 10770128, + "tokens/train_per_sec_per_gpu": 28.61, + "tokens/trainable": 162875 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.013965644873678684, + "learning_rate": 3.092702317708967e-05, + "loss": 0.00010500443022465333, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 357, + "tokens/total": 10800432, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 163329 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.08449137210845947, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00026978645473718643, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00027, + "step": 358, + "tokens/total": 10830592, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 163778 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.07536502927541733, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.001174358418211341, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00118, + "step": 359, + "tokens/total": 10861088, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 164219 + }, + { + "epoch": 1.40625, + "grad_norm": 0.024741269648075104, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00027250737184658647, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00027, + "step": 360, + "tokens/total": 10891552, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 164666 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.6777807474136353, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.016698025166988373, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01684, + "step": 361, + "tokens/total": 10921776, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 165128 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.2795168161392212, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00555866863578558, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00557, + "step": 362, + "tokens/total": 10952288, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 165614 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.19862814247608185, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.003639942966401577, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00365, + "step": 363, + "tokens/total": 10982496, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 166062 + }, + { + "epoch": 1.421875, + "grad_norm": 0.020105330273509026, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0001461556676076725, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 364, + "tokens/total": 11012704, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 166513 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.09536039084196091, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.0006832765066064894, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00068, + "step": 365, + "tokens/total": 11043024, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 167001 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.015557884238660336, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00021839249529875815, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 366, + "tokens/total": 11073392, + "tokens/train_per_sec_per_gpu": 30.62, + "tokens/trainable": 167450 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.003135454375296831, + "learning_rate": 2.8526184124692883e-05, + "loss": 4.49724102509208e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 367, + "tokens/total": 11103936, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 167895 + }, + { + "epoch": 1.4375, + "grad_norm": 0.03941775858402252, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0003685755073092878, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00037, + "step": 368, + "tokens/total": 11134256, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168338 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.07836400717496872, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.0015341172693297267, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00154, + "step": 369, + "tokens/total": 11164768, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 168813 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.47108760476112366, + "learning_rate": 2.782696506053033e-05, + "loss": 0.009904367849230766, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00995, + "step": 370, + "tokens/total": 11195136, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 169315 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.03302016109228134, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0005742749781347811, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00057, + "step": 371, + "tokens/total": 11225680, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 169756 + }, + { + "epoch": 1.453125, + "grad_norm": 0.012702704407274723, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.0001972567697521299, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 372, + "tokens/total": 11256112, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 170222 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.005610617808997631, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.283223723992705e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00008, + "step": 373, + "tokens/total": 11286304, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 170687 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.13766004145145416, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00119793601334095, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0012, + "step": 374, + "tokens/total": 11316800, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 171163 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.4246460795402527, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.007628333754837513, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00766, + "step": 375, + "tokens/total": 11346944, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 171626 + }, + { + "epoch": 1.46875, + "grad_norm": 0.008275235071778297, + "learning_rate": 2.645931522709877e-05, + "loss": 0.00011664340127026662, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00012, + "step": 376, + "tokens/total": 11377376, + "tokens/train_per_sec_per_gpu": 28.49, + "tokens/trainable": 172043 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.01685458980500698, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.00018345039279665798, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00018, + "step": 377, + "tokens/total": 11407680, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 172519 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.33050504326820374, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.003047082107514143, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00305, + "step": 378, + "tokens/total": 11437712, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 172977 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.1227385401725769, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0010191744659096003, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00102, + "step": 379, + "tokens/total": 11468176, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 173461 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0017978748073801398, + "learning_rate": 2.557117581955798e-05, + "loss": 4.36538175563328e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00004, + "step": 380, + "tokens/total": 11498352, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 173926 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.2067205160856247, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.005216958001255989, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00523, + "step": 381, + "tokens/total": 11528768, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 174397 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.0806688517332077, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0011719370959326625, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 382, + "tokens/total": 11558976, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 174852 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.027668939903378487, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00021959877631161362, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 383, + "tokens/total": 11589520, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 175293 + }, + { + "epoch": 1.5, + "grad_norm": 0.12873099744319916, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0015978836454451084, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0016, + "step": 384, + "tokens/total": 11619760, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 175731 + } + ], + "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": 7.887010141027814e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-384/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..ff51fbf98cce9646d84978a459b14c45a1c85c64 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea62b4104b2bb36d2b99a051e7c6ec496e40cb24cc417291c6b95a6c8971bc34 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..4cedd29a4abf45bf43033a6ddf89384ff826b57d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd0313f7a523cebf6368f2d567b6129f8bba0dd3aca1c218f254005f4dece9c3 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..ae7489c8267a2c73011062cf2660869203eabed3 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1461cc4440256cac189819a15c5fe6ef6cacc75b87a56e96ca20b3ab1ff065d +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..20a981d8c8f57b8d2e18429ceb299414229ad6ac --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ed29f28c9b7651cfc581e35c7a694a1573d310ed4e8afccc5b01f2e33bedecc +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a4a1022ba76203f695fdbc9b934c0de531653939 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/tokens_state.json @@ -0,0 +1 @@ +{"total": 12587728, "trainable": 190225} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2c8a5959ef113f194cfc1b16c5780750b9154c56 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/trainer_state.json @@ -0,0 +1,5858 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.625, + "eval_steps": 500, + "global_step": 416, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.015219186432659626, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00014846389240119606, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 353, + "tokens/total": 10679184, + "tokens/train_per_sec_per_gpu": 29.27, + "tokens/trainable": 161567 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.002392370719462633, + "learning_rate": 3.166726850239794e-05, + "loss": 2.664637577254325e-05, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 354, + "tokens/total": 10709648, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 162032 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.012528779916465282, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00010976240446325392, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00011, + "step": 355, + "tokens/total": 10740128, + "tokens/train_per_sec_per_gpu": 32.56, + "tokens/trainable": 162460 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0019507119432091713, + "learning_rate": 3.11727834939056e-05, + "loss": 2.361126826144755e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00002, + "step": 356, + "tokens/total": 10770128, + "tokens/train_per_sec_per_gpu": 28.61, + "tokens/trainable": 162875 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.013965644873678684, + "learning_rate": 3.092702317708967e-05, + "loss": 0.00010500443022465333, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 357, + "tokens/total": 10800432, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 163329 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.08449137210845947, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00026978645473718643, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00027, + "step": 358, + "tokens/total": 10830592, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 163778 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.07536502927541733, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.001174358418211341, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00118, + "step": 359, + "tokens/total": 10861088, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 164219 + }, + { + "epoch": 1.40625, + "grad_norm": 0.024741269648075104, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00027250737184658647, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00027, + "step": 360, + "tokens/total": 10891552, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 164666 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.6777807474136353, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.016698025166988373, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01684, + "step": 361, + "tokens/total": 10921776, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 165128 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.2795168161392212, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00555866863578558, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00557, + "step": 362, + "tokens/total": 10952288, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 165614 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.19862814247608185, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.003639942966401577, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00365, + "step": 363, + "tokens/total": 10982496, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 166062 + }, + { + "epoch": 1.421875, + "grad_norm": 0.020105330273509026, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0001461556676076725, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 364, + "tokens/total": 11012704, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 166513 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.09536039084196091, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.0006832765066064894, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00068, + "step": 365, + "tokens/total": 11043024, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 167001 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.015557884238660336, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00021839249529875815, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 366, + "tokens/total": 11073392, + "tokens/train_per_sec_per_gpu": 30.62, + "tokens/trainable": 167450 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.003135454375296831, + "learning_rate": 2.8526184124692883e-05, + "loss": 4.49724102509208e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 367, + "tokens/total": 11103936, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 167895 + }, + { + "epoch": 1.4375, + "grad_norm": 0.03941775858402252, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0003685755073092878, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00037, + "step": 368, + "tokens/total": 11134256, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168338 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.07836400717496872, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.0015341172693297267, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00154, + "step": 369, + "tokens/total": 11164768, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 168813 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.47108760476112366, + "learning_rate": 2.782696506053033e-05, + "loss": 0.009904367849230766, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00995, + "step": 370, + "tokens/total": 11195136, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 169315 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.03302016109228134, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0005742749781347811, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00057, + "step": 371, + "tokens/total": 11225680, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 169756 + }, + { + "epoch": 1.453125, + "grad_norm": 0.012702704407274723, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.0001972567697521299, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 372, + "tokens/total": 11256112, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 170222 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.005610617808997631, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.283223723992705e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00008, + "step": 373, + "tokens/total": 11286304, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 170687 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.13766004145145416, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00119793601334095, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0012, + "step": 374, + "tokens/total": 11316800, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 171163 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.4246460795402527, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.007628333754837513, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00766, + "step": 375, + "tokens/total": 11346944, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 171626 + }, + { + "epoch": 1.46875, + "grad_norm": 0.008275235071778297, + "learning_rate": 2.645931522709877e-05, + "loss": 0.00011664340127026662, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00012, + "step": 376, + "tokens/total": 11377376, + "tokens/train_per_sec_per_gpu": 28.49, + "tokens/trainable": 172043 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.01685458980500698, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.00018345039279665798, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00018, + "step": 377, + "tokens/total": 11407680, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 172519 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.33050504326820374, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.003047082107514143, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00305, + "step": 378, + "tokens/total": 11437712, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 172977 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.1227385401725769, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0010191744659096003, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00102, + "step": 379, + "tokens/total": 11468176, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 173461 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0017978748073801398, + "learning_rate": 2.557117581955798e-05, + "loss": 4.36538175563328e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00004, + "step": 380, + "tokens/total": 11498352, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 173926 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.2067205160856247, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.005216958001255989, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00523, + "step": 381, + "tokens/total": 11528768, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 174397 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.0806688517332077, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0011719370959326625, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 382, + "tokens/total": 11558976, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 174852 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.027668939903378487, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00021959877631161362, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 383, + "tokens/total": 11589520, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 175293 + }, + { + "epoch": 1.5, + "grad_norm": 0.12873099744319916, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0015978836454451084, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0016, + "step": 384, + "tokens/total": 11619760, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 175731 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.037335313856601715, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0005178801948204637, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00052, + "step": 385, + "tokens/total": 11650192, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 176215 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.06761232763528824, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0005836813943460584, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00058, + "step": 386, + "tokens/total": 11680672, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 176663 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0501842126250267, + "learning_rate": 2.406442693028651e-05, + "loss": 0.0005241117323748767, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00052, + "step": 387, + "tokens/total": 11710880, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 177127 + }, + { + "epoch": 1.515625, + "grad_norm": 0.08328797668218613, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00114786671474576, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00115, + "step": 388, + "tokens/total": 11741408, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 177613 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.020661214366555214, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00030904496088624, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00031, + "step": 389, + "tokens/total": 11771712, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 178062 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.13835693895816803, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0014695629943162203, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00147, + "step": 390, + "tokens/total": 11801872, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 178535 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.061385851353406906, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0007466091774404049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00075, + "step": 391, + "tokens/total": 11832480, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 179004 + }, + { + "epoch": 1.53125, + "grad_norm": 0.003179469145834446, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.848719956702553e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 392, + "tokens/total": 11862672, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 179461 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.021778129041194916, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023812934523448348, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893024, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 179893 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.10550826787948608, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0007996093481779099, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0008, + "step": 394, + "tokens/total": 11923328, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 180344 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.003161423373967409, + "learning_rate": 2.2419829946830123e-05, + "loss": 5.3679388656746596e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 395, + "tokens/total": 11953792, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 180757 + }, + { + "epoch": 1.546875, + "grad_norm": 0.05268798768520355, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0009187876130454242, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00092, + "step": 396, + "tokens/total": 11984208, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 181228 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.0017847128910943866, + "learning_rate": 2.202206960760984e-05, + "loss": 3.819329140242189e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 397, + "tokens/total": 12014608, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 181716 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.1685003638267517, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0024524193722754717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00246, + "step": 398, + "tokens/total": 12045088, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 182152 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.11644069850444794, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.001032137661240995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00103, + "step": 399, + "tokens/total": 12075280, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 182623 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004440188407897949, + "learning_rate": 2.1435742029681725e-05, + "loss": 7.53562999307178e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105792, + "tokens/train_per_sec_per_gpu": 26.92, + "tokens/trainable": 183039 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.06607124954462051, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0008609758224338293, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00086, + "step": 401, + "tokens/total": 12135680, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 183487 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.0036999934818595648, + "learning_rate": 2.105182715082638e-05, + "loss": 6.075722922105342e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 402, + "tokens/total": 12165920, + "tokens/train_per_sec_per_gpu": 39.03, + "tokens/trainable": 183987 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.010572529397904873, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0001372473343508318, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00014, + "step": 403, + "tokens/total": 12196080, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 184459 + }, + { + "epoch": 1.578125, + "grad_norm": 0.3721093237400055, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.003381735645234585, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00339, + "step": 404, + "tokens/total": 12226128, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 184887 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.08975216001272202, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0008267756784334779, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00083, + "step": 405, + "tokens/total": 12256176, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 185320 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.4965755343437195, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.012792397290468216, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01287, + "step": 406, + "tokens/total": 12286416, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185774 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.0017864442197605968, + "learning_rate": 2.011689980574966e-05, + "loss": 3.962808841606602e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 407, + "tokens/total": 12316720, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 186233 + }, + { + "epoch": 1.59375, + "grad_norm": 0.006781951989978552, + "learning_rate": 1.993423839463052e-05, + "loss": 6.552625563926995e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12347072, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 186671 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.02034713886678219, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00025357812410220504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00025, + "step": 409, + "tokens/total": 12377504, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 187102 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.003248689230531454, + "learning_rate": 1.957330080137385e-05, + "loss": 5.506081652129069e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 410, + "tokens/total": 12408048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 187549 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.16686522960662842, + "learning_rate": 1.9395039639322864e-05, + "loss": 0.0007388527155853808, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00074, + "step": 411, + "tokens/total": 12438256, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 187990 + }, + { + "epoch": 1.609375, + "grad_norm": 0.005520866252481937, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.00011023284605471417, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 412, + "tokens/total": 12466608, + "tokens/train_per_sec_per_gpu": 39.21, + "tokens/trainable": 188426 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.1133890226483345, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0016134778270497918, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 413, + "tokens/total": 12497136, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 188877 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.09548613429069519, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.00023090933973435313, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00023, + "step": 414, + "tokens/total": 12527120, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 189346 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0033725884277373552, + "learning_rate": 1.869688492349885e-05, + "loss": 4.779225855600089e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00005, + "step": 415, + "tokens/total": 12557248, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 189759 + }, + { + "epoch": 1.625, + "grad_norm": 0.0031356913968920708, + "learning_rate": 1.85261050441233e-05, + "loss": 4.390432150103152e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 416, + "tokens/total": 12587728, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 190225 + } + ], + "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": 8.544026588199736e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-416/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..daa9f8890442dcd3f116ff582a47d4d36ad4d765 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a279c5bd2f0a48ab2686d389d172094097e6ce39537831fdd4584d5ce2426262 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..93624cad36a5c060ee253919fcd09f4c81b598c6 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dd6d93b933b286e10d27546f816090842f90583619a559a1d4a34e74de48bfff +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..7780e878f7b769c0ef5a0a7ca54f5eb349c6517b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ba699bdea782fd2e8a001c27eb6fa2ca27017d8663d424f40c8c3b86434cb26 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2045249a1135f9f2d87c87c3e02e6227697d0cfa --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:457821a0c6da6ac211fae3b76339963abddd2e9839d13c28173a0924fca59503 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..9fe7d74ed1255ff810a443f60519fc1de1ff028b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/tokens_state.json @@ -0,0 +1 @@ +{"total": 13558752, "trainable": 204868} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..cc85ecd9d566846f6be9d18a15c8ed563d11eef3 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/trainer_state.json @@ -0,0 +1,6306 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.75, + "eval_steps": 500, + "global_step": 448, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.015219186432659626, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00014846389240119606, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 353, + "tokens/total": 10679184, + "tokens/train_per_sec_per_gpu": 29.27, + "tokens/trainable": 161567 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.002392370719462633, + "learning_rate": 3.166726850239794e-05, + "loss": 2.664637577254325e-05, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 354, + "tokens/total": 10709648, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 162032 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.012528779916465282, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00010976240446325392, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00011, + "step": 355, + "tokens/total": 10740128, + "tokens/train_per_sec_per_gpu": 32.56, + "tokens/trainable": 162460 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0019507119432091713, + "learning_rate": 3.11727834939056e-05, + "loss": 2.361126826144755e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00002, + "step": 356, + "tokens/total": 10770128, + "tokens/train_per_sec_per_gpu": 28.61, + "tokens/trainable": 162875 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.013965644873678684, + "learning_rate": 3.092702317708967e-05, + "loss": 0.00010500443022465333, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 357, + "tokens/total": 10800432, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 163329 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.08449137210845947, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00026978645473718643, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00027, + "step": 358, + "tokens/total": 10830592, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 163778 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.07536502927541733, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.001174358418211341, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00118, + "step": 359, + "tokens/total": 10861088, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 164219 + }, + { + "epoch": 1.40625, + "grad_norm": 0.024741269648075104, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00027250737184658647, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00027, + "step": 360, + "tokens/total": 10891552, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 164666 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.6777807474136353, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.016698025166988373, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01684, + "step": 361, + "tokens/total": 10921776, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 165128 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.2795168161392212, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00555866863578558, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00557, + "step": 362, + "tokens/total": 10952288, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 165614 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.19862814247608185, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.003639942966401577, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00365, + "step": 363, + "tokens/total": 10982496, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 166062 + }, + { + "epoch": 1.421875, + "grad_norm": 0.020105330273509026, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0001461556676076725, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 364, + "tokens/total": 11012704, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 166513 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.09536039084196091, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.0006832765066064894, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00068, + "step": 365, + "tokens/total": 11043024, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 167001 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.015557884238660336, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00021839249529875815, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 366, + "tokens/total": 11073392, + "tokens/train_per_sec_per_gpu": 30.62, + "tokens/trainable": 167450 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.003135454375296831, + "learning_rate": 2.8526184124692883e-05, + "loss": 4.49724102509208e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 367, + "tokens/total": 11103936, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 167895 + }, + { + "epoch": 1.4375, + "grad_norm": 0.03941775858402252, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0003685755073092878, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00037, + "step": 368, + "tokens/total": 11134256, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168338 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.07836400717496872, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.0015341172693297267, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00154, + "step": 369, + "tokens/total": 11164768, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 168813 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.47108760476112366, + "learning_rate": 2.782696506053033e-05, + "loss": 0.009904367849230766, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00995, + "step": 370, + "tokens/total": 11195136, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 169315 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.03302016109228134, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0005742749781347811, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00057, + "step": 371, + "tokens/total": 11225680, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 169756 + }, + { + "epoch": 1.453125, + "grad_norm": 0.012702704407274723, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.0001972567697521299, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 372, + "tokens/total": 11256112, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 170222 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.005610617808997631, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.283223723992705e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00008, + "step": 373, + "tokens/total": 11286304, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 170687 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.13766004145145416, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00119793601334095, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0012, + "step": 374, + "tokens/total": 11316800, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 171163 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.4246460795402527, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.007628333754837513, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00766, + "step": 375, + "tokens/total": 11346944, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 171626 + }, + { + "epoch": 1.46875, + "grad_norm": 0.008275235071778297, + "learning_rate": 2.645931522709877e-05, + "loss": 0.00011664340127026662, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00012, + "step": 376, + "tokens/total": 11377376, + "tokens/train_per_sec_per_gpu": 28.49, + "tokens/trainable": 172043 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.01685458980500698, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.00018345039279665798, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00018, + "step": 377, + "tokens/total": 11407680, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 172519 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.33050504326820374, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.003047082107514143, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00305, + "step": 378, + "tokens/total": 11437712, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 172977 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.1227385401725769, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0010191744659096003, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00102, + "step": 379, + "tokens/total": 11468176, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 173461 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0017978748073801398, + "learning_rate": 2.557117581955798e-05, + "loss": 4.36538175563328e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00004, + "step": 380, + "tokens/total": 11498352, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 173926 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.2067205160856247, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.005216958001255989, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00523, + "step": 381, + "tokens/total": 11528768, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 174397 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.0806688517332077, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0011719370959326625, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 382, + "tokens/total": 11558976, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 174852 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.027668939903378487, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00021959877631161362, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 383, + "tokens/total": 11589520, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 175293 + }, + { + "epoch": 1.5, + "grad_norm": 0.12873099744319916, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0015978836454451084, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0016, + "step": 384, + "tokens/total": 11619760, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 175731 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.037335313856601715, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0005178801948204637, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00052, + "step": 385, + "tokens/total": 11650192, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 176215 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.06761232763528824, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0005836813943460584, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00058, + "step": 386, + "tokens/total": 11680672, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 176663 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0501842126250267, + "learning_rate": 2.406442693028651e-05, + "loss": 0.0005241117323748767, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00052, + "step": 387, + "tokens/total": 11710880, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 177127 + }, + { + "epoch": 1.515625, + "grad_norm": 0.08328797668218613, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00114786671474576, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00115, + "step": 388, + "tokens/total": 11741408, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 177613 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.020661214366555214, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00030904496088624, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00031, + "step": 389, + "tokens/total": 11771712, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 178062 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.13835693895816803, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0014695629943162203, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00147, + "step": 390, + "tokens/total": 11801872, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 178535 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.061385851353406906, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0007466091774404049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00075, + "step": 391, + "tokens/total": 11832480, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 179004 + }, + { + "epoch": 1.53125, + "grad_norm": 0.003179469145834446, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.848719956702553e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 392, + "tokens/total": 11862672, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 179461 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.021778129041194916, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023812934523448348, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893024, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 179893 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.10550826787948608, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0007996093481779099, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0008, + "step": 394, + "tokens/total": 11923328, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 180344 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.003161423373967409, + "learning_rate": 2.2419829946830123e-05, + "loss": 5.3679388656746596e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 395, + "tokens/total": 11953792, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 180757 + }, + { + "epoch": 1.546875, + "grad_norm": 0.05268798768520355, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0009187876130454242, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00092, + "step": 396, + "tokens/total": 11984208, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 181228 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.0017847128910943866, + "learning_rate": 2.202206960760984e-05, + "loss": 3.819329140242189e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 397, + "tokens/total": 12014608, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 181716 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.1685003638267517, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0024524193722754717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00246, + "step": 398, + "tokens/total": 12045088, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 182152 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.11644069850444794, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.001032137661240995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00103, + "step": 399, + "tokens/total": 12075280, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 182623 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004440188407897949, + "learning_rate": 2.1435742029681725e-05, + "loss": 7.53562999307178e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105792, + "tokens/train_per_sec_per_gpu": 26.92, + "tokens/trainable": 183039 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.06607124954462051, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0008609758224338293, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00086, + "step": 401, + "tokens/total": 12135680, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 183487 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.0036999934818595648, + "learning_rate": 2.105182715082638e-05, + "loss": 6.075722922105342e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 402, + "tokens/total": 12165920, + "tokens/train_per_sec_per_gpu": 39.03, + "tokens/trainable": 183987 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.010572529397904873, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0001372473343508318, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00014, + "step": 403, + "tokens/total": 12196080, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 184459 + }, + { + "epoch": 1.578125, + "grad_norm": 0.3721093237400055, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.003381735645234585, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00339, + "step": 404, + "tokens/total": 12226128, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 184887 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.08975216001272202, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0008267756784334779, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00083, + "step": 405, + "tokens/total": 12256176, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 185320 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.4965755343437195, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.012792397290468216, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01287, + "step": 406, + "tokens/total": 12286416, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185774 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.0017864442197605968, + "learning_rate": 2.011689980574966e-05, + "loss": 3.962808841606602e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 407, + "tokens/total": 12316720, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 186233 + }, + { + "epoch": 1.59375, + "grad_norm": 0.006781951989978552, + "learning_rate": 1.993423839463052e-05, + "loss": 6.552625563926995e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12347072, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 186671 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.02034713886678219, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00025357812410220504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00025, + "step": 409, + "tokens/total": 12377504, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 187102 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.003248689230531454, + "learning_rate": 1.957330080137385e-05, + "loss": 5.506081652129069e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 410, + "tokens/total": 12408048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 187549 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.16686522960662842, + "learning_rate": 1.9395039639322864e-05, + "loss": 0.0007388527155853808, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00074, + "step": 411, + "tokens/total": 12438256, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 187990 + }, + { + "epoch": 1.609375, + "grad_norm": 0.005520866252481937, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.00011023284605471417, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 412, + "tokens/total": 12466608, + "tokens/train_per_sec_per_gpu": 39.21, + "tokens/trainable": 188426 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.1133890226483345, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0016134778270497918, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 413, + "tokens/total": 12497136, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 188877 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.09548613429069519, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.00023090933973435313, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00023, + "step": 414, + "tokens/total": 12527120, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 189346 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0033725884277373552, + "learning_rate": 1.869688492349885e-05, + "loss": 4.779225855600089e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00005, + "step": 415, + "tokens/total": 12557248, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 189759 + }, + { + "epoch": 1.625, + "grad_norm": 0.0031356913968920708, + "learning_rate": 1.85261050441233e-05, + "loss": 4.390432150103152e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 416, + "tokens/total": 12587728, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 190225 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0025502180214971304, + "learning_rate": 1.8356842992397304e-05, + "loss": 5.062075433670543e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00005, + "step": 417, + "tokens/total": 12618160, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 190717 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.006292750593274832, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.369729377795011e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00009, + "step": 418, + "tokens/total": 12648672, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 191135 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.1299043893814087, + "learning_rate": 1.802290048317732e-05, + "loss": 0.002147798193618655, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00215, + "step": 419, + "tokens/total": 12678944, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 191559 + }, + { + "epoch": 1.640625, + "grad_norm": 0.017163539305329323, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0001932488230522722, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 420, + "tokens/total": 12709408, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 192033 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.0045894598588347435, + "learning_rate": 1.7695112982104225e-05, + "loss": 5.246631917543709e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 421, + "tokens/total": 12739632, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 192482 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.048289310187101364, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0002731278946157545, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00027, + "step": 422, + "tokens/total": 12770224, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 192937 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.033917587250471115, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0003945066418964416, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 423, + "tokens/total": 12800640, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 193444 + }, + { + "epoch": 1.65625, + "grad_norm": 0.011636070907115936, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00016518094344064593, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00017, + "step": 424, + "tokens/total": 12831104, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 193925 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.06534689664840698, + "learning_rate": 1.705822021773101e-05, + "loss": 0.0008454776252619922, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 425, + "tokens/total": 12861296, + "tokens/train_per_sec_per_gpu": 29.68, + "tokens/trainable": 194342 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.07199558615684509, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0007431323174387217, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00074, + "step": 426, + "tokens/total": 12891504, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 194830 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.030936896800994873, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00027934886747971177, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00028, + "step": 427, + "tokens/total": 12921840, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 195296 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0030303315725177526, + "learning_rate": 1.659710580175893e-05, + "loss": 4.713955058832653e-05, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12952224, + "tokens/train_per_sec_per_gpu": 32.33, + "tokens/trainable": 195714 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.055711835622787476, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0008513920474797487, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00085, + "step": 429, + "tokens/total": 12982416, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 196179 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.014902536757290363, + "learning_rate": 1.629767603613508e-05, + "loss": 0.00024280000070575625, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00024, + "step": 430, + "tokens/total": 13012800, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 196660 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.003592823399230838, + "learning_rate": 1.615037389740547e-05, + "loss": 5.272522321320139e-05, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.00005, + "step": 431, + "tokens/total": 13043392, + "tokens/train_per_sec_per_gpu": 31.51, + "tokens/trainable": 197091 + }, + { + "epoch": 1.6875, + "grad_norm": 0.03395218774676323, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003875850234180689, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 432, + "tokens/total": 13073984, + "tokens/train_per_sec_per_gpu": 39.07, + "tokens/trainable": 197589 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.06874402612447739, + "learning_rate": 1.5860625757072092e-05, + "loss": 0.0007863138453103602, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00079, + "step": 433, + "tokens/total": 13104512, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 198047 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04059433192014694, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00044966390123590827, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00045, + "step": 434, + "tokens/total": 13134656, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 198465 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.1531708687543869, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0016073889564722776, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00161, + "step": 435, + "tokens/total": 13165248, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 198937 + }, + { + "epoch": 1.703125, + "grad_norm": 0.03812495246529579, + "learning_rate": 1.543823381344311e-05, + "loss": 0.0005045874277129769, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0005, + "step": 436, + "tokens/total": 13195648, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 199385 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.3262580633163452, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008203462697565556, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00824, + "step": 437, + "tokens/total": 13226176, + "tokens/train_per_sec_per_gpu": 34.78, + "tokens/trainable": 199869 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0010299552232027054, + "learning_rate": 1.5164861051607254e-05, + "loss": 1.8928842109744437e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 438, + "tokens/total": 13256480, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 200377 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.02154897339642048, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.00018170180555898696, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00018, + "step": 439, + "tokens/total": 13286528, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 200824 + }, + { + "epoch": 1.71875, + "grad_norm": 0.02101851999759674, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00032509121228940785, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00033, + "step": 440, + "tokens/total": 13317024, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 201297 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.007308815140277147, + "learning_rate": 1.476724846845306e-05, + "loss": 0.00010265046876156703, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 441, + "tokens/total": 13347440, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 201764 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.04583185538649559, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0005665539065375924, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00057, + "step": 442, + "tokens/total": 13377664, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 202204 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.2436366230249405, + "learning_rate": 1.451053546535705e-05, + "loss": 0.004352548159658909, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00436, + "step": 443, + "tokens/total": 13408080, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 202642 + }, + { + "epoch": 1.734375, + "grad_norm": 0.015320447273552418, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0002279828186146915, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00023, + "step": 444, + "tokens/total": 13437824, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 203050 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0014951409539207816, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.6277737561031245e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 445, + "tokens/total": 13468032, + "tokens/train_per_sec_per_gpu": 34.63, + "tokens/trainable": 203509 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009211640805006027, + "learning_rate": 1.413811586531508e-05, + "loss": 6.653230229858309e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00007, + "step": 446, + "tokens/total": 13498096, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 203976 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.15356089174747467, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0011966564925387502, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0012, + "step": 447, + "tokens/total": 13528288, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 204412 + }, + { + "epoch": 1.75, + "grad_norm": 0.008474756963551044, + "learning_rate": 1.3898329670629645e-05, + "loss": 9.093034896068275e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 448, + "tokens/total": 13558752, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 204868 + } + ], + "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": 9.203117321156475e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-448/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..5983b130634d903229e456888e225781d2ee1c7f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22ac8b2b52585ed380e1f2aa30b6bb208f1f4a2a3aa592010c7f62607726c8bb +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..54c6e2cbafa396a1ae641d1fcf4bf0aa000cd09f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:353dcf0b0448f14f813bae9f64cd704699a1d7c1ae42be01eb17b23acbd2409c +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..86c2a2a9d7c6dda7ea34b7dae9932816a5489705 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ffba478bd77760378b05056d94bb66357c9e20d58141b04bc1161a1024df90b +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..ce1c4e63a6c6f8be8c137d3add4eda184d59a043 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b68903e42777cffa699a4ceb5e57a17a53fd50d8cdf046737093de36719dc01 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e66cde141413ff27e7f5aaab5419ac67ee073645 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/tokens_state.json @@ -0,0 +1 @@ +{"total": 14524000, "trainable": 219387} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..558367fde6cb9fe1c849c602ac41109faa1370f6 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/trainer_state.json @@ -0,0 +1,6754 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.875, + "eval_steps": 500, + "global_step": 480, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.015219186432659626, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00014846389240119606, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 353, + "tokens/total": 10679184, + "tokens/train_per_sec_per_gpu": 29.27, + "tokens/trainable": 161567 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.002392370719462633, + "learning_rate": 3.166726850239794e-05, + "loss": 2.664637577254325e-05, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 354, + "tokens/total": 10709648, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 162032 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.012528779916465282, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00010976240446325392, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00011, + "step": 355, + "tokens/total": 10740128, + "tokens/train_per_sec_per_gpu": 32.56, + "tokens/trainable": 162460 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0019507119432091713, + "learning_rate": 3.11727834939056e-05, + "loss": 2.361126826144755e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00002, + "step": 356, + "tokens/total": 10770128, + "tokens/train_per_sec_per_gpu": 28.61, + "tokens/trainable": 162875 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.013965644873678684, + "learning_rate": 3.092702317708967e-05, + "loss": 0.00010500443022465333, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 357, + "tokens/total": 10800432, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 163329 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.08449137210845947, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00026978645473718643, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00027, + "step": 358, + "tokens/total": 10830592, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 163778 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.07536502927541733, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.001174358418211341, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00118, + "step": 359, + "tokens/total": 10861088, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 164219 + }, + { + "epoch": 1.40625, + "grad_norm": 0.024741269648075104, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00027250737184658647, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00027, + "step": 360, + "tokens/total": 10891552, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 164666 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.6777807474136353, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.016698025166988373, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01684, + "step": 361, + "tokens/total": 10921776, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 165128 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.2795168161392212, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00555866863578558, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00557, + "step": 362, + "tokens/total": 10952288, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 165614 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.19862814247608185, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.003639942966401577, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00365, + "step": 363, + "tokens/total": 10982496, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 166062 + }, + { + "epoch": 1.421875, + "grad_norm": 0.020105330273509026, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0001461556676076725, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 364, + "tokens/total": 11012704, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 166513 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.09536039084196091, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.0006832765066064894, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00068, + "step": 365, + "tokens/total": 11043024, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 167001 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.015557884238660336, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00021839249529875815, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 366, + "tokens/total": 11073392, + "tokens/train_per_sec_per_gpu": 30.62, + "tokens/trainable": 167450 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.003135454375296831, + "learning_rate": 2.8526184124692883e-05, + "loss": 4.49724102509208e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 367, + "tokens/total": 11103936, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 167895 + }, + { + "epoch": 1.4375, + "grad_norm": 0.03941775858402252, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0003685755073092878, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00037, + "step": 368, + "tokens/total": 11134256, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168338 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.07836400717496872, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.0015341172693297267, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00154, + "step": 369, + "tokens/total": 11164768, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 168813 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.47108760476112366, + "learning_rate": 2.782696506053033e-05, + "loss": 0.009904367849230766, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00995, + "step": 370, + "tokens/total": 11195136, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 169315 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.03302016109228134, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0005742749781347811, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00057, + "step": 371, + "tokens/total": 11225680, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 169756 + }, + { + "epoch": 1.453125, + "grad_norm": 0.012702704407274723, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.0001972567697521299, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 372, + "tokens/total": 11256112, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 170222 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.005610617808997631, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.283223723992705e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00008, + "step": 373, + "tokens/total": 11286304, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 170687 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.13766004145145416, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00119793601334095, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0012, + "step": 374, + "tokens/total": 11316800, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 171163 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.4246460795402527, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.007628333754837513, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00766, + "step": 375, + "tokens/total": 11346944, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 171626 + }, + { + "epoch": 1.46875, + "grad_norm": 0.008275235071778297, + "learning_rate": 2.645931522709877e-05, + "loss": 0.00011664340127026662, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00012, + "step": 376, + "tokens/total": 11377376, + "tokens/train_per_sec_per_gpu": 28.49, + "tokens/trainable": 172043 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.01685458980500698, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.00018345039279665798, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00018, + "step": 377, + "tokens/total": 11407680, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 172519 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.33050504326820374, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.003047082107514143, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00305, + "step": 378, + "tokens/total": 11437712, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 172977 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.1227385401725769, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0010191744659096003, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00102, + "step": 379, + "tokens/total": 11468176, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 173461 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0017978748073801398, + "learning_rate": 2.557117581955798e-05, + "loss": 4.36538175563328e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00004, + "step": 380, + "tokens/total": 11498352, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 173926 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.2067205160856247, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.005216958001255989, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00523, + "step": 381, + "tokens/total": 11528768, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 174397 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.0806688517332077, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0011719370959326625, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 382, + "tokens/total": 11558976, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 174852 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.027668939903378487, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00021959877631161362, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 383, + "tokens/total": 11589520, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 175293 + }, + { + "epoch": 1.5, + "grad_norm": 0.12873099744319916, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0015978836454451084, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0016, + "step": 384, + "tokens/total": 11619760, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 175731 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.037335313856601715, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0005178801948204637, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00052, + "step": 385, + "tokens/total": 11650192, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 176215 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.06761232763528824, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0005836813943460584, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00058, + "step": 386, + "tokens/total": 11680672, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 176663 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0501842126250267, + "learning_rate": 2.406442693028651e-05, + "loss": 0.0005241117323748767, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00052, + "step": 387, + "tokens/total": 11710880, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 177127 + }, + { + "epoch": 1.515625, + "grad_norm": 0.08328797668218613, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00114786671474576, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00115, + "step": 388, + "tokens/total": 11741408, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 177613 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.020661214366555214, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00030904496088624, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00031, + "step": 389, + "tokens/total": 11771712, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 178062 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.13835693895816803, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0014695629943162203, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00147, + "step": 390, + "tokens/total": 11801872, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 178535 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.061385851353406906, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0007466091774404049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00075, + "step": 391, + "tokens/total": 11832480, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 179004 + }, + { + "epoch": 1.53125, + "grad_norm": 0.003179469145834446, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.848719956702553e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 392, + "tokens/total": 11862672, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 179461 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.021778129041194916, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023812934523448348, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893024, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 179893 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.10550826787948608, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0007996093481779099, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0008, + "step": 394, + "tokens/total": 11923328, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 180344 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.003161423373967409, + "learning_rate": 2.2419829946830123e-05, + "loss": 5.3679388656746596e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 395, + "tokens/total": 11953792, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 180757 + }, + { + "epoch": 1.546875, + "grad_norm": 0.05268798768520355, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0009187876130454242, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00092, + "step": 396, + "tokens/total": 11984208, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 181228 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.0017847128910943866, + "learning_rate": 2.202206960760984e-05, + "loss": 3.819329140242189e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 397, + "tokens/total": 12014608, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 181716 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.1685003638267517, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0024524193722754717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00246, + "step": 398, + "tokens/total": 12045088, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 182152 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.11644069850444794, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.001032137661240995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00103, + "step": 399, + "tokens/total": 12075280, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 182623 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004440188407897949, + "learning_rate": 2.1435742029681725e-05, + "loss": 7.53562999307178e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105792, + "tokens/train_per_sec_per_gpu": 26.92, + "tokens/trainable": 183039 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.06607124954462051, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0008609758224338293, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00086, + "step": 401, + "tokens/total": 12135680, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 183487 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.0036999934818595648, + "learning_rate": 2.105182715082638e-05, + "loss": 6.075722922105342e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 402, + "tokens/total": 12165920, + "tokens/train_per_sec_per_gpu": 39.03, + "tokens/trainable": 183987 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.010572529397904873, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0001372473343508318, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00014, + "step": 403, + "tokens/total": 12196080, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 184459 + }, + { + "epoch": 1.578125, + "grad_norm": 0.3721093237400055, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.003381735645234585, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00339, + "step": 404, + "tokens/total": 12226128, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 184887 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.08975216001272202, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0008267756784334779, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00083, + "step": 405, + "tokens/total": 12256176, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 185320 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.4965755343437195, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.012792397290468216, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01287, + "step": 406, + "tokens/total": 12286416, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185774 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.0017864442197605968, + "learning_rate": 2.011689980574966e-05, + "loss": 3.962808841606602e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 407, + "tokens/total": 12316720, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 186233 + }, + { + "epoch": 1.59375, + "grad_norm": 0.006781951989978552, + "learning_rate": 1.993423839463052e-05, + "loss": 6.552625563926995e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12347072, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 186671 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.02034713886678219, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00025357812410220504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00025, + "step": 409, + "tokens/total": 12377504, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 187102 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.003248689230531454, + "learning_rate": 1.957330080137385e-05, + "loss": 5.506081652129069e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 410, + "tokens/total": 12408048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 187549 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.16686522960662842, + "learning_rate": 1.9395039639322864e-05, + "loss": 0.0007388527155853808, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00074, + "step": 411, + "tokens/total": 12438256, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 187990 + }, + { + "epoch": 1.609375, + "grad_norm": 0.005520866252481937, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.00011023284605471417, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 412, + "tokens/total": 12466608, + "tokens/train_per_sec_per_gpu": 39.21, + "tokens/trainable": 188426 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.1133890226483345, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0016134778270497918, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 413, + "tokens/total": 12497136, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 188877 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.09548613429069519, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.00023090933973435313, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00023, + "step": 414, + "tokens/total": 12527120, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 189346 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0033725884277373552, + "learning_rate": 1.869688492349885e-05, + "loss": 4.779225855600089e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00005, + "step": 415, + "tokens/total": 12557248, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 189759 + }, + { + "epoch": 1.625, + "grad_norm": 0.0031356913968920708, + "learning_rate": 1.85261050441233e-05, + "loss": 4.390432150103152e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 416, + "tokens/total": 12587728, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 190225 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0025502180214971304, + "learning_rate": 1.8356842992397304e-05, + "loss": 5.062075433670543e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00005, + "step": 417, + "tokens/total": 12618160, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 190717 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.006292750593274832, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.369729377795011e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00009, + "step": 418, + "tokens/total": 12648672, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 191135 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.1299043893814087, + "learning_rate": 1.802290048317732e-05, + "loss": 0.002147798193618655, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00215, + "step": 419, + "tokens/total": 12678944, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 191559 + }, + { + "epoch": 1.640625, + "grad_norm": 0.017163539305329323, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0001932488230522722, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 420, + "tokens/total": 12709408, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 192033 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.0045894598588347435, + "learning_rate": 1.7695112982104225e-05, + "loss": 5.246631917543709e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 421, + "tokens/total": 12739632, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 192482 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.048289310187101364, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0002731278946157545, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00027, + "step": 422, + "tokens/total": 12770224, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 192937 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.033917587250471115, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0003945066418964416, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 423, + "tokens/total": 12800640, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 193444 + }, + { + "epoch": 1.65625, + "grad_norm": 0.011636070907115936, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00016518094344064593, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00017, + "step": 424, + "tokens/total": 12831104, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 193925 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.06534689664840698, + "learning_rate": 1.705822021773101e-05, + "loss": 0.0008454776252619922, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 425, + "tokens/total": 12861296, + "tokens/train_per_sec_per_gpu": 29.68, + "tokens/trainable": 194342 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.07199558615684509, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0007431323174387217, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00074, + "step": 426, + "tokens/total": 12891504, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 194830 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.030936896800994873, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00027934886747971177, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00028, + "step": 427, + "tokens/total": 12921840, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 195296 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0030303315725177526, + "learning_rate": 1.659710580175893e-05, + "loss": 4.713955058832653e-05, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12952224, + "tokens/train_per_sec_per_gpu": 32.33, + "tokens/trainable": 195714 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.055711835622787476, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0008513920474797487, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00085, + "step": 429, + "tokens/total": 12982416, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 196179 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.014902536757290363, + "learning_rate": 1.629767603613508e-05, + "loss": 0.00024280000070575625, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00024, + "step": 430, + "tokens/total": 13012800, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 196660 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.003592823399230838, + "learning_rate": 1.615037389740547e-05, + "loss": 5.272522321320139e-05, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.00005, + "step": 431, + "tokens/total": 13043392, + "tokens/train_per_sec_per_gpu": 31.51, + "tokens/trainable": 197091 + }, + { + "epoch": 1.6875, + "grad_norm": 0.03395218774676323, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003875850234180689, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 432, + "tokens/total": 13073984, + "tokens/train_per_sec_per_gpu": 39.07, + "tokens/trainable": 197589 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.06874402612447739, + "learning_rate": 1.5860625757072092e-05, + "loss": 0.0007863138453103602, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00079, + "step": 433, + "tokens/total": 13104512, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 198047 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04059433192014694, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00044966390123590827, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00045, + "step": 434, + "tokens/total": 13134656, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 198465 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.1531708687543869, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0016073889564722776, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00161, + "step": 435, + "tokens/total": 13165248, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 198937 + }, + { + "epoch": 1.703125, + "grad_norm": 0.03812495246529579, + "learning_rate": 1.543823381344311e-05, + "loss": 0.0005045874277129769, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0005, + "step": 436, + "tokens/total": 13195648, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 199385 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.3262580633163452, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008203462697565556, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00824, + "step": 437, + "tokens/total": 13226176, + "tokens/train_per_sec_per_gpu": 34.78, + "tokens/trainable": 199869 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0010299552232027054, + "learning_rate": 1.5164861051607254e-05, + "loss": 1.8928842109744437e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 438, + "tokens/total": 13256480, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 200377 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.02154897339642048, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.00018170180555898696, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00018, + "step": 439, + "tokens/total": 13286528, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 200824 + }, + { + "epoch": 1.71875, + "grad_norm": 0.02101851999759674, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00032509121228940785, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00033, + "step": 440, + "tokens/total": 13317024, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 201297 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.007308815140277147, + "learning_rate": 1.476724846845306e-05, + "loss": 0.00010265046876156703, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 441, + "tokens/total": 13347440, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 201764 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.04583185538649559, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0005665539065375924, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00057, + "step": 442, + "tokens/total": 13377664, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 202204 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.2436366230249405, + "learning_rate": 1.451053546535705e-05, + "loss": 0.004352548159658909, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00436, + "step": 443, + "tokens/total": 13408080, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 202642 + }, + { + "epoch": 1.734375, + "grad_norm": 0.015320447273552418, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0002279828186146915, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00023, + "step": 444, + "tokens/total": 13437824, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 203050 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0014951409539207816, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.6277737561031245e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 445, + "tokens/total": 13468032, + "tokens/train_per_sec_per_gpu": 34.63, + "tokens/trainable": 203509 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009211640805006027, + "learning_rate": 1.413811586531508e-05, + "loss": 6.653230229858309e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00007, + "step": 446, + "tokens/total": 13498096, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 203976 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.15356089174747467, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0011966564925387502, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0012, + "step": 447, + "tokens/total": 13528288, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 204412 + }, + { + "epoch": 1.75, + "grad_norm": 0.008474756963551044, + "learning_rate": 1.3898329670629645e-05, + "loss": 9.093034896068275e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 448, + "tokens/total": 13558752, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 204868 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.015406596474349499, + "learning_rate": 1.3780999708818058e-05, + "loss": 0.0001275191898457706, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00013, + "step": 449, + "tokens/total": 13589136, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 205338 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.12964943051338196, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.0008908645249903202, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00089, + "step": 450, + "tokens/total": 13619472, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 205764 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.010256104171276093, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00014732024283148348, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 451, + "tokens/total": 13649872, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 206201 + }, + { + "epoch": 1.765625, + "grad_norm": 0.006175735965371132, + "learning_rate": 1.3439320741704075e-05, + "loss": 9.223521192325279e-05, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 452, + "tokens/total": 13680064, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 206657 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.060654036700725555, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.0005743277142755687, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00057, + "step": 453, + "tokens/total": 13710592, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 207138 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.003730883589014411, + "learning_rate": 1.3220174411609587e-05, + "loss": 5.224360211286694e-05, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 454, + "tokens/total": 13740960, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 207583 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.09384409338235855, + "learning_rate": 1.3113206928331471e-05, + "loss": 0.0008255710126832128, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00083, + "step": 455, + "tokens/total": 13771296, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 208025 + }, + { + "epoch": 1.78125, + "grad_norm": 0.11239906400442123, + "learning_rate": 1.300798252548806e-05, + "loss": 0.0013365334598347545, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00134, + "step": 456, + "tokens/total": 13801664, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 208453 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.007983297109603882, + "learning_rate": 1.2904505581896265e-05, + "loss": 6.189515988808125e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 457, + "tokens/total": 13831936, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 208947 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.0063696797005832195, + "learning_rate": 1.2802780403654082e-05, + "loss": 9.090925595955923e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00009, + "step": 458, + "tokens/total": 13862160, + "tokens/train_per_sec_per_gpu": 35.33, + "tokens/trainable": 209435 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.008154391311109066, + "learning_rate": 1.2702811223961408e-05, + "loss": 7.762061432003975e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00008, + "step": 459, + "tokens/total": 13892320, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 209885 + }, + { + "epoch": 1.796875, + "grad_norm": 0.015743529424071312, + "learning_rate": 1.2604602202943861e-05, + "loss": 0.00018653822189662606, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00019, + "step": 460, + "tokens/total": 13922752, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 210344 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.011715077795088291, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.00016241846606135368, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00016, + "step": 461, + "tokens/total": 13952816, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 210814 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.04297679290175438, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0005623188917525113, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00056, + "step": 462, + "tokens/total": 13981040, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 211238 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.004571136087179184, + "learning_rate": 1.2320576593470082e-05, + "loss": 4.7571604227414355e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 463, + "tokens/total": 14011248, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 211700 + }, + { + "epoch": 1.8125, + "grad_norm": 0.0022084659431129694, + "learning_rate": 1.2229448340928828e-05, + "loss": 1.850002445280552e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00002, + "step": 464, + "tokens/total": 14041504, + "tokens/train_per_sec_per_gpu": 29.47, + "tokens/trainable": 212138 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.08595507591962814, + "learning_rate": 1.2140099945624458e-05, + "loss": 0.0008459303062409163, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00085, + "step": 465, + "tokens/total": 14071728, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 212623 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0007654759101569653, + "learning_rate": 1.205253512570841e-05, + "loss": 1.6132753444253467e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00002, + "step": 466, + "tokens/total": 14101824, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 213062 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0005283593782223761, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.0958694474538788e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 467, + "tokens/total": 14132080, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 213501 + }, + { + "epoch": 1.828125, + "grad_norm": 0.0003768271126318723, + "learning_rate": 1.1882770713386095e-05, + "loss": 1.1069126230722759e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 468, + "tokens/total": 14162112, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 213942 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.010821828618645668, + "learning_rate": 1.180057818556998e-05, + "loss": 7.607591396663338e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00008, + "step": 469, + "tokens/total": 14192784, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 214410 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.008473122492432594, + "learning_rate": 1.1720183362028494e-05, + "loss": 0.00010939891217276454, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 470, + "tokens/total": 14221008, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 214844 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.0007591558387503028, + "learning_rate": 1.1641589588318387e-05, + "loss": 1.7881218809634447e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 471, + "tokens/total": 14251536, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 215336 + }, + { + "epoch": 1.84375, + "grad_norm": 0.05926339700818062, + "learning_rate": 1.1564800135047418e-05, + "loss": 0.0005082086427137256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00051, + "step": 472, + "tokens/total": 14281664, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 215777 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.12773646414279938, + "learning_rate": 1.148981819773816e-05, + "loss": 0.002721701515838504, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 473, + "tokens/total": 14312048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 216222 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.028139308094978333, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.00013170376769267023, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00013, + "step": 474, + "tokens/total": 14342112, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 216662 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.003854855429381132, + "learning_rate": 1.1345289276874717e-05, + "loss": 5.479517130879685e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 475, + "tokens/total": 14372176, + "tokens/train_per_sec_per_gpu": 31.26, + "tokens/trainable": 217068 + }, + { + "epoch": 1.859375, + "grad_norm": 0.0007855606963858008, + "learning_rate": 1.1275748307758873e-05, + "loss": 1.5755449567222968e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00002, + "step": 476, + "tokens/total": 14402592, + "tokens/train_per_sec_per_gpu": 38.21, + "tokens/trainable": 217558 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.004396429285407066, + "learning_rate": 1.1208026883231147e-05, + "loss": 5.772991426056251e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00006, + "step": 477, + "tokens/total": 14432816, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 218025 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.03181939572095871, + "learning_rate": 1.1142127821456433e-05, + "loss": 0.00036515528336167336, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00037, + "step": 478, + "tokens/total": 14463424, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 218498 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.004733108915388584, + "learning_rate": 1.1078053864763674e-05, + "loss": 4.52871281595435e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 479, + "tokens/total": 14493440, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 218946 + }, + { + "epoch": 1.875, + "grad_norm": 0.0026765780057758093, + "learning_rate": 1.1015807679531756e-05, + "loss": 3.31170012941584e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 480, + "tokens/total": 14524000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 219387 + } + ], + "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": 9.85828754537856e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-480/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..0f72ae23581071a26affd9e78eb4218588d15102 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c5b55d2b972cb95c7ac1f9b990f0200f755538382f699ff8db68276809b571fa +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..24621dc14ac549951e2e7615300178a008b9a878 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:623e28ce7e6c707ff84aa20683ab63515fb64ba836512961aaa1eb2158390016 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..1532de23e9f9e12eb874859ec3aeb02203d607f4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0aa6f06cb517ef29e346b941fccc40135c628920735642bfda469e00ed15de41 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..22325b3b3d08e0697784fe21b8c321485a51f7b5 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:697fe8894f7795467da8b1a7ccebf8570b28357dc457e2ba8ccdb525507cfef4 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e853992de2edf5fcb474e774f7cbd1e1e7b9a0b2 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/tokens_state.json @@ -0,0 +1 @@ +{"total": 15493248, "trainable": 234100} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3cc81a78a39dc00e8eb73a543c776e4d308de91c --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/trainer_state.json @@ -0,0 +1,7202 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 2.0, + "eval_steps": 500, + "global_step": 512, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + }, + { + "epoch": 0.37890625, + "grad_norm": 0.44522419571876526, + "learning_rate": 9.53619478457953e-05, + "loss": 0.0070595573633909225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00708, + "step": 97, + "tokens/total": 2935040, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 44295 + }, + { + "epoch": 0.3828125, + "grad_norm": 1.1640557050704956, + "learning_rate": 9.523275153154695e-05, + "loss": 0.06137344613671303, + "memory/device_reserved (GiB)": 35.19, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0633, + "step": 98, + "tokens/total": 2965360, + "tokens/train_per_sec_per_gpu": 39.11, + "tokens/trainable": 44786 + }, + { + "epoch": 0.38671875, + "grad_norm": 0.5285153985023499, + "learning_rate": 9.51018809682839e-05, + "loss": 0.005951949395239353, + "memory/device_reserved (GiB)": 35.76, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00597, + "step": 99, + "tokens/total": 2995952, + "tokens/train_per_sec_per_gpu": 36.34, + "tokens/trainable": 45249 + }, + { + "epoch": 0.390625, + "grad_norm": 0.40244120359420776, + "learning_rate": 9.49693416020645e-05, + "loss": 0.01774679310619831, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01791, + "step": 100, + "tokens/total": 3026320, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 45693 + }, + { + "epoch": 0.39453125, + "grad_norm": 0.3558945953845978, + "learning_rate": 9.483513894839276e-05, + "loss": 0.011360588483512402, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01143, + "step": 101, + "tokens/total": 3056560, + "tokens/train_per_sec_per_gpu": 33.5, + "tokens/trainable": 46136 + }, + { + "epoch": 0.3984375, + "grad_norm": 0.4544995129108429, + "learning_rate": 9.469927859198888e-05, + "loss": 0.0114982221275568, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01156, + "step": 102, + "tokens/total": 3087136, + "tokens/train_per_sec_per_gpu": 32.88, + "tokens/trainable": 46592 + }, + { + "epoch": 0.40234375, + "grad_norm": 0.42753517627716064, + "learning_rate": 9.456176618655689e-05, + "loss": 0.012740183621644974, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01282, + "step": 103, + "tokens/total": 3117712, + "tokens/train_per_sec_per_gpu": 32.13, + "tokens/trainable": 47059 + }, + { + "epoch": 0.40625, + "grad_norm": 0.990300714969635, + "learning_rate": 9.442260745454927e-05, + "loss": 0.03431129455566406, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03491, + "step": 104, + "tokens/total": 3147984, + "tokens/train_per_sec_per_gpu": 37.98, + "tokens/trainable": 47548 + }, + { + "epoch": 0.41015625, + "grad_norm": 0.275656521320343, + "learning_rate": 9.428180818692884e-05, + "loss": 0.010381961241364479, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01044, + "step": 105, + "tokens/total": 3178352, + "tokens/train_per_sec_per_gpu": 36.54, + "tokens/trainable": 48041 + }, + { + "epoch": 0.4140625, + "grad_norm": 0.38448187708854675, + "learning_rate": 9.413937424292791e-05, + "loss": 0.018732387572526932, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01891, + "step": 106, + "tokens/total": 3206608, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 48492 + }, + { + "epoch": 0.41796875, + "grad_norm": 0.6006481647491455, + "learning_rate": 9.399531154980424e-05, + "loss": 0.030589262023568153, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.03106, + "step": 107, + "tokens/total": 3236704, + "tokens/train_per_sec_per_gpu": 31.11, + "tokens/trainable": 48926 + }, + { + "epoch": 0.421875, + "grad_norm": 0.21727009117603302, + "learning_rate": 9.384962610259455e-05, + "loss": 0.013438574969768524, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01353, + "step": 108, + "tokens/total": 3267072, + "tokens/train_per_sec_per_gpu": 38.4, + "tokens/trainable": 49397 + }, + { + "epoch": 0.42578125, + "grad_norm": 0.22535420954227448, + "learning_rate": 9.370232396386494e-05, + "loss": 0.021586481481790543, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02182, + "step": 109, + "tokens/total": 3297200, + "tokens/train_per_sec_per_gpu": 37.26, + "tokens/trainable": 49864 + }, + { + "epoch": 0.4296875, + "grad_norm": 0.37579283118247986, + "learning_rate": 9.355341126345868e-05, + "loss": 0.02149398997426033, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.02173, + "step": 110, + "tokens/total": 3327616, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 50303 + }, + { + "epoch": 0.43359375, + "grad_norm": 0.2575673758983612, + "learning_rate": 9.340289419824107e-05, + "loss": 0.029223870486021042, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02966, + "step": 111, + "tokens/total": 3357824, + "tokens/train_per_sec_per_gpu": 33.8, + "tokens/trainable": 50746 + }, + { + "epoch": 0.4375, + "grad_norm": 0.346432626247406, + "learning_rate": 9.325077903184159e-05, + "loss": 0.022741328924894333, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.023, + "step": 112, + "tokens/total": 3387856, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 51192 + }, + { + "epoch": 0.44140625, + "grad_norm": 0.3252893090248108, + "learning_rate": 9.30970720943932e-05, + "loss": 0.012311354279518127, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01239, + "step": 113, + "tokens/total": 3418304, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 51648 + }, + { + "epoch": 0.4453125, + "grad_norm": 0.39176976680755615, + "learning_rate": 9.2941779782269e-05, + "loss": 0.012435214594006538, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01251, + "step": 114, + "tokens/total": 3448560, + "tokens/train_per_sec_per_gpu": 30.79, + "tokens/trainable": 52061 + }, + { + "epoch": 0.44921875, + "grad_norm": 0.18729569017887115, + "learning_rate": 9.278490855781596e-05, + "loss": 0.006094799842685461, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00611, + "step": 115, + "tokens/total": 3478928, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 52550 + }, + { + "epoch": 0.453125, + "grad_norm": 0.15628787875175476, + "learning_rate": 9.262646494908604e-05, + "loss": 0.0027563392650336027, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00276, + "step": 116, + "tokens/total": 3509232, + "tokens/train_per_sec_per_gpu": 38.98, + "tokens/trainable": 53051 + }, + { + "epoch": 0.45703125, + "grad_norm": 0.3185792863368988, + "learning_rate": 9.246645554956457e-05, + "loss": 0.006228582933545113, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00625, + "step": 117, + "tokens/total": 3539680, + "tokens/train_per_sec_per_gpu": 35.85, + "tokens/trainable": 53510 + }, + { + "epoch": 0.4609375, + "grad_norm": 0.4676451086997986, + "learning_rate": 9.230488701789578e-05, + "loss": 0.010252914391458035, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01031, + "step": 118, + "tokens/total": 3570320, + "tokens/train_per_sec_per_gpu": 33.24, + "tokens/trainable": 53966 + }, + { + "epoch": 0.46484375, + "grad_norm": 0.18005269765853882, + "learning_rate": 9.214176607760577e-05, + "loss": 0.0011447753058746457, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00115, + "step": 119, + "tokens/total": 3600384, + "tokens/train_per_sec_per_gpu": 35.02, + "tokens/trainable": 54399 + }, + { + "epoch": 0.46875, + "grad_norm": 0.42614251375198364, + "learning_rate": 9.197709951682268e-05, + "loss": 0.01829312928020954, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01846, + "step": 120, + "tokens/total": 3630672, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 54875 + }, + { + "epoch": 0.47265625, + "grad_norm": 0.06879743933677673, + "learning_rate": 9.181089418799428e-05, + "loss": 0.0005500271217897534, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00055, + "step": 121, + "tokens/total": 3660992, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 55294 + }, + { + "epoch": 0.4765625, + "grad_norm": 0.3329726457595825, + "learning_rate": 9.164315700760271e-05, + "loss": 0.009491996839642525, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00954, + "step": 122, + "tokens/total": 3691248, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 55734 + }, + { + "epoch": 0.48046875, + "grad_norm": 0.7758689522743225, + "learning_rate": 9.147389495587671e-05, + "loss": 0.013467054814100266, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.01356, + "step": 123, + "tokens/total": 3721248, + "tokens/train_per_sec_per_gpu": 33.68, + "tokens/trainable": 56178 + }, + { + "epoch": 0.484375, + "grad_norm": 0.591164231300354, + "learning_rate": 9.130311507650116e-05, + "loss": 0.011679252609610558, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01175, + "step": 124, + "tokens/total": 3751440, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 56623 + }, + { + "epoch": 0.48828125, + "grad_norm": 0.5725088119506836, + "learning_rate": 9.113082447632394e-05, + "loss": 0.00860346108675003, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00864, + "step": 125, + "tokens/total": 3781696, + "tokens/train_per_sec_per_gpu": 37.12, + "tokens/trainable": 57115 + }, + { + "epoch": 0.4921875, + "grad_norm": 0.4081246554851532, + "learning_rate": 9.09570303250602e-05, + "loss": 0.010806378908455372, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01086, + "step": 126, + "tokens/total": 3812096, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 57561 + }, + { + "epoch": 0.49609375, + "grad_norm": 0.6531717777252197, + "learning_rate": 9.078173985499394e-05, + "loss": 0.029426712542772293, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02986, + "step": 127, + "tokens/total": 3842720, + "tokens/train_per_sec_per_gpu": 35.26, + "tokens/trainable": 58039 + }, + { + "epoch": 0.5, + "grad_norm": 0.23760652542114258, + "learning_rate": 9.060496036067713e-05, + "loss": 0.009020479395985603, + "memory/device_reserved (GiB)": 35.99, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.00906, + "step": 128, + "tokens/total": 3871072, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 58488 + }, + { + "epoch": 0.50390625, + "grad_norm": 0.5177896022796631, + "learning_rate": 9.042669919862615e-05, + "loss": 0.013876531273126602, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01397, + "step": 129, + "tokens/total": 3901504, + "tokens/train_per_sec_per_gpu": 33.07, + "tokens/trainable": 58938 + }, + { + "epoch": 0.5078125, + "grad_norm": 0.2801472842693329, + "learning_rate": 9.024696378701557e-05, + "loss": 0.02102840505540371, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02125, + "step": 130, + "tokens/total": 3931968, + "tokens/train_per_sec_per_gpu": 37.3, + "tokens/trainable": 59406 + }, + { + "epoch": 0.51171875, + "grad_norm": 0.4326750636100769, + "learning_rate": 9.006576160536948e-05, + "loss": 0.024094708263874054, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.02439, + "step": 131, + "tokens/total": 3962048, + "tokens/train_per_sec_per_gpu": 32.69, + "tokens/trainable": 59842 + }, + { + "epoch": 0.515625, + "grad_norm": 0.16691909730434418, + "learning_rate": 8.988310019425035e-05, + "loss": 0.007350304163992405, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00738, + "step": 132, + "tokens/total": 3992512, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 60316 + }, + { + "epoch": 0.51953125, + "grad_norm": 0.2076587677001953, + "learning_rate": 8.969898715494506e-05, + "loss": 0.01380444597452879, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0139, + "step": 133, + "tokens/total": 4022864, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 60809 + }, + { + "epoch": 0.5234375, + "grad_norm": 0.21502523124217987, + "learning_rate": 8.951343014914869e-05, + "loss": 0.016695577651262283, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.39, + "memory/max_allocated (GiB)": 33.39, + "ppl": 1.01684, + "step": 134, + "tokens/total": 4051344, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 61254 + }, + { + "epoch": 0.52734375, + "grad_norm": 0.307210236787796, + "learning_rate": 8.932643689864568e-05, + "loss": 0.01635798066854477, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01649, + "step": 135, + "tokens/total": 4081728, + "tokens/train_per_sec_per_gpu": 36.41, + "tokens/trainable": 61725 + }, + { + "epoch": 0.53125, + "grad_norm": 0.2396101951599121, + "learning_rate": 8.913801518498845e-05, + "loss": 0.014302338473498821, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.7, + "memory/max_allocated (GiB)": 33.7, + "ppl": 1.01441, + "step": 136, + "tokens/total": 4111904, + "tokens/train_per_sec_per_gpu": 38.67, + "tokens/trainable": 62200 + }, + { + "epoch": 0.53515625, + "grad_norm": 0.1934954673051834, + "learning_rate": 8.894817284917364e-05, + "loss": 0.011029968038201332, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01109, + "step": 137, + "tokens/total": 4142208, + "tokens/train_per_sec_per_gpu": 36.9, + "tokens/trainable": 62698 + }, + { + "epoch": 0.5390625, + "grad_norm": 0.30764421820640564, + "learning_rate": 8.875691779131569e-05, + "loss": 0.020420396700501442, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.02063, + "step": 138, + "tokens/total": 4172512, + "tokens/train_per_sec_per_gpu": 37.02, + "tokens/trainable": 63158 + }, + { + "epoch": 0.54296875, + "grad_norm": 0.23006600141525269, + "learning_rate": 8.856425797031829e-05, + "loss": 0.007682272233068943, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00771, + "step": 139, + "tokens/total": 4202944, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 63638 + }, + { + "epoch": 0.546875, + "grad_norm": 0.2612818777561188, + "learning_rate": 8.837020140354295e-05, + "loss": 0.011739096604287624, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.42, + "memory/max_allocated (GiB)": 33.42, + "ppl": 1.01181, + "step": 140, + "tokens/total": 4231264, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 64055 + }, + { + "epoch": 0.55078125, + "grad_norm": 0.25878170132637024, + "learning_rate": 8.817475616647554e-05, + "loss": 0.008031821809709072, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00806, + "step": 141, + "tokens/total": 4261488, + "tokens/train_per_sec_per_gpu": 35.41, + "tokens/trainable": 64512 + }, + { + "epoch": 0.5546875, + "grad_norm": 0.11936136335134506, + "learning_rate": 8.797793039239017e-05, + "loss": 0.0028082651551812887, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00281, + "step": 142, + "tokens/total": 4292032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 64950 + }, + { + "epoch": 0.55859375, + "grad_norm": 0.17141154408454895, + "learning_rate": 8.777973227201069e-05, + "loss": 0.004208846017718315, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00422, + "step": 143, + "tokens/total": 4322032, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 65429 + }, + { + "epoch": 0.5625, + "grad_norm": 0.1641281396150589, + "learning_rate": 8.758017005316988e-05, + "loss": 0.0030169994570314884, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00302, + "step": 144, + "tokens/total": 4352128, + "tokens/train_per_sec_per_gpu": 37.17, + "tokens/trainable": 65895 + }, + { + "epoch": 0.56640625, + "grad_norm": 0.1339448094367981, + "learning_rate": 8.737925204046629e-05, + "loss": 0.0016135971527546644, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00161, + "step": 145, + "tokens/total": 4382368, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 66327 + }, + { + "epoch": 0.5703125, + "grad_norm": 0.5660213828086853, + "learning_rate": 8.717698659491851e-05, + "loss": 0.00862264446914196, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00866, + "step": 146, + "tokens/total": 4412640, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 66778 + }, + { + "epoch": 0.57421875, + "grad_norm": 0.42477667331695557, + "learning_rate": 8.697338213361735e-05, + "loss": 0.008355698548257351, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00839, + "step": 147, + "tokens/total": 4442896, + "tokens/train_per_sec_per_gpu": 35.67, + "tokens/trainable": 67192 + }, + { + "epoch": 0.578125, + "grad_norm": 0.40101227164268494, + "learning_rate": 8.676844712937552e-05, + "loss": 0.01053987629711628, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.0106, + "step": 148, + "tokens/total": 4473248, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 67677 + }, + { + "epoch": 0.58203125, + "grad_norm": 0.1790342479944229, + "learning_rate": 8.656219011037509e-05, + "loss": 0.0022937909234315157, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0023, + "step": 149, + "tokens/total": 4503664, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 68121 + }, + { + "epoch": 0.5859375, + "grad_norm": 0.49002259969711304, + "learning_rate": 8.63546196598125e-05, + "loss": 0.0021040919236838818, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00211, + "step": 150, + "tokens/total": 4533760, + "tokens/train_per_sec_per_gpu": 33.44, + "tokens/trainable": 68589 + }, + { + "epoch": 0.58984375, + "grad_norm": 0.635802686214447, + "learning_rate": 8.614574441554145e-05, + "loss": 0.02125491574406624, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.02148, + "step": 151, + "tokens/total": 4563872, + "tokens/train_per_sec_per_gpu": 30.28, + "tokens/trainable": 69055 + }, + { + "epoch": 0.59375, + "grad_norm": 0.8153690099716187, + "learning_rate": 8.593557306971349e-05, + "loss": 0.01092493161559105, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01098, + "step": 152, + "tokens/total": 4594384, + "tokens/train_per_sec_per_gpu": 35.77, + "tokens/trainable": 69501 + }, + { + "epoch": 0.59765625, + "grad_norm": 0.5217874646186829, + "learning_rate": 8.572411436841618e-05, + "loss": 0.009749074466526508, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0098, + "step": 153, + "tokens/total": 4624816, + "tokens/train_per_sec_per_gpu": 34.26, + "tokens/trainable": 69964 + }, + { + "epoch": 0.6015625, + "grad_norm": 0.2303430438041687, + "learning_rate": 8.551137711130922e-05, + "loss": 0.0031966331880539656, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0032, + "step": 154, + "tokens/total": 4655136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 70431 + }, + { + "epoch": 0.60546875, + "grad_norm": 1.3286867141723633, + "learning_rate": 8.529737015125824e-05, + "loss": 0.04919695854187012, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.05043, + "step": 155, + "tokens/total": 4683488, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 70875 + }, + { + "epoch": 0.609375, + "grad_norm": 0.6995537877082825, + "learning_rate": 8.508210239396639e-05, + "loss": 0.02087758481502533, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0211, + "step": 156, + "tokens/total": 4714032, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 71333 + }, + { + "epoch": 0.61328125, + "grad_norm": 0.2431785613298416, + "learning_rate": 8.486558279760375e-05, + "loss": 0.003255764953792095, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00326, + "step": 157, + "tokens/total": 4744192, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 71795 + }, + { + "epoch": 0.6171875, + "grad_norm": 0.5777634978294373, + "learning_rate": 8.464782037243449e-05, + "loss": 0.015976078808307648, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0161, + "step": 158, + "tokens/total": 4774592, + "tokens/train_per_sec_per_gpu": 32.59, + "tokens/trainable": 72231 + }, + { + "epoch": 0.62109375, + "grad_norm": 0.3740072548389435, + "learning_rate": 8.442882418044202e-05, + "loss": 0.015244726091623306, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.01536, + "step": 159, + "tokens/total": 4804672, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 72660 + }, + { + "epoch": 0.625, + "grad_norm": 0.5423697233200073, + "learning_rate": 8.420860333495179e-05, + "loss": 0.02623123675584793, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02658, + "step": 160, + "tokens/total": 4835056, + "tokens/train_per_sec_per_gpu": 31.94, + "tokens/trainable": 73115 + }, + { + "epoch": 0.62890625, + "grad_norm": 0.20633140206336975, + "learning_rate": 8.398716700025208e-05, + "loss": 0.009905166923999786, + "memory/device_reserved (GiB)": 37.89, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00995, + "step": 161, + "tokens/total": 4865296, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 73536 + }, + { + "epoch": 0.6328125, + "grad_norm": 0.2795397937297821, + "learning_rate": 8.376452439121266e-05, + "loss": 0.011865551583468914, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.01194, + "step": 162, + "tokens/total": 4895872, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 73985 + }, + { + "epoch": 0.63671875, + "grad_norm": 0.2866630256175995, + "learning_rate": 8.354068477290124e-05, + "loss": 0.012883953750133514, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01297, + "step": 163, + "tokens/total": 4926192, + "tokens/train_per_sec_per_gpu": 35.7, + "tokens/trainable": 74449 + }, + { + "epoch": 0.640625, + "grad_norm": 0.1895613968372345, + "learning_rate": 8.331565746019807e-05, + "loss": 0.010326865129172802, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01038, + "step": 164, + "tokens/total": 4956784, + "tokens/train_per_sec_per_gpu": 33.78, + "tokens/trainable": 74900 + }, + { + "epoch": 0.64453125, + "grad_norm": 0.27947714924812317, + "learning_rate": 8.308945181740812e-05, + "loss": 0.008086128160357475, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00812, + "step": 165, + "tokens/total": 4987120, + "tokens/train_per_sec_per_gpu": 30.91, + "tokens/trainable": 75324 + }, + { + "epoch": 0.6484375, + "grad_norm": 0.20866429805755615, + "learning_rate": 8.286207725787153e-05, + "loss": 0.0023722797632217407, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00238, + "step": 166, + "tokens/total": 5017424, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 75765 + }, + { + "epoch": 0.65234375, + "grad_norm": 0.16256311535835266, + "learning_rate": 8.263354324357182e-05, + "loss": 0.006408158224076033, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00643, + "step": 167, + "tokens/total": 5048096, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 76226 + }, + { + "epoch": 0.65625, + "grad_norm": 0.24209903180599213, + "learning_rate": 8.240385928474219e-05, + "loss": 0.015991341322660446, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.01612, + "step": 168, + "tokens/total": 5078608, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 76677 + }, + { + "epoch": 0.66015625, + "grad_norm": 0.09416277706623077, + "learning_rate": 8.217303493946967e-05, + "loss": 0.0014367344556376338, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00144, + "step": 169, + "tokens/total": 5108944, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 77109 + }, + { + "epoch": 0.6640625, + "grad_norm": 0.35095950961112976, + "learning_rate": 8.194107981329746e-05, + "loss": 0.018648192286491394, + "memory/device_reserved (GiB)": 36.01, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01882, + "step": 170, + "tokens/total": 5139344, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 77558 + }, + { + "epoch": 0.66796875, + "grad_norm": 0.5683367252349854, + "learning_rate": 8.170800355882518e-05, + "loss": 0.017304519191384315, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01746, + "step": 171, + "tokens/total": 5169600, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 78018 + }, + { + "epoch": 0.671875, + "grad_norm": 0.43827077746391296, + "learning_rate": 8.147381587530713e-05, + "loss": 0.00554292369633913, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00556, + "step": 172, + "tokens/total": 5200032, + "tokens/train_per_sec_per_gpu": 30.15, + "tokens/trainable": 78472 + }, + { + "epoch": 0.67578125, + "grad_norm": 0.3213059902191162, + "learning_rate": 8.123852650824877e-05, + "loss": 0.007639693561941385, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00767, + "step": 173, + "tokens/total": 5230464, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 78919 + }, + { + "epoch": 0.6796875, + "grad_norm": 0.237381711602211, + "learning_rate": 8.100214524900103e-05, + "loss": 0.005053661298006773, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.62, + "memory/max_allocated (GiB)": 33.62, + "ppl": 1.00507, + "step": 174, + "tokens/total": 5260192, + "tokens/train_per_sec_per_gpu": 30.4, + "tokens/trainable": 79336 + }, + { + "epoch": 0.68359375, + "grad_norm": 0.11023443192243576, + "learning_rate": 8.076468193435301e-05, + "loss": 0.0017536291852593422, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00176, + "step": 175, + "tokens/total": 5290480, + "tokens/train_per_sec_per_gpu": 32.91, + "tokens/trainable": 79806 + }, + { + "epoch": 0.6875, + "grad_norm": 0.5749457478523254, + "learning_rate": 8.052614644612253e-05, + "loss": 0.009171009063720703, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00921, + "step": 176, + "tokens/total": 5320992, + "tokens/train_per_sec_per_gpu": 36.4, + "tokens/trainable": 80284 + }, + { + "epoch": 0.69140625, + "grad_norm": 0.09884531795978546, + "learning_rate": 8.028654871074489e-05, + "loss": 0.0016485131345689297, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00165, + "step": 177, + "tokens/total": 5351328, + "tokens/train_per_sec_per_gpu": 32.5, + "tokens/trainable": 80731 + }, + { + "epoch": 0.6953125, + "grad_norm": 0.28975823521614075, + "learning_rate": 8.004589869885986e-05, + "loss": 0.0046661170199513435, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00468, + "step": 178, + "tokens/total": 5381536, + "tokens/train_per_sec_per_gpu": 34.93, + "tokens/trainable": 81199 + }, + { + "epoch": 0.69921875, + "grad_norm": 0.22287319600582123, + "learning_rate": 7.980420642489674e-05, + "loss": 0.0017163918819278479, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00172, + "step": 179, + "tokens/total": 5411536, + "tokens/train_per_sec_per_gpu": 33.18, + "tokens/trainable": 81646 + }, + { + "epoch": 0.703125, + "grad_norm": 0.015606923960149288, + "learning_rate": 7.95614819466576e-05, + "loss": 0.00040557183092460036, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00041, + "step": 180, + "tokens/total": 5441712, + "tokens/train_per_sec_per_gpu": 34.69, + "tokens/trainable": 82119 + }, + { + "epoch": 0.70703125, + "grad_norm": 0.06141373887658119, + "learning_rate": 7.931773536489872e-05, + "loss": 0.00117710349150002, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00118, + "step": 181, + "tokens/total": 5471968, + "tokens/train_per_sec_per_gpu": 39.13, + "tokens/trainable": 82636 + }, + { + "epoch": 0.7109375, + "grad_norm": 0.348529189825058, + "learning_rate": 7.907297682291035e-05, + "loss": 0.0045830621384084225, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00459, + "step": 182, + "tokens/total": 5502320, + "tokens/train_per_sec_per_gpu": 34.4, + "tokens/trainable": 83116 + }, + { + "epoch": 0.71484375, + "grad_norm": 0.3752739727497101, + "learning_rate": 7.882721650609442e-05, + "loss": 0.00952779222279787, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00957, + "step": 183, + "tokens/total": 5532512, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 83563 + }, + { + "epoch": 0.71875, + "grad_norm": 0.00500837154686451, + "learning_rate": 7.85804646415409e-05, + "loss": 0.00013146433047950268, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00013, + "step": 184, + "tokens/total": 5562992, + "tokens/train_per_sec_per_gpu": 35.95, + "tokens/trainable": 84069 + }, + { + "epoch": 0.72265625, + "grad_norm": 0.04002415016293526, + "learning_rate": 7.833273149760207e-05, + "loss": 0.000704923877492547, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00071, + "step": 185, + "tokens/total": 5593200, + "tokens/train_per_sec_per_gpu": 34.77, + "tokens/trainable": 84544 + }, + { + "epoch": 0.7265625, + "grad_norm": 0.24958014488220215, + "learning_rate": 7.808402738346527e-05, + "loss": 0.0023764465004205704, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00238, + "step": 186, + "tokens/total": 5623568, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 85005 + }, + { + "epoch": 0.73046875, + "grad_norm": 0.03483692184090614, + "learning_rate": 7.783436264872382e-05, + "loss": 0.0004840679175686091, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00048, + "step": 187, + "tokens/total": 5653856, + "tokens/train_per_sec_per_gpu": 32.14, + "tokens/trainable": 85445 + }, + { + "epoch": 0.734375, + "grad_norm": 0.8145651817321777, + "learning_rate": 7.758374768294647e-05, + "loss": 0.024159617722034454, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02445, + "step": 188, + "tokens/total": 5684112, + "tokens/train_per_sec_per_gpu": 34.38, + "tokens/trainable": 85935 + }, + { + "epoch": 0.73828125, + "grad_norm": 0.5074039697647095, + "learning_rate": 7.733219291524489e-05, + "loss": 0.010505304671823978, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.01056, + "step": 189, + "tokens/total": 5714400, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 86450 + }, + { + "epoch": 0.7421875, + "grad_norm": 0.4441295862197876, + "learning_rate": 7.707970881383977e-05, + "loss": 0.012071937322616577, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01215, + "step": 190, + "tokens/total": 5744992, + "tokens/train_per_sec_per_gpu": 34.21, + "tokens/trainable": 86921 + }, + { + "epoch": 0.74609375, + "grad_norm": 0.25269120931625366, + "learning_rate": 7.682630588562518e-05, + "loss": 0.005623019300401211, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00564, + "step": 191, + "tokens/total": 5775568, + "tokens/train_per_sec_per_gpu": 36.5, + "tokens/trainable": 87376 + }, + { + "epoch": 0.75, + "grad_norm": 0.40644046664237976, + "learning_rate": 7.657199467573129e-05, + "loss": 0.01618378609418869, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01632, + "step": 192, + "tokens/total": 5806000, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 87806 + }, + { + "epoch": 0.75390625, + "grad_norm": 0.2630307376384735, + "learning_rate": 7.631678576708561e-05, + "loss": 0.00848501268774271, + "memory/device_reserved (GiB)": 36.02, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00852, + "step": 193, + "tokens/total": 5836432, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 88289 + }, + { + "epoch": 0.7578125, + "grad_norm": 0.10661368817090988, + "learning_rate": 7.606068977997255e-05, + "loss": 0.00248312926851213, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00249, + "step": 194, + "tokens/total": 5866752, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 88760 + }, + { + "epoch": 0.76171875, + "grad_norm": 0.15434609353542328, + "learning_rate": 7.580371737159148e-05, + "loss": 0.005539691541343927, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00556, + "step": 195, + "tokens/total": 5894976, + "tokens/train_per_sec_per_gpu": 30.83, + "tokens/trainable": 89193 + }, + { + "epoch": 0.765625, + "grad_norm": 0.15394753217697144, + "learning_rate": 7.554587923561324e-05, + "loss": 0.0035556911025196314, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00356, + "step": 196, + "tokens/total": 5925376, + "tokens/train_per_sec_per_gpu": 37.72, + "tokens/trainable": 89655 + }, + { + "epoch": 0.76953125, + "grad_norm": 0.2555786371231079, + "learning_rate": 7.528718610173511e-05, + "loss": 0.0075815883465111256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.34, + "memory/max_allocated (GiB)": 33.34, + "ppl": 1.00761, + "step": 197, + "tokens/total": 5953696, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 90103 + }, + { + "epoch": 0.7734375, + "grad_norm": 0.319821834564209, + "learning_rate": 7.502764873523431e-05, + "loss": 0.013781344518065453, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.01388, + "step": 198, + "tokens/total": 5983952, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 90564 + }, + { + "epoch": 0.77734375, + "grad_norm": 0.25230011343955994, + "learning_rate": 7.476727793652011e-05, + "loss": 0.009122922085225582, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00916, + "step": 199, + "tokens/total": 6014704, + "tokens/train_per_sec_per_gpu": 36.87, + "tokens/trainable": 91043 + }, + { + "epoch": 0.78125, + "grad_norm": 0.30380022525787354, + "learning_rate": 7.450608454068415e-05, + "loss": 0.007986492477357388, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00802, + "step": 200, + "tokens/total": 6044960, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 91519 + }, + { + "epoch": 0.78515625, + "grad_norm": 0.07847452163696289, + "learning_rate": 7.424407941704987e-05, + "loss": 0.0024348655715584755, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00244, + "step": 201, + "tokens/total": 6074960, + "tokens/train_per_sec_per_gpu": 34.43, + "tokens/trainable": 91965 + }, + { + "epoch": 0.7890625, + "grad_norm": 0.15240712463855743, + "learning_rate": 7.398127346871986e-05, + "loss": 0.005189534276723862, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0052, + "step": 202, + "tokens/total": 6105312, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 92430 + }, + { + "epoch": 0.79296875, + "grad_norm": 0.33844277262687683, + "learning_rate": 7.371767763212238e-05, + "loss": 0.023467905819416046, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02375, + "step": 203, + "tokens/total": 6135392, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 92877 + }, + { + "epoch": 0.796875, + "grad_norm": 0.19864974915981293, + "learning_rate": 7.345330287655617e-05, + "loss": 0.00668431306257844, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00671, + "step": 204, + "tokens/total": 6165936, + "tokens/train_per_sec_per_gpu": 34.46, + "tokens/trainable": 93350 + }, + { + "epoch": 0.80078125, + "grad_norm": 0.20530343055725098, + "learning_rate": 7.31881602037339e-05, + "loss": 0.007711844518780708, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00774, + "step": 205, + "tokens/total": 6196128, + "tokens/train_per_sec_per_gpu": 33.89, + "tokens/trainable": 93774 + }, + { + "epoch": 0.8046875, + "grad_norm": 0.6116178631782532, + "learning_rate": 7.29222606473245e-05, + "loss": 0.044379763305187225, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.04538, + "step": 206, + "tokens/total": 6226288, + "tokens/train_per_sec_per_gpu": 38.11, + "tokens/trainable": 94253 + }, + { + "epoch": 0.80859375, + "grad_norm": 0.22179003059864044, + "learning_rate": 7.265561527249383e-05, + "loss": 0.009884262457489967, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00993, + "step": 207, + "tokens/total": 6256448, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 94743 + }, + { + "epoch": 0.8125, + "grad_norm": 0.13754810392856598, + "learning_rate": 7.238823517544436e-05, + "loss": 0.004908200353384018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00492, + "step": 208, + "tokens/total": 6286736, + "tokens/train_per_sec_per_gpu": 34.16, + "tokens/trainable": 95228 + }, + { + "epoch": 0.81640625, + "grad_norm": 0.11986561864614487, + "learning_rate": 7.212013148295333e-05, + "loss": 0.004731586202979088, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00474, + "step": 209, + "tokens/total": 6317376, + "tokens/train_per_sec_per_gpu": 29.53, + "tokens/trainable": 95629 + }, + { + "epoch": 0.8203125, + "grad_norm": 0.18956328928470612, + "learning_rate": 7.185131535190975e-05, + "loss": 0.009558689780533314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.0096, + "step": 210, + "tokens/total": 6347632, + "tokens/train_per_sec_per_gpu": 35.79, + "tokens/trainable": 96086 + }, + { + "epoch": 0.82421875, + "grad_norm": 0.2573530673980713, + "learning_rate": 7.158179796885005e-05, + "loss": 0.013298135250806808, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.01339, + "step": 211, + "tokens/total": 6377616, + "tokens/train_per_sec_per_gpu": 34.56, + "tokens/trainable": 96507 + }, + { + "epoch": 0.828125, + "grad_norm": 0.24280081689357758, + "learning_rate": 7.131159054949273e-05, + "loss": 0.00786699540913105, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0079, + "step": 212, + "tokens/total": 6408128, + "tokens/train_per_sec_per_gpu": 32.92, + "tokens/trainable": 96931 + }, + { + "epoch": 0.83203125, + "grad_norm": 0.1858828216791153, + "learning_rate": 7.104070433827139e-05, + "loss": 0.007124112918972969, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00715, + "step": 213, + "tokens/total": 6438560, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 97369 + }, + { + "epoch": 0.8359375, + "grad_norm": 0.38198429346084595, + "learning_rate": 7.076915060786705e-05, + "loss": 0.008005295880138874, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00804, + "step": 214, + "tokens/total": 6468832, + "tokens/train_per_sec_per_gpu": 36.86, + "tokens/trainable": 97858 + }, + { + "epoch": 0.83984375, + "grad_norm": 0.16806401312351227, + "learning_rate": 7.049694065873882e-05, + "loss": 0.004250777885317802, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00426, + "step": 215, + "tokens/total": 6498992, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 98308 + }, + { + "epoch": 0.84375, + "grad_norm": 0.011974122375249863, + "learning_rate": 7.022408581865382e-05, + "loss": 0.0003424361639190465, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00034, + "step": 216, + "tokens/total": 6529344, + "tokens/train_per_sec_per_gpu": 36.89, + "tokens/trainable": 98781 + }, + { + "epoch": 0.84765625, + "grad_norm": 0.07292452454566956, + "learning_rate": 6.99505974422157e-05, + "loss": 0.0010470495326444507, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00105, + "step": 217, + "tokens/total": 6559696, + "tokens/train_per_sec_per_gpu": 37.21, + "tokens/trainable": 99266 + }, + { + "epoch": 0.8515625, + "grad_norm": 0.023741189390420914, + "learning_rate": 6.967648691039213e-05, + "loss": 0.0004652513307519257, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00047, + "step": 218, + "tokens/total": 6588048, + "tokens/train_per_sec_per_gpu": 39.98, + "tokens/trainable": 99739 + }, + { + "epoch": 0.85546875, + "grad_norm": 0.020997121930122375, + "learning_rate": 6.940176563004123e-05, + "loss": 0.000519782246556133, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00052, + "step": 219, + "tokens/total": 6618480, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 100172 + }, + { + "epoch": 0.859375, + "grad_norm": 0.06872506439685822, + "learning_rate": 6.912644503343682e-05, + "loss": 0.0009209583513438702, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00092, + "step": 220, + "tokens/total": 6648880, + "tokens/train_per_sec_per_gpu": 37.96, + "tokens/trainable": 100646 + }, + { + "epoch": 0.86328125, + "grad_norm": 0.12261921912431717, + "learning_rate": 6.885053657779273e-05, + "loss": 0.0011635493719950318, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00116, + "step": 221, + "tokens/total": 6679040, + "tokens/train_per_sec_per_gpu": 32.05, + "tokens/trainable": 101097 + }, + { + "epoch": 0.8671875, + "grad_norm": 0.21586517989635468, + "learning_rate": 6.857405174478604e-05, + "loss": 0.004545507952570915, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00456, + "step": 222, + "tokens/total": 6709552, + "tokens/train_per_sec_per_gpu": 38.27, + "tokens/trainable": 101563 + }, + { + "epoch": 0.87109375, + "grad_norm": 0.8474878668785095, + "learning_rate": 6.82970020400792e-05, + "loss": 0.018102096393704414, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01827, + "step": 223, + "tokens/total": 6739936, + "tokens/train_per_sec_per_gpu": 31.91, + "tokens/trainable": 102004 + }, + { + "epoch": 0.875, + "grad_norm": 0.04830281436443329, + "learning_rate": 6.801939899284132e-05, + "loss": 0.0005447077564895153, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00054, + "step": 224, + "tokens/total": 6770144, + "tokens/train_per_sec_per_gpu": 35.91, + "tokens/trainable": 102452 + }, + { + "epoch": 0.87890625, + "grad_norm": 0.6089279651641846, + "learning_rate": 6.774125415526827e-05, + "loss": 0.007109163329005241, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00713, + "step": 225, + "tokens/total": 6800688, + "tokens/train_per_sec_per_gpu": 29.54, + "tokens/trainable": 102868 + }, + { + "epoch": 0.8828125, + "grad_norm": 0.3519023358821869, + "learning_rate": 6.746257910210214e-05, + "loss": 0.004347245674580336, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00436, + "step": 226, + "tokens/total": 6831376, + "tokens/train_per_sec_per_gpu": 37.74, + "tokens/trainable": 103375 + }, + { + "epoch": 0.88671875, + "grad_norm": 0.6944298148155212, + "learning_rate": 6.718338543014937e-05, + "loss": 0.011605200357735157, + "memory/device_reserved (GiB)": 35.17, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.01167, + "step": 227, + "tokens/total": 6861392, + "tokens/train_per_sec_per_gpu": 39.83, + "tokens/trainable": 103868 + }, + { + "epoch": 0.890625, + "grad_norm": 0.061347056180238724, + "learning_rate": 6.69036847577983e-05, + "loss": 0.0006434498354792595, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00064, + "step": 228, + "tokens/total": 6891776, + "tokens/train_per_sec_per_gpu": 36.72, + "tokens/trainable": 104331 + }, + { + "epoch": 0.89453125, + "grad_norm": 0.1004457101225853, + "learning_rate": 6.662348872453553e-05, + "loss": 0.0012449490604922175, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00125, + "step": 229, + "tokens/total": 6921776, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 104796 + }, + { + "epoch": 0.8984375, + "grad_norm": 0.47870421409606934, + "learning_rate": 6.63428089904618e-05, + "loss": 0.012319323606789112, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0124, + "step": 230, + "tokens/total": 6952048, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 105264 + }, + { + "epoch": 0.90234375, + "grad_norm": 0.4258827865123749, + "learning_rate": 6.60616572358065e-05, + "loss": 0.00830936897546053, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00834, + "step": 231, + "tokens/total": 6982384, + "tokens/train_per_sec_per_gpu": 34.25, + "tokens/trainable": 105727 + }, + { + "epoch": 0.90625, + "grad_norm": 0.07725202292203903, + "learning_rate": 6.578004516044172e-05, + "loss": 0.0011023205006495118, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0011, + "step": 232, + "tokens/total": 7012784, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 106203 + }, + { + "epoch": 0.91015625, + "grad_norm": 0.07384540885686874, + "learning_rate": 6.549798448339548e-05, + "loss": 0.0012245059479027987, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00123, + "step": 233, + "tokens/total": 7043184, + "tokens/train_per_sec_per_gpu": 38.46, + "tokens/trainable": 106708 + }, + { + "epoch": 0.9140625, + "grad_norm": 0.3441248834133148, + "learning_rate": 6.521548694236384e-05, + "loss": 0.006349029019474983, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00637, + "step": 234, + "tokens/total": 7073168, + "tokens/train_per_sec_per_gpu": 34.19, + "tokens/trainable": 107146 + }, + { + "epoch": 0.91796875, + "grad_norm": 0.019105862826108932, + "learning_rate": 6.493256429322259e-05, + "loss": 0.0003932910331059247, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00039, + "step": 235, + "tokens/total": 7103616, + "tokens/train_per_sec_per_gpu": 33.1, + "tokens/trainable": 107586 + }, + { + "epoch": 0.921875, + "grad_norm": 0.1013171374797821, + "learning_rate": 6.464922830953799e-05, + "loss": 0.0018943212926387787, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.0019, + "step": 236, + "tokens/total": 7133840, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 108051 + }, + { + "epoch": 0.92578125, + "grad_norm": 0.07892157137393951, + "learning_rate": 6.436549078207688e-05, + "loss": 0.0005698923487216234, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00057, + "step": 237, + "tokens/total": 7164016, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 108479 + }, + { + "epoch": 0.9296875, + "grad_norm": 0.45241886377334595, + "learning_rate": 6.408136351831592e-05, + "loss": 0.0160800963640213, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01621, + "step": 238, + "tokens/total": 7194400, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 108907 + }, + { + "epoch": 0.93359375, + "grad_norm": 0.23932066559791565, + "learning_rate": 6.379685834195036e-05, + "loss": 0.009671966545283794, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00972, + "step": 239, + "tokens/total": 7224992, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 109348 + }, + { + "epoch": 0.9375, + "grad_norm": 0.04965788871049881, + "learning_rate": 6.351198709240186e-05, + "loss": 0.0012405012967064977, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00124, + "step": 240, + "tokens/total": 7255424, + "tokens/train_per_sec_per_gpu": 32.71, + "tokens/trainable": 109812 + }, + { + "epoch": 0.94140625, + "grad_norm": 0.29825523495674133, + "learning_rate": 6.32267616243259e-05, + "loss": 0.005660452414304018, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00568, + "step": 241, + "tokens/total": 7285696, + "tokens/train_per_sec_per_gpu": 38.88, + "tokens/trainable": 110270 + }, + { + "epoch": 0.9453125, + "grad_norm": 0.26985254883766174, + "learning_rate": 6.294119380711849e-05, + "loss": 0.0047082495875656605, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00472, + "step": 242, + "tokens/total": 7316000, + "tokens/train_per_sec_per_gpu": 28.25, + "tokens/trainable": 110682 + }, + { + "epoch": 0.94921875, + "grad_norm": 0.43075069785118103, + "learning_rate": 6.265529552442209e-05, + "loss": 0.003754058852791786, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00376, + "step": 243, + "tokens/total": 7346304, + "tokens/train_per_sec_per_gpu": 39.54, + "tokens/trainable": 111176 + }, + { + "epoch": 0.953125, + "grad_norm": 0.27702149748802185, + "learning_rate": 6.236907867363127e-05, + "loss": 0.008292095735669136, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00833, + "step": 244, + "tokens/total": 7376944, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 111633 + }, + { + "epoch": 0.95703125, + "grad_norm": 0.07920246571302414, + "learning_rate": 6.208255516539749e-05, + "loss": 0.0018851464847102761, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00189, + "step": 245, + "tokens/total": 7407184, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 112115 + }, + { + "epoch": 0.9609375, + "grad_norm": 0.306937575340271, + "learning_rate": 6.179573692313344e-05, + "loss": 0.008207850158214569, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00824, + "step": 246, + "tokens/total": 7437760, + "tokens/train_per_sec_per_gpu": 33.59, + "tokens/trainable": 112589 + }, + { + "epoch": 0.96484375, + "grad_norm": 0.22587862610816956, + "learning_rate": 6.150863588251694e-05, + "loss": 0.004535369109362364, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00455, + "step": 247, + "tokens/total": 7466080, + "tokens/train_per_sec_per_gpu": 38.2, + "tokens/trainable": 113043 + }, + { + "epoch": 0.96875, + "grad_norm": 0.1514762043952942, + "learning_rate": 6.122126399099419e-05, + "loss": 0.0027371919713914394, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00274, + "step": 248, + "tokens/total": 7496672, + "tokens/train_per_sec_per_gpu": 35.66, + "tokens/trainable": 113484 + }, + { + "epoch": 0.97265625, + "grad_norm": 0.09306064248085022, + "learning_rate": 6.0933633207282615e-05, + "loss": 0.002224899362772703, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00223, + "step": 249, + "tokens/total": 7526816, + "tokens/train_per_sec_per_gpu": 33.02, + "tokens/trainable": 113928 + }, + { + "epoch": 0.9765625, + "grad_norm": 0.06742242723703384, + "learning_rate": 6.064575550087316e-05, + "loss": 0.0009518311126157641, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00095, + "step": 250, + "tokens/total": 7555120, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 114360 + }, + { + "epoch": 0.98046875, + "grad_norm": 0.09547543525695801, + "learning_rate": 6.0357642851532245e-05, + "loss": 0.0044240644201636314, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00443, + "step": 251, + "tokens/total": 7585632, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 114850 + }, + { + "epoch": 0.984375, + "grad_norm": 0.35386043787002563, + "learning_rate": 6.0069307248803294e-05, + "loss": 0.009106422774493694, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00915, + "step": 252, + "tokens/total": 7615888, + "tokens/train_per_sec_per_gpu": 35.56, + "tokens/trainable": 115298 + }, + { + "epoch": 0.98828125, + "grad_norm": 0.047934144735336304, + "learning_rate": 5.9780760691507635e-05, + "loss": 0.0007137035136111081, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00071, + "step": 253, + "tokens/total": 7646464, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 115719 + }, + { + "epoch": 0.9921875, + "grad_norm": 0.2060946822166443, + "learning_rate": 5.9492015187245334e-05, + "loss": 0.0057786800898611546, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0058, + "step": 254, + "tokens/total": 7676896, + "tokens/train_per_sec_per_gpu": 29.56, + "tokens/trainable": 116151 + }, + { + "epoch": 0.99609375, + "grad_norm": 0.02973618358373642, + "learning_rate": 5.920308275189541e-05, + "loss": 0.000541117915417999, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00054, + "step": 255, + "tokens/total": 7707328, + "tokens/train_per_sec_per_gpu": 37.19, + "tokens/trainable": 116602 + }, + { + "epoch": 1.0, + "grad_norm": 0.026812957599759102, + "learning_rate": 5.8913975409115874e-05, + "loss": 0.0003775915829464793, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00038, + "step": 256, + "tokens/total": 7737920, + "tokens/train_per_sec_per_gpu": 32.45, + "tokens/trainable": 117050 + }, + { + "epoch": 1.00390625, + "grad_norm": 0.13118499517440796, + "learning_rate": 5.8624705189843395e-05, + "loss": 0.0015422197757288814, + "memory/device_reserved (GiB)": 36.21, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00154, + "step": 257, + "tokens/total": 7768256, + "tokens/train_per_sec_per_gpu": 32.54, + "tokens/trainable": 117506 + }, + { + "epoch": 1.0078125, + "grad_norm": 0.007149726152420044, + "learning_rate": 5.833528413179249e-05, + "loss": 0.0001414301514159888, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00014, + "step": 258, + "tokens/total": 7798544, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 117936 + }, + { + "epoch": 1.01171875, + "grad_norm": 0.25255724787712097, + "learning_rate": 5.80457242789548e-05, + "loss": 0.0026415749453008175, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00265, + "step": 259, + "tokens/total": 7828880, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 118377 + }, + { + "epoch": 1.015625, + "grad_norm": 0.0023047819267958403, + "learning_rate": 5.77560376810977e-05, + "loss": 4.6871849917806685e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 260, + "tokens/total": 7859312, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 118820 + }, + { + "epoch": 1.01953125, + "grad_norm": 0.1961309015750885, + "learning_rate": 5.7466236393263005e-05, + "loss": 0.0034356648102402687, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00344, + "step": 261, + "tokens/total": 7889776, + "tokens/train_per_sec_per_gpu": 34.37, + "tokens/trainable": 119267 + }, + { + "epoch": 1.0234375, + "grad_norm": 0.06521005183458328, + "learning_rate": 5.717633247526522e-05, + "loss": 0.0006203935481607914, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00062, + "step": 262, + "tokens/total": 7920256, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 119776 + }, + { + "epoch": 1.02734375, + "grad_norm": 0.2380313128232956, + "learning_rate": 5.688633799118971e-05, + "loss": 0.003634552238509059, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00364, + "step": 263, + "tokens/total": 7950608, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 120265 + }, + { + "epoch": 1.03125, + "grad_norm": 0.22213542461395264, + "learning_rate": 5.659626500889066e-05, + "loss": 0.0028508887626230717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00285, + "step": 264, + "tokens/total": 7981168, + "tokens/train_per_sec_per_gpu": 37.54, + "tokens/trainable": 120742 + }, + { + "epoch": 1.03515625, + "grad_norm": 0.007840439677238464, + "learning_rate": 5.6306125599488905e-05, + "loss": 0.00010933953308267519, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00011, + "step": 265, + "tokens/total": 8011504, + "tokens/train_per_sec_per_gpu": 40.03, + "tokens/trainable": 121209 + }, + { + "epoch": 1.0390625, + "grad_norm": 0.06023910269141197, + "learning_rate": 5.601593183686955e-05, + "loss": 0.0008161423029378057, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 266, + "tokens/total": 8042048, + "tokens/train_per_sec_per_gpu": 32.67, + "tokens/trainable": 121683 + }, + { + "epoch": 1.04296875, + "grad_norm": 0.014270068146288395, + "learning_rate": 5.572569579717961e-05, + "loss": 0.00010447163367643952, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0001, + "step": 267, + "tokens/total": 8072336, + "tokens/train_per_sec_per_gpu": 40.53, + "tokens/trainable": 122187 + }, + { + "epoch": 1.046875, + "grad_norm": 0.036489423364400864, + "learning_rate": 5.543542955832538e-05, + "loss": 0.0003805297310464084, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00038, + "step": 268, + "tokens/total": 8102592, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 122649 + }, + { + "epoch": 1.05078125, + "grad_norm": 0.04869162663817406, + "learning_rate": 5.514514519946986e-05, + "loss": 0.0005166378105059266, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00052, + "step": 269, + "tokens/total": 8132752, + "tokens/train_per_sec_per_gpu": 33.52, + "tokens/trainable": 123109 + }, + { + "epoch": 1.0546875, + "grad_norm": 0.0886458307504654, + "learning_rate": 5.485485480053015e-05, + "loss": 0.0009586921078152955, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00096, + "step": 270, + "tokens/total": 8163232, + "tokens/train_per_sec_per_gpu": 36.69, + "tokens/trainable": 123599 + }, + { + "epoch": 1.05859375, + "grad_norm": 0.07190798223018646, + "learning_rate": 5.4564570441674645e-05, + "loss": 0.0005164776812307537, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00052, + "step": 271, + "tokens/total": 8193344, + "tokens/train_per_sec_per_gpu": 38.15, + "tokens/trainable": 124081 + }, + { + "epoch": 1.0625, + "grad_norm": 0.0680084079504013, + "learning_rate": 5.42743042028204e-05, + "loss": 0.00043266740976832807, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00043, + "step": 272, + "tokens/total": 8223632, + "tokens/train_per_sec_per_gpu": 35.19, + "tokens/trainable": 124551 + }, + { + "epoch": 1.06640625, + "grad_norm": 0.008596232160925865, + "learning_rate": 5.3984068163130464e-05, + "loss": 5.5010747018968686e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00006, + "step": 273, + "tokens/total": 8253680, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 124981 + }, + { + "epoch": 1.0703125, + "grad_norm": 0.12593501806259155, + "learning_rate": 5.369387440051111e-05, + "loss": 0.0006858849665150046, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00069, + "step": 274, + "tokens/total": 8284256, + "tokens/train_per_sec_per_gpu": 35.0, + "tokens/trainable": 125442 + }, + { + "epoch": 1.07421875, + "grad_norm": 0.002708524465560913, + "learning_rate": 5.340373499110935e-05, + "loss": 2.155889706045855e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00002, + "step": 275, + "tokens/total": 8314896, + "tokens/train_per_sec_per_gpu": 31.81, + "tokens/trainable": 125893 + }, + { + "epoch": 1.078125, + "grad_norm": 0.6500045657157898, + "learning_rate": 5.3113662008810304e-05, + "loss": 0.0032175867818295956, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00322, + "step": 276, + "tokens/total": 8345264, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 126367 + }, + { + "epoch": 1.08203125, + "grad_norm": 0.0035925847478210926, + "learning_rate": 5.282366752473479e-05, + "loss": 4.369133966974914e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 277, + "tokens/total": 8375728, + "tokens/train_per_sec_per_gpu": 34.39, + "tokens/trainable": 126832 + }, + { + "epoch": 1.0859375, + "grad_norm": 0.06011431664228439, + "learning_rate": 5.2533763606737005e-05, + "loss": 0.00032412572181783617, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00032, + "step": 278, + "tokens/total": 8405952, + "tokens/train_per_sec_per_gpu": 38.74, + "tokens/trainable": 127335 + }, + { + "epoch": 1.08984375, + "grad_norm": 0.0041359406895935535, + "learning_rate": 5.224396231890232e-05, + "loss": 4.0772643842501566e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00004, + "step": 279, + "tokens/total": 8436096, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 127793 + }, + { + "epoch": 1.09375, + "grad_norm": 0.0019470915431156754, + "learning_rate": 5.195427572104522e-05, + "loss": 2.092921204166487e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00002, + "step": 280, + "tokens/total": 8466464, + "tokens/train_per_sec_per_gpu": 33.21, + "tokens/trainable": 128242 + }, + { + "epoch": 1.09765625, + "grad_norm": 0.0017560214037075639, + "learning_rate": 5.166471586820751e-05, + "loss": 2.1481781004695222e-05, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 281, + "tokens/total": 8496736, + "tokens/train_per_sec_per_gpu": 36.63, + "tokens/trainable": 128724 + }, + { + "epoch": 1.1015625, + "grad_norm": 0.01728290133178234, + "learning_rate": 5.1375294810156615e-05, + "loss": 0.00014254249981604517, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00014, + "step": 282, + "tokens/total": 8526928, + "tokens/train_per_sec_per_gpu": 35.5, + "tokens/trainable": 129198 + }, + { + "epoch": 1.10546875, + "grad_norm": 0.014790019951760769, + "learning_rate": 5.1086024590884144e-05, + "loss": 6.786123412894085e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00007, + "step": 283, + "tokens/total": 8557296, + "tokens/train_per_sec_per_gpu": 33.38, + "tokens/trainable": 129648 + }, + { + "epoch": 1.109375, + "grad_norm": 0.4627615809440613, + "learning_rate": 5.079691724810461e-05, + "loss": 0.004037827253341675, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00405, + "step": 284, + "tokens/total": 8587696, + "tokens/train_per_sec_per_gpu": 32.72, + "tokens/trainable": 130095 + }, + { + "epoch": 1.11328125, + "grad_norm": 0.3512204587459564, + "learning_rate": 5.0507984812754684e-05, + "loss": 0.002691472414880991, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0027, + "step": 285, + "tokens/total": 8618080, + "tokens/train_per_sec_per_gpu": 31.74, + "tokens/trainable": 130545 + }, + { + "epoch": 1.1171875, + "grad_norm": 0.009706776589155197, + "learning_rate": 5.021923930849237e-05, + "loss": 8.887588774086908e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00009, + "step": 286, + "tokens/total": 8648288, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 131002 + }, + { + "epoch": 1.12109375, + "grad_norm": 0.07987845689058304, + "learning_rate": 4.99306927511967e-05, + "loss": 0.0007880277116782963, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00079, + "step": 287, + "tokens/total": 8678464, + "tokens/train_per_sec_per_gpu": 34.31, + "tokens/trainable": 131458 + }, + { + "epoch": 1.125, + "grad_norm": 0.0013652736088261008, + "learning_rate": 4.964235714846775e-05, + "loss": 1.4229615771910176e-05, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 288, + "tokens/total": 8708816, + "tokens/train_per_sec_per_gpu": 36.32, + "tokens/trainable": 131932 + }, + { + "epoch": 1.12890625, + "grad_norm": 0.00022536546748597175, + "learning_rate": 4.9354244499126866e-05, + "loss": 5.348314971342916e-06, + "memory/device_reserved (GiB)": 35.91, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00001, + "step": 289, + "tokens/total": 8738992, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 132406 + }, + { + "epoch": 1.1328125, + "grad_norm": 0.00034249111195094883, + "learning_rate": 4.90663667927174e-05, + "loss": 6.50786478217924e-06, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00001, + "step": 290, + "tokens/total": 8769424, + "tokens/train_per_sec_per_gpu": 36.83, + "tokens/trainable": 132855 + }, + { + "epoch": 1.13671875, + "grad_norm": 0.0011629040818661451, + "learning_rate": 4.877873600900581e-05, + "loss": 2.100312485708855e-05, + "memory/device_reserved (GiB)": 35.57, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00002, + "step": 291, + "tokens/total": 8799680, + "tokens/train_per_sec_per_gpu": 36.18, + "tokens/trainable": 133356 + }, + { + "epoch": 1.140625, + "grad_norm": 0.0002511514176148921, + "learning_rate": 4.849136411748306e-05, + "loss": 5.5127738960436545e-06, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00001, + "step": 292, + "tokens/total": 8830000, + "tokens/train_per_sec_per_gpu": 34.54, + "tokens/trainable": 133788 + }, + { + "epoch": 1.14453125, + "grad_norm": 0.005342632066458464, + "learning_rate": 4.8204263076866574e-05, + "loss": 5.693141429219395e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00006, + "step": 293, + "tokens/total": 8860320, + "tokens/train_per_sec_per_gpu": 36.67, + "tokens/trainable": 134260 + }, + { + "epoch": 1.1484375, + "grad_norm": 0.015129564329981804, + "learning_rate": 4.791744483460251e-05, + "loss": 5.992315709590912e-05, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00006, + "step": 294, + "tokens/total": 8891072, + "tokens/train_per_sec_per_gpu": 34.01, + "tokens/trainable": 134684 + }, + { + "epoch": 1.15234375, + "grad_norm": 0.2627723515033722, + "learning_rate": 4.7630921326368736e-05, + "loss": 0.0013240529224276543, + "memory/device_reserved (GiB)": 35.59, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00132, + "step": 295, + "tokens/total": 8921504, + "tokens/train_per_sec_per_gpu": 36.05, + "tokens/trainable": 135166 + }, + { + "epoch": 1.15625, + "grad_norm": 0.00017540185945108533, + "learning_rate": 4.7344704475577916e-05, + "loss": 4.621510925062466e-06, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.0, + "step": 296, + "tokens/total": 8951808, + "tokens/train_per_sec_per_gpu": 33.62, + "tokens/trainable": 135600 + }, + { + "epoch": 1.16015625, + "grad_norm": 0.34399038553237915, + "learning_rate": 4.705880619288153e-05, + "loss": 0.007753564510494471, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00778, + "step": 297, + "tokens/total": 8982080, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 136039 + }, + { + "epoch": 1.1640625, + "grad_norm": 0.010148555971682072, + "learning_rate": 4.677323837567412e-05, + "loss": 5.8421592257218435e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 298, + "tokens/total": 9012560, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 136496 + }, + { + "epoch": 1.16796875, + "grad_norm": 0.0038862715009599924, + "learning_rate": 4.6488012907598146e-05, + "loss": 3.437784835114144e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 299, + "tokens/total": 9042944, + "tokens/train_per_sec_per_gpu": 35.75, + "tokens/trainable": 136954 + }, + { + "epoch": 1.171875, + "grad_norm": 0.0015254979953169823, + "learning_rate": 4.620314165804964e-05, + "loss": 2.051112096523866e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00002, + "step": 300, + "tokens/total": 9073248, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 137404 + }, + { + "epoch": 1.17578125, + "grad_norm": 0.4777839481830597, + "learning_rate": 4.591863648168407e-05, + "loss": 0.006096641067415476, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00612, + "step": 301, + "tokens/total": 9103728, + "tokens/train_per_sec_per_gpu": 36.12, + "tokens/trainable": 137899 + }, + { + "epoch": 1.1796875, + "grad_norm": 0.09574320912361145, + "learning_rate": 4.5634509217923135e-05, + "loss": 0.00047178880777209997, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00047, + "step": 302, + "tokens/total": 9134000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 138375 + }, + { + "epoch": 1.18359375, + "grad_norm": 0.003620963543653488, + "learning_rate": 4.535077169046201e-05, + "loss": 4.1847168176900595e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 303, + "tokens/total": 9164400, + "tokens/train_per_sec_per_gpu": 36.82, + "tokens/trainable": 138834 + }, + { + "epoch": 1.1875, + "grad_norm": 0.004095857497304678, + "learning_rate": 4.506743570677743e-05, + "loss": 3.6887995520373806e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00004, + "step": 304, + "tokens/total": 9194672, + "tokens/train_per_sec_per_gpu": 30.99, + "tokens/trainable": 139259 + }, + { + "epoch": 1.19140625, + "grad_norm": 0.004100849851965904, + "learning_rate": 4.478451305763618e-05, + "loss": 4.464022640604526e-05, + "memory/device_reserved (GiB)": 35.67, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00004, + "step": 305, + "tokens/total": 9224848, + "tokens/train_per_sec_per_gpu": 33.42, + "tokens/trainable": 139717 + }, + { + "epoch": 1.1953125, + "grad_norm": 0.22271950542926788, + "learning_rate": 4.450201551660454e-05, + "loss": 0.002607254544273019, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00261, + "step": 306, + "tokens/total": 9255440, + "tokens/train_per_sec_per_gpu": 37.37, + "tokens/trainable": 140207 + }, + { + "epoch": 1.19921875, + "grad_norm": 0.05722619220614433, + "learning_rate": 4.4219954839558276e-05, + "loss": 0.00047086336417123675, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.19, + "memory/max_allocated (GiB)": 33.19, + "ppl": 1.00047, + "step": 307, + "tokens/total": 9283344, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 140645 + }, + { + "epoch": 1.203125, + "grad_norm": 0.6939759850502014, + "learning_rate": 4.393834276419352e-05, + "loss": 0.004533157218247652, + "memory/device_reserved (GiB)": 35.9, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00454, + "step": 308, + "tokens/total": 9313856, + "tokens/train_per_sec_per_gpu": 30.34, + "tokens/trainable": 141096 + }, + { + "epoch": 1.20703125, + "grad_norm": 0.003842511447146535, + "learning_rate": 4.36571910095382e-05, + "loss": 4.202376658213325e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00004, + "step": 309, + "tokens/total": 9344064, + "tokens/train_per_sec_per_gpu": 29.26, + "tokens/trainable": 141518 + }, + { + "epoch": 1.2109375, + "grad_norm": 0.002201825613155961, + "learning_rate": 4.337651127546448e-05, + "loss": 4.229221667628735e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00004, + "step": 310, + "tokens/total": 9374368, + "tokens/train_per_sec_per_gpu": 35.84, + "tokens/trainable": 141952 + }, + { + "epoch": 1.21484375, + "grad_norm": 0.0005678636953234673, + "learning_rate": 4.3096315242201736e-05, + "loss": 1.2881277143605985e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00001, + "step": 311, + "tokens/total": 9404624, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 142416 + }, + { + "epoch": 1.21875, + "grad_norm": 0.0005692228442057967, + "learning_rate": 4.2816614569850635e-05, + "loss": 1.2092226825188845e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00001, + "step": 312, + "tokens/total": 9434784, + "tokens/train_per_sec_per_gpu": 38.63, + "tokens/trainable": 142882 + }, + { + "epoch": 1.22265625, + "grad_norm": 0.00879319105297327, + "learning_rate": 4.2537420897897864e-05, + "loss": 9.982152550946921e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0001, + "step": 313, + "tokens/total": 9465168, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 143356 + }, + { + "epoch": 1.2265625, + "grad_norm": 0.008733258582651615, + "learning_rate": 4.225874584473174e-05, + "loss": 0.00010937166371149942, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00011, + "step": 314, + "tokens/total": 9495488, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 143812 + }, + { + "epoch": 1.23046875, + "grad_norm": 0.016755763441324234, + "learning_rate": 4.19806010071587e-05, + "loss": 0.00017517709056846797, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00018, + "step": 315, + "tokens/total": 9525968, + "tokens/train_per_sec_per_gpu": 34.95, + "tokens/trainable": 144294 + }, + { + "epoch": 1.234375, + "grad_norm": 0.01962393708527088, + "learning_rate": 4.170299795992081e-05, + "loss": 0.0001436145685147494, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00014, + "step": 316, + "tokens/total": 9556368, + "tokens/train_per_sec_per_gpu": 35.87, + "tokens/trainable": 144751 + }, + { + "epoch": 1.23828125, + "grad_norm": 0.0018895679386332631, + "learning_rate": 4.142594825521398e-05, + "loss": 3.0470149795291945e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 317, + "tokens/total": 9586800, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 145251 + }, + { + "epoch": 1.2421875, + "grad_norm": 0.030299803242087364, + "learning_rate": 4.114946342220728e-05, + "loss": 0.0003194157325197011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00032, + "step": 318, + "tokens/total": 9617152, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 145711 + }, + { + "epoch": 1.24609375, + "grad_norm": 0.04471183568239212, + "learning_rate": 4.087355496656321e-05, + "loss": 0.0005480494000948966, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00055, + "step": 319, + "tokens/total": 9647424, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 146205 + }, + { + "epoch": 1.25, + "grad_norm": 1.604000210762024, + "learning_rate": 4.05982343699588e-05, + "loss": 0.003160722553730011, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00317, + "step": 320, + "tokens/total": 9677776, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 146678 + }, + { + "epoch": 1.25390625, + "grad_norm": 0.0033273256849497557, + "learning_rate": 4.0323513089607876e-05, + "loss": 3.962405025959015e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 321, + "tokens/total": 9707936, + "tokens/train_per_sec_per_gpu": 35.71, + "tokens/trainable": 147124 + }, + { + "epoch": 1.2578125, + "grad_norm": 0.005383767187595367, + "learning_rate": 4.004940255778431e-05, + "loss": 7.032878784229979e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 322, + "tokens/total": 9738448, + "tokens/train_per_sec_per_gpu": 35.58, + "tokens/trainable": 147574 + }, + { + "epoch": 1.26171875, + "grad_norm": 0.0028380267322063446, + "learning_rate": 3.977591418134619e-05, + "loss": 5.103685543872416e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 323, + "tokens/total": 9768864, + "tokens/train_per_sec_per_gpu": 35.65, + "tokens/trainable": 148030 + }, + { + "epoch": 1.265625, + "grad_norm": 0.025233343243598938, + "learning_rate": 3.95030593412612e-05, + "loss": 0.00011621347948675975, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00012, + "step": 324, + "tokens/total": 9799184, + "tokens/train_per_sec_per_gpu": 37.22, + "tokens/trainable": 148478 + }, + { + "epoch": 1.26953125, + "grad_norm": 0.04950147494673729, + "learning_rate": 3.923084939213296e-05, + "loss": 0.0003831804497167468, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00038, + "step": 325, + "tokens/total": 9829392, + "tokens/train_per_sec_per_gpu": 36.95, + "tokens/trainable": 148941 + }, + { + "epoch": 1.2734375, + "grad_norm": 0.0017629251815378666, + "learning_rate": 3.895929566172861e-05, + "loss": 2.4212906282627955e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 326, + "tokens/total": 9859696, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 149435 + }, + { + "epoch": 1.27734375, + "grad_norm": 0.025092069059610367, + "learning_rate": 3.868840945050728e-05, + "loss": 0.0002421572571620345, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00024, + "step": 327, + "tokens/total": 9889648, + "tokens/train_per_sec_per_gpu": 39.82, + "tokens/trainable": 149888 + }, + { + "epoch": 1.28125, + "grad_norm": 0.04499012976884842, + "learning_rate": 3.841820203114995e-05, + "loss": 0.0005073954816907644, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00051, + "step": 328, + "tokens/total": 9919968, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 150362 + }, + { + "epoch": 1.28515625, + "grad_norm": 0.016609711572527885, + "learning_rate": 3.814868464809027e-05, + "loss": 0.00013441528426483274, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00013, + "step": 329, + "tokens/total": 9950368, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 150837 + }, + { + "epoch": 1.2890625, + "grad_norm": 0.003686394076794386, + "learning_rate": 3.787986851704667e-05, + "loss": 4.67524841951672e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 330, + "tokens/total": 9980688, + "tokens/train_per_sec_per_gpu": 34.36, + "tokens/trainable": 151273 + }, + { + "epoch": 1.29296875, + "grad_norm": 0.0016950524877756834, + "learning_rate": 3.7611764824555654e-05, + "loss": 2.6381265342934057e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00003, + "step": 331, + "tokens/total": 10011104, + "tokens/train_per_sec_per_gpu": 37.63, + "tokens/trainable": 151741 + }, + { + "epoch": 1.296875, + "grad_norm": 0.004727163352072239, + "learning_rate": 3.734438472750619e-05, + "loss": 4.7812289267312735e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00005, + "step": 332, + "tokens/total": 10041536, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 152191 + }, + { + "epoch": 1.30078125, + "grad_norm": 0.09555032104253769, + "learning_rate": 3.707773935267552e-05, + "loss": 0.0008173320675268769, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00082, + "step": 333, + "tokens/total": 10072112, + "tokens/train_per_sec_per_gpu": 32.24, + "tokens/trainable": 152630 + }, + { + "epoch": 1.3046875, + "grad_norm": 0.02793118543922901, + "learning_rate": 3.68118397962661e-05, + "loss": 0.0002140044525731355, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00021, + "step": 334, + "tokens/total": 10102592, + "tokens/train_per_sec_per_gpu": 33.85, + "tokens/trainable": 153047 + }, + { + "epoch": 1.30859375, + "grad_norm": 0.2730892598628998, + "learning_rate": 3.654669712344384e-05, + "loss": 0.004334757570177317, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00434, + "step": 335, + "tokens/total": 10132736, + "tokens/train_per_sec_per_gpu": 30.89, + "tokens/trainable": 153477 + }, + { + "epoch": 1.3125, + "grad_norm": 0.002235857769846916, + "learning_rate": 3.628232236787763e-05, + "loss": 9.928971849149093e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00001, + "step": 336, + "tokens/total": 10162928, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 153929 + }, + { + "epoch": 1.31640625, + "grad_norm": 0.0533645860850811, + "learning_rate": 3.6018726531280144e-05, + "loss": 0.0006771332118660212, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00068, + "step": 337, + "tokens/total": 10193552, + "tokens/train_per_sec_per_gpu": 35.34, + "tokens/trainable": 154416 + }, + { + "epoch": 1.3203125, + "grad_norm": 0.003405389143154025, + "learning_rate": 3.575592058295017e-05, + "loss": 4.594672645907849e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 338, + "tokens/total": 10223808, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 154865 + }, + { + "epoch": 1.32421875, + "grad_norm": 0.0007159400847740471, + "learning_rate": 3.549391545931585e-05, + "loss": 1.3502348338079173e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 339, + "tokens/total": 10254272, + "tokens/train_per_sec_per_gpu": 30.21, + "tokens/trainable": 155290 + }, + { + "epoch": 1.328125, + "grad_norm": 0.013892000541090965, + "learning_rate": 3.5232722063479914e-05, + "loss": 7.453892612829804e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00007, + "step": 340, + "tokens/total": 10284656, + "tokens/train_per_sec_per_gpu": 31.6, + "tokens/trainable": 155703 + }, + { + "epoch": 1.33203125, + "grad_norm": 0.005680185277014971, + "learning_rate": 3.49723512647657e-05, + "loss": 7.792153337504715e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00008, + "step": 341, + "tokens/total": 10314976, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 156193 + }, + { + "epoch": 1.3359375, + "grad_norm": 0.015012380667030811, + "learning_rate": 3.471281389826491e-05, + "loss": 0.0001705413160379976, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00017, + "step": 342, + "tokens/total": 10345360, + "tokens/train_per_sec_per_gpu": 34.09, + "tokens/trainable": 156640 + }, + { + "epoch": 1.33984375, + "grad_norm": 0.0032200925052165985, + "learning_rate": 3.4454120764386764e-05, + "loss": 2.8803326131310314e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00003, + "step": 343, + "tokens/total": 10375648, + "tokens/train_per_sec_per_gpu": 34.14, + "tokens/trainable": 157102 + }, + { + "epoch": 1.34375, + "grad_norm": 0.007129024248570204, + "learning_rate": 3.4196282628408526e-05, + "loss": 5.837350181536749e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 344, + "tokens/total": 10406144, + "tokens/train_per_sec_per_gpu": 33.51, + "tokens/trainable": 157544 + }, + { + "epoch": 1.34765625, + "grad_norm": 0.30264967679977417, + "learning_rate": 3.3939310220027456e-05, + "loss": 0.007746525574475527, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00778, + "step": 345, + "tokens/total": 10436528, + "tokens/train_per_sec_per_gpu": 35.6, + "tokens/trainable": 157985 + }, + { + "epoch": 1.3515625, + "grad_norm": 0.014983746223151684, + "learning_rate": 3.3683214232914404e-05, + "loss": 0.0001249729102710262, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00012, + "step": 346, + "tokens/total": 10466864, + "tokens/train_per_sec_per_gpu": 29.48, + "tokens/trainable": 158402 + }, + { + "epoch": 1.35546875, + "grad_norm": 0.0023245092015713453, + "learning_rate": 3.342800532426873e-05, + "loss": 2.6811658244696446e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00003, + "step": 347, + "tokens/total": 10497008, + "tokens/train_per_sec_per_gpu": 36.08, + "tokens/trainable": 158860 + }, + { + "epoch": 1.359375, + "grad_norm": 0.01576041616499424, + "learning_rate": 3.317369411437484e-05, + "loss": 0.00016933181905187666, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.0, + "memory/max_allocated (GiB)": 34.0, + "ppl": 1.00017, + "step": 348, + "tokens/total": 10527728, + "tokens/train_per_sec_per_gpu": 29.09, + "tokens/trainable": 159282 + }, + { + "epoch": 1.36328125, + "grad_norm": 0.20484071969985962, + "learning_rate": 3.292029118616024e-05, + "loss": 0.0017917631193995476, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00179, + "step": 349, + "tokens/total": 10557856, + "tokens/train_per_sec_per_gpu": 32.95, + "tokens/trainable": 159732 + }, + { + "epoch": 1.3671875, + "grad_norm": 0.0009161452180705965, + "learning_rate": 3.266780708475511e-05, + "loss": 1.5678642739658244e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00002, + "step": 350, + "tokens/total": 10588368, + "tokens/train_per_sec_per_gpu": 35.68, + "tokens/trainable": 160199 + }, + { + "epoch": 1.37109375, + "grad_norm": 0.0552787259221077, + "learning_rate": 3.241625231705354e-05, + "loss": 0.0003493681433610618, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00035, + "step": 351, + "tokens/total": 10618608, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 160657 + }, + { + "epoch": 1.375, + "grad_norm": 0.01718018762767315, + "learning_rate": 3.216563735127618e-05, + "loss": 0.00022284504666458815, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00022, + "step": 352, + "tokens/total": 10648864, + "tokens/train_per_sec_per_gpu": 34.98, + "tokens/trainable": 161114 + }, + { + "epoch": 1.37890625, + "grad_norm": 0.015219186432659626, + "learning_rate": 3.191597261653475e-05, + "loss": 0.00014846389240119606, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00015, + "step": 353, + "tokens/total": 10679184, + "tokens/train_per_sec_per_gpu": 29.27, + "tokens/trainable": 161567 + }, + { + "epoch": 1.3828125, + "grad_norm": 0.002392370719462633, + "learning_rate": 3.166726850239794e-05, + "loss": 2.664637577254325e-05, + "memory/device_reserved (GiB)": 35.41, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00003, + "step": 354, + "tokens/total": 10709648, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 162032 + }, + { + "epoch": 1.38671875, + "grad_norm": 0.012528779916465282, + "learning_rate": 3.141953535845912e-05, + "loss": 0.00010976240446325392, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00011, + "step": 355, + "tokens/total": 10740128, + "tokens/train_per_sec_per_gpu": 32.56, + "tokens/trainable": 162460 + }, + { + "epoch": 1.390625, + "grad_norm": 0.0019507119432091713, + "learning_rate": 3.11727834939056e-05, + "loss": 2.361126826144755e-05, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.61, + "memory/max_allocated (GiB)": 33.61, + "ppl": 1.00002, + "step": 356, + "tokens/total": 10770128, + "tokens/train_per_sec_per_gpu": 28.61, + "tokens/trainable": 162875 + }, + { + "epoch": 1.39453125, + "grad_norm": 0.013965644873678684, + "learning_rate": 3.092702317708967e-05, + "loss": 0.00010500443022465333, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 357, + "tokens/total": 10800432, + "tokens/train_per_sec_per_gpu": 35.43, + "tokens/trainable": 163329 + }, + { + "epoch": 1.3984375, + "grad_norm": 0.08449137210845947, + "learning_rate": 3.0682264635101276e-05, + "loss": 0.00026978645473718643, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00027, + "step": 358, + "tokens/total": 10830592, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 163778 + }, + { + "epoch": 1.40234375, + "grad_norm": 0.07536502927541733, + "learning_rate": 3.0438518053342407e-05, + "loss": 0.001174358418211341, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00118, + "step": 359, + "tokens/total": 10861088, + "tokens/train_per_sec_per_gpu": 32.17, + "tokens/trainable": 164219 + }, + { + "epoch": 1.40625, + "grad_norm": 0.024741269648075104, + "learning_rate": 3.0195793575103266e-05, + "loss": 0.00027250737184658647, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00027, + "step": 360, + "tokens/total": 10891552, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 164666 + }, + { + "epoch": 1.41015625, + "grad_norm": 0.6777807474136353, + "learning_rate": 2.9954101301140146e-05, + "loss": 0.016698025166988373, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.01684, + "step": 361, + "tokens/total": 10921776, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 165128 + }, + { + "epoch": 1.4140625, + "grad_norm": 0.2795168161392212, + "learning_rate": 2.9713451289255123e-05, + "loss": 0.00555866863578558, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00557, + "step": 362, + "tokens/total": 10952288, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 165614 + }, + { + "epoch": 1.41796875, + "grad_norm": 0.19862814247608185, + "learning_rate": 2.9473853553877484e-05, + "loss": 0.003639942966401577, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00365, + "step": 363, + "tokens/total": 10982496, + "tokens/train_per_sec_per_gpu": 34.59, + "tokens/trainable": 166062 + }, + { + "epoch": 1.421875, + "grad_norm": 0.020105330273509026, + "learning_rate": 2.9235318065647e-05, + "loss": 0.0001461556676076725, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 364, + "tokens/total": 11012704, + "tokens/train_per_sec_per_gpu": 34.55, + "tokens/trainable": 166513 + }, + { + "epoch": 1.42578125, + "grad_norm": 0.09536039084196091, + "learning_rate": 2.8997854750998964e-05, + "loss": 0.0006832765066064894, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00068, + "step": 365, + "tokens/total": 11043024, + "tokens/train_per_sec_per_gpu": 36.77, + "tokens/trainable": 167001 + }, + { + "epoch": 1.4296875, + "grad_norm": 0.015557884238660336, + "learning_rate": 2.8761473491751258e-05, + "loss": 0.00021839249529875815, + "memory/device_reserved (GiB)": 35.84, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00022, + "step": 366, + "tokens/total": 11073392, + "tokens/train_per_sec_per_gpu": 30.62, + "tokens/trainable": 167450 + }, + { + "epoch": 1.43359375, + "grad_norm": 0.003135454375296831, + "learning_rate": 2.8526184124692883e-05, + "loss": 4.49724102509208e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00004, + "step": 367, + "tokens/total": 11103936, + "tokens/train_per_sec_per_gpu": 31.79, + "tokens/trainable": 167895 + }, + { + "epoch": 1.4375, + "grad_norm": 0.03941775858402252, + "learning_rate": 2.829199644117484e-05, + "loss": 0.0003685755073092878, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00037, + "step": 368, + "tokens/total": 11134256, + "tokens/train_per_sec_per_gpu": 36.0, + "tokens/trainable": 168338 + }, + { + "epoch": 1.44140625, + "grad_norm": 0.07836400717496872, + "learning_rate": 2.8058920186702553e-05, + "loss": 0.0015341172693297267, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00154, + "step": 369, + "tokens/total": 11164768, + "tokens/train_per_sec_per_gpu": 32.61, + "tokens/trainable": 168813 + }, + { + "epoch": 1.4453125, + "grad_norm": 0.47108760476112366, + "learning_rate": 2.782696506053033e-05, + "loss": 0.009904367849230766, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00995, + "step": 370, + "tokens/total": 11195136, + "tokens/train_per_sec_per_gpu": 37.27, + "tokens/trainable": 169315 + }, + { + "epoch": 1.44921875, + "grad_norm": 0.03302016109228134, + "learning_rate": 2.7596140715257824e-05, + "loss": 0.0005742749781347811, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00057, + "step": 371, + "tokens/total": 11225680, + "tokens/train_per_sec_per_gpu": 35.47, + "tokens/trainable": 169756 + }, + { + "epoch": 1.453125, + "grad_norm": 0.012702704407274723, + "learning_rate": 2.7366456756428184e-05, + "loss": 0.0001972567697521299, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.0002, + "step": 372, + "tokens/total": 11256112, + "tokens/train_per_sec_per_gpu": 33.3, + "tokens/trainable": 170222 + }, + { + "epoch": 1.45703125, + "grad_norm": 0.005610617808997631, + "learning_rate": 2.7137922742128486e-05, + "loss": 8.283223723992705e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00008, + "step": 373, + "tokens/total": 11286304, + "tokens/train_per_sec_per_gpu": 33.6, + "tokens/trainable": 170687 + }, + { + "epoch": 1.4609375, + "grad_norm": 0.13766004145145416, + "learning_rate": 2.691054818259188e-05, + "loss": 0.00119793601334095, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.0012, + "step": 374, + "tokens/total": 11316800, + "tokens/train_per_sec_per_gpu": 32.43, + "tokens/trainable": 171163 + }, + { + "epoch": 1.46484375, + "grad_norm": 0.4246460795402527, + "learning_rate": 2.6684342539801933e-05, + "loss": 0.007628333754837513, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00766, + "step": 375, + "tokens/total": 11346944, + "tokens/train_per_sec_per_gpu": 34.48, + "tokens/trainable": 171626 + }, + { + "epoch": 1.46875, + "grad_norm": 0.008275235071778297, + "learning_rate": 2.645931522709877e-05, + "loss": 0.00011664340127026662, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00012, + "step": 376, + "tokens/total": 11377376, + "tokens/train_per_sec_per_gpu": 28.49, + "tokens/trainable": 172043 + }, + { + "epoch": 1.47265625, + "grad_norm": 0.01685458980500698, + "learning_rate": 2.6235475608787365e-05, + "loss": 0.00018345039279665798, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00018, + "step": 377, + "tokens/total": 11407680, + "tokens/train_per_sec_per_gpu": 36.55, + "tokens/trainable": 172519 + }, + { + "epoch": 1.4765625, + "grad_norm": 0.33050504326820374, + "learning_rate": 2.6012832999747916e-05, + "loss": 0.003047082107514143, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00305, + "step": 378, + "tokens/total": 11437712, + "tokens/train_per_sec_per_gpu": 34.88, + "tokens/trainable": 172977 + }, + { + "epoch": 1.48046875, + "grad_norm": 0.1227385401725769, + "learning_rate": 2.579139666504821e-05, + "loss": 0.0010191744659096003, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00102, + "step": 379, + "tokens/total": 11468176, + "tokens/train_per_sec_per_gpu": 36.22, + "tokens/trainable": 173461 + }, + { + "epoch": 1.484375, + "grad_norm": 0.0017978748073801398, + "learning_rate": 2.557117581955798e-05, + "loss": 4.36538175563328e-05, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00004, + "step": 380, + "tokens/total": 11498352, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 173926 + }, + { + "epoch": 1.48828125, + "grad_norm": 0.2067205160856247, + "learning_rate": 2.5352179627565532e-05, + "loss": 0.005216958001255989, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00523, + "step": 381, + "tokens/total": 11528768, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 174397 + }, + { + "epoch": 1.4921875, + "grad_norm": 0.0806688517332077, + "learning_rate": 2.5134417202396277e-05, + "loss": 0.0011719370959326625, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00117, + "step": 382, + "tokens/total": 11558976, + "tokens/train_per_sec_per_gpu": 33.27, + "tokens/trainable": 174852 + }, + { + "epoch": 1.49609375, + "grad_norm": 0.027668939903378487, + "learning_rate": 2.491789760603361e-05, + "loss": 0.00021959877631161362, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00022, + "step": 383, + "tokens/total": 11589520, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 175293 + }, + { + "epoch": 1.5, + "grad_norm": 0.12873099744319916, + "learning_rate": 2.4702629848741764e-05, + "loss": 0.0015978836454451084, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0016, + "step": 384, + "tokens/total": 11619760, + "tokens/train_per_sec_per_gpu": 32.87, + "tokens/trainable": 175731 + }, + { + "epoch": 1.50390625, + "grad_norm": 0.037335313856601715, + "learning_rate": 2.4488622888690785e-05, + "loss": 0.0005178801948204637, + "memory/device_reserved (GiB)": 35.89, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00052, + "step": 385, + "tokens/total": 11650192, + "tokens/train_per_sec_per_gpu": 35.62, + "tokens/trainable": 176215 + }, + { + "epoch": 1.5078125, + "grad_norm": 0.06761232763528824, + "learning_rate": 2.427588563158384e-05, + "loss": 0.0005836813943460584, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00058, + "step": 386, + "tokens/total": 11680672, + "tokens/train_per_sec_per_gpu": 34.8, + "tokens/trainable": 176663 + }, + { + "epoch": 1.51171875, + "grad_norm": 0.0501842126250267, + "learning_rate": 2.406442693028651e-05, + "loss": 0.0005241117323748767, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00052, + "step": 387, + "tokens/total": 11710880, + "tokens/train_per_sec_per_gpu": 33.11, + "tokens/trainable": 177127 + }, + { + "epoch": 1.515625, + "grad_norm": 0.08328797668218613, + "learning_rate": 2.3854255584458547e-05, + "loss": 0.00114786671474576, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00115, + "step": 388, + "tokens/total": 11741408, + "tokens/train_per_sec_per_gpu": 36.56, + "tokens/trainable": 177613 + }, + { + "epoch": 1.51953125, + "grad_norm": 0.020661214366555214, + "learning_rate": 2.3645380340187508e-05, + "loss": 0.00030904496088624, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00031, + "step": 389, + "tokens/total": 11771712, + "tokens/train_per_sec_per_gpu": 33.47, + "tokens/trainable": 178062 + }, + { + "epoch": 1.5234375, + "grad_norm": 0.13835693895816803, + "learning_rate": 2.3437809889624914e-05, + "loss": 0.0014695629943162203, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00147, + "step": 390, + "tokens/total": 11801872, + "tokens/train_per_sec_per_gpu": 36.96, + "tokens/trainable": 178535 + }, + { + "epoch": 1.52734375, + "grad_norm": 0.061385851353406906, + "learning_rate": 2.3231552870624487e-05, + "loss": 0.0007466091774404049, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00075, + "step": 391, + "tokens/total": 11832480, + "tokens/train_per_sec_per_gpu": 34.42, + "tokens/trainable": 179004 + }, + { + "epoch": 1.53125, + "grad_norm": 0.003179469145834446, + "learning_rate": 2.3026617866382657e-05, + "loss": 5.848719956702553e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00006, + "step": 392, + "tokens/total": 11862672, + "tokens/train_per_sec_per_gpu": 33.49, + "tokens/trainable": 179461 + }, + { + "epoch": 1.53515625, + "grad_norm": 0.021778129041194916, + "learning_rate": 2.2823013405081507e-05, + "loss": 0.00023812934523448348, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00024, + "step": 393, + "tokens/total": 11893024, + "tokens/train_per_sec_per_gpu": 33.22, + "tokens/trainable": 179893 + }, + { + "epoch": 1.5390625, + "grad_norm": 0.10550826787948608, + "learning_rate": 2.2620747959533722e-05, + "loss": 0.0007996093481779099, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.0008, + "step": 394, + "tokens/total": 11923328, + "tokens/train_per_sec_per_gpu": 34.41, + "tokens/trainable": 180344 + }, + { + "epoch": 1.54296875, + "grad_norm": 0.003161423373967409, + "learning_rate": 2.2419829946830123e-05, + "loss": 5.3679388656746596e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00005, + "step": 395, + "tokens/total": 11953792, + "tokens/train_per_sec_per_gpu": 32.21, + "tokens/trainable": 180757 + }, + { + "epoch": 1.546875, + "grad_norm": 0.05268798768520355, + "learning_rate": 2.2220267727989325e-05, + "loss": 0.0009187876130454242, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00092, + "step": 396, + "tokens/total": 11984208, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 181228 + }, + { + "epoch": 1.55078125, + "grad_norm": 0.0017847128910943866, + "learning_rate": 2.202206960760984e-05, + "loss": 3.819329140242189e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00004, + "step": 397, + "tokens/total": 12014608, + "tokens/train_per_sec_per_gpu": 33.66, + "tokens/trainable": 181716 + }, + { + "epoch": 1.5546875, + "grad_norm": 0.1685003638267517, + "learning_rate": 2.182524383352446e-05, + "loss": 0.0024524193722754717, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00246, + "step": 398, + "tokens/total": 12045088, + "tokens/train_per_sec_per_gpu": 36.01, + "tokens/trainable": 182152 + }, + { + "epoch": 1.55859375, + "grad_norm": 0.11644069850444794, + "learning_rate": 2.1629798596457056e-05, + "loss": 0.001032137661240995, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00103, + "step": 399, + "tokens/total": 12075280, + "tokens/train_per_sec_per_gpu": 35.37, + "tokens/trainable": 182623 + }, + { + "epoch": 1.5625, + "grad_norm": 0.004440188407897949, + "learning_rate": 2.1435742029681725e-05, + "loss": 7.53562999307178e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00008, + "step": 400, + "tokens/total": 12105792, + "tokens/train_per_sec_per_gpu": 26.92, + "tokens/trainable": 183039 + }, + { + "epoch": 1.56640625, + "grad_norm": 0.06607124954462051, + "learning_rate": 2.124308220868431e-05, + "loss": 0.0008609758224338293, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.71, + "memory/max_allocated (GiB)": 33.71, + "ppl": 1.00086, + "step": 401, + "tokens/total": 12135680, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 183487 + }, + { + "epoch": 1.5703125, + "grad_norm": 0.0036999934818595648, + "learning_rate": 2.105182715082638e-05, + "loss": 6.075722922105342e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00006, + "step": 402, + "tokens/total": 12165920, + "tokens/train_per_sec_per_gpu": 39.03, + "tokens/trainable": 183987 + }, + { + "epoch": 1.57421875, + "grad_norm": 0.010572529397904873, + "learning_rate": 2.0861984815011552e-05, + "loss": 0.0001372473343508318, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00014, + "step": 403, + "tokens/total": 12196080, + "tokens/train_per_sec_per_gpu": 32.86, + "tokens/trainable": 184459 + }, + { + "epoch": 1.578125, + "grad_norm": 0.3721093237400055, + "learning_rate": 2.0673563101354323e-05, + "loss": 0.003381735645234585, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00339, + "step": 404, + "tokens/total": 12226128, + "tokens/train_per_sec_per_gpu": 31.95, + "tokens/trainable": 184887 + }, + { + "epoch": 1.58203125, + "grad_norm": 0.08975216001272202, + "learning_rate": 2.0486569850851317e-05, + "loss": 0.0008267756784334779, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.00083, + "step": 405, + "tokens/total": 12256176, + "tokens/train_per_sec_per_gpu": 29.59, + "tokens/trainable": 185320 + }, + { + "epoch": 1.5859375, + "grad_norm": 0.4965755343437195, + "learning_rate": 2.0301012845054956e-05, + "loss": 0.012792397290468216, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.01287, + "step": 406, + "tokens/total": 12286416, + "tokens/train_per_sec_per_gpu": 31.77, + "tokens/trainable": 185774 + }, + { + "epoch": 1.58984375, + "grad_norm": 0.0017864442197605968, + "learning_rate": 2.011689980574966e-05, + "loss": 3.962808841606602e-05, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00004, + "step": 407, + "tokens/total": 12316720, + "tokens/train_per_sec_per_gpu": 37.15, + "tokens/trainable": 186233 + }, + { + "epoch": 1.59375, + "grad_norm": 0.006781951989978552, + "learning_rate": 1.993423839463052e-05, + "loss": 6.552625563926995e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00007, + "step": 408, + "tokens/total": 12347072, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 186671 + }, + { + "epoch": 1.59765625, + "grad_norm": 0.02034713886678219, + "learning_rate": 1.975303621298445e-05, + "loss": 0.00025357812410220504, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00025, + "step": 409, + "tokens/total": 12377504, + "tokens/train_per_sec_per_gpu": 34.29, + "tokens/trainable": 187102 + }, + { + "epoch": 1.6015625, + "grad_norm": 0.003248689230531454, + "learning_rate": 1.957330080137385e-05, + "loss": 5.506081652129069e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00006, + "step": 410, + "tokens/total": 12408048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 187549 + }, + { + "epoch": 1.60546875, + "grad_norm": 0.16686522960662842, + "learning_rate": 1.9395039639322864e-05, + "loss": 0.0007388527155853808, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00074, + "step": 411, + "tokens/total": 12438256, + "tokens/train_per_sec_per_gpu": 34.61, + "tokens/trainable": 187990 + }, + { + "epoch": 1.609375, + "grad_norm": 0.005520866252481937, + "learning_rate": 1.9218260145006073e-05, + "loss": 0.00011023284605471417, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00011, + "step": 412, + "tokens/total": 12466608, + "tokens/train_per_sec_per_gpu": 39.21, + "tokens/trainable": 188426 + }, + { + "epoch": 1.61328125, + "grad_norm": 0.1133890226483345, + "learning_rate": 1.904296967493982e-05, + "loss": 0.0016134778270497918, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00161, + "step": 413, + "tokens/total": 12497136, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 188877 + }, + { + "epoch": 1.6171875, + "grad_norm": 0.09548613429069519, + "learning_rate": 1.8869175523676064e-05, + "loss": 0.00023090933973435313, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00023, + "step": 414, + "tokens/total": 12527120, + "tokens/train_per_sec_per_gpu": 37.84, + "tokens/trainable": 189346 + }, + { + "epoch": 1.62109375, + "grad_norm": 0.0033725884277373552, + "learning_rate": 1.869688492349885e-05, + "loss": 4.779225855600089e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00005, + "step": 415, + "tokens/total": 12557248, + "tokens/train_per_sec_per_gpu": 32.34, + "tokens/trainable": 189759 + }, + { + "epoch": 1.625, + "grad_norm": 0.0031356913968920708, + "learning_rate": 1.85261050441233e-05, + "loss": 4.390432150103152e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00004, + "step": 416, + "tokens/total": 12587728, + "tokens/train_per_sec_per_gpu": 34.5, + "tokens/trainable": 190225 + }, + { + "epoch": 1.62890625, + "grad_norm": 0.0025502180214971304, + "learning_rate": 1.8356842992397304e-05, + "loss": 5.062075433670543e-05, + "memory/device_reserved (GiB)": 35.96, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00005, + "step": 417, + "tokens/total": 12618160, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 190717 + }, + { + "epoch": 1.6328125, + "grad_norm": 0.006292750593274832, + "learning_rate": 1.8189105812005714e-05, + "loss": 9.369729377795011e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00009, + "step": 418, + "tokens/total": 12648672, + "tokens/train_per_sec_per_gpu": 31.28, + "tokens/trainable": 191135 + }, + { + "epoch": 1.63671875, + "grad_norm": 0.1299043893814087, + "learning_rate": 1.802290048317732e-05, + "loss": 0.002147798193618655, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00215, + "step": 419, + "tokens/total": 12678944, + "tokens/train_per_sec_per_gpu": 31.86, + "tokens/trainable": 191559 + }, + { + "epoch": 1.640625, + "grad_norm": 0.017163539305329323, + "learning_rate": 1.785823392239424e-05, + "loss": 0.0001932488230522722, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00019, + "step": 420, + "tokens/total": 12709408, + "tokens/train_per_sec_per_gpu": 36.98, + "tokens/trainable": 192033 + }, + { + "epoch": 1.64453125, + "grad_norm": 0.0045894598588347435, + "learning_rate": 1.7695112982104225e-05, + "loss": 5.246631917543709e-05, + "memory/device_reserved (GiB)": 35.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 421, + "tokens/total": 12739632, + "tokens/train_per_sec_per_gpu": 31.53, + "tokens/trainable": 192482 + }, + { + "epoch": 1.6484375, + "grad_norm": 0.048289310187101364, + "learning_rate": 1.7533544450435433e-05, + "loss": 0.0002731278946157545, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00027, + "step": 422, + "tokens/total": 12770224, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 192937 + }, + { + "epoch": 1.65234375, + "grad_norm": 0.033917587250471115, + "learning_rate": 1.7373535050913946e-05, + "loss": 0.0003945066418964416, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 423, + "tokens/total": 12800640, + "tokens/train_per_sec_per_gpu": 37.81, + "tokens/trainable": 193444 + }, + { + "epoch": 1.65625, + "grad_norm": 0.011636070907115936, + "learning_rate": 1.721509144218405e-05, + "loss": 0.00016518094344064593, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00017, + "step": 424, + "tokens/total": 12831104, + "tokens/train_per_sec_per_gpu": 36.92, + "tokens/trainable": 193925 + }, + { + "epoch": 1.66015625, + "grad_norm": 0.06534689664840698, + "learning_rate": 1.705822021773101e-05, + "loss": 0.0008454776252619922, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00085, + "step": 425, + "tokens/total": 12861296, + "tokens/train_per_sec_per_gpu": 29.68, + "tokens/trainable": 194342 + }, + { + "epoch": 1.6640625, + "grad_norm": 0.07199558615684509, + "learning_rate": 1.69029279056068e-05, + "loss": 0.0007431323174387217, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00074, + "step": 426, + "tokens/total": 12891504, + "tokens/train_per_sec_per_gpu": 35.4, + "tokens/trainable": 194830 + }, + { + "epoch": 1.66796875, + "grad_norm": 0.030936896800994873, + "learning_rate": 1.6749220968158415e-05, + "loss": 0.00027934886747971177, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00028, + "step": 427, + "tokens/total": 12921840, + "tokens/train_per_sec_per_gpu": 34.47, + "tokens/trainable": 195296 + }, + { + "epoch": 1.671875, + "grad_norm": 0.0030303315725177526, + "learning_rate": 1.659710580175893e-05, + "loss": 4.713955058832653e-05, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00005, + "step": 428, + "tokens/total": 12952224, + "tokens/train_per_sec_per_gpu": 32.33, + "tokens/trainable": 195714 + }, + { + "epoch": 1.67578125, + "grad_norm": 0.055711835622787476, + "learning_rate": 1.644658873654133e-05, + "loss": 0.0008513920474797487, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.00085, + "step": 429, + "tokens/total": 12982416, + "tokens/train_per_sec_per_gpu": 33.31, + "tokens/trainable": 196179 + }, + { + "epoch": 1.6796875, + "grad_norm": 0.014902536757290363, + "learning_rate": 1.629767603613508e-05, + "loss": 0.00024280000070575625, + "memory/device_reserved (GiB)": 35.83, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00024, + "step": 430, + "tokens/total": 13012800, + "tokens/train_per_sec_per_gpu": 32.82, + "tokens/trainable": 196660 + }, + { + "epoch": 1.68359375, + "grad_norm": 0.003592823399230838, + "learning_rate": 1.615037389740547e-05, + "loss": 5.272522321320139e-05, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.00005, + "step": 431, + "tokens/total": 13043392, + "tokens/train_per_sec_per_gpu": 31.51, + "tokens/trainable": 197091 + }, + { + "epoch": 1.6875, + "grad_norm": 0.03395218774676323, + "learning_rate": 1.600468845019576e-05, + "loss": 0.0003875850234180689, + "memory/device_reserved (GiB)": 36.35, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00039, + "step": 432, + "tokens/total": 13073984, + "tokens/train_per_sec_per_gpu": 39.07, + "tokens/trainable": 197589 + }, + { + "epoch": 1.69140625, + "grad_norm": 0.06874402612447739, + "learning_rate": 1.5860625757072092e-05, + "loss": 0.0007863138453103602, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.00079, + "step": 433, + "tokens/total": 13104512, + "tokens/train_per_sec_per_gpu": 33.15, + "tokens/trainable": 198047 + }, + { + "epoch": 1.6953125, + "grad_norm": 0.04059433192014694, + "learning_rate": 1.571819181307116e-05, + "loss": 0.00044966390123590827, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00045, + "step": 434, + "tokens/total": 13134656, + "tokens/train_per_sec_per_gpu": 32.89, + "tokens/trainable": 198465 + }, + { + "epoch": 1.69921875, + "grad_norm": 0.1531708687543869, + "learning_rate": 1.557739254545075e-05, + "loss": 0.0016073889564722776, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00161, + "step": 435, + "tokens/total": 13165248, + "tokens/train_per_sec_per_gpu": 34.75, + "tokens/trainable": 198937 + }, + { + "epoch": 1.703125, + "grad_norm": 0.03812495246529579, + "learning_rate": 1.543823381344311e-05, + "loss": 0.0005045874277129769, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.0005, + "step": 436, + "tokens/total": 13195648, + "tokens/train_per_sec_per_gpu": 33.7, + "tokens/trainable": 199385 + }, + { + "epoch": 1.70703125, + "grad_norm": 0.3262580633163452, + "learning_rate": 1.5300721408011114e-05, + "loss": 0.008203462697565556, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00824, + "step": 437, + "tokens/total": 13226176, + "tokens/train_per_sec_per_gpu": 34.78, + "tokens/trainable": 199869 + }, + { + "epoch": 1.7109375, + "grad_norm": 0.0010299552232027054, + "learning_rate": 1.5164861051607254e-05, + "loss": 1.8928842109744437e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 438, + "tokens/total": 13256480, + "tokens/train_per_sec_per_gpu": 35.38, + "tokens/trainable": 200377 + }, + { + "epoch": 1.71484375, + "grad_norm": 0.02154897339642048, + "learning_rate": 1.5030658397935521e-05, + "loss": 0.00018170180555898696, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00018, + "step": 439, + "tokens/total": 13286528, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 200824 + }, + { + "epoch": 1.71875, + "grad_norm": 0.02101851999759674, + "learning_rate": 1.4898119031716104e-05, + "loss": 0.00032509121228940785, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00033, + "step": 440, + "tokens/total": 13317024, + "tokens/train_per_sec_per_gpu": 35.51, + "tokens/trainable": 201297 + }, + { + "epoch": 1.72265625, + "grad_norm": 0.007308815140277147, + "learning_rate": 1.476724846845306e-05, + "loss": 0.00010265046876156703, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0001, + "step": 441, + "tokens/total": 13347440, + "tokens/train_per_sec_per_gpu": 34.67, + "tokens/trainable": 201764 + }, + { + "epoch": 1.7265625, + "grad_norm": 0.04583185538649559, + "learning_rate": 1.463805215420471e-05, + "loss": 0.0005665539065375924, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00057, + "step": 442, + "tokens/total": 13377664, + "tokens/train_per_sec_per_gpu": 30.84, + "tokens/trainable": 202204 + }, + { + "epoch": 1.73046875, + "grad_norm": 0.2436366230249405, + "learning_rate": 1.451053546535705e-05, + "loss": 0.004352548159658909, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00436, + "step": 443, + "tokens/total": 13408080, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 202642 + }, + { + "epoch": 1.734375, + "grad_norm": 0.015320447273552418, + "learning_rate": 1.438470370840001e-05, + "loss": 0.0002279828186146915, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.64, + "memory/max_allocated (GiB)": 33.64, + "ppl": 1.00023, + "step": 444, + "tokens/total": 13437824, + "tokens/train_per_sec_per_gpu": 31.22, + "tokens/trainable": 203050 + }, + { + "epoch": 1.73828125, + "grad_norm": 0.0014951409539207816, + "learning_rate": 1.4260562119706606e-05, + "loss": 2.6277737561031245e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00003, + "step": 445, + "tokens/total": 13468032, + "tokens/train_per_sec_per_gpu": 34.63, + "tokens/trainable": 203509 + }, + { + "epoch": 1.7421875, + "grad_norm": 0.009211640805006027, + "learning_rate": 1.413811586531508e-05, + "loss": 6.653230229858309e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00007, + "step": 446, + "tokens/total": 13498096, + "tokens/train_per_sec_per_gpu": 35.78, + "tokens/trainable": 203976 + }, + { + "epoch": 1.74609375, + "grad_norm": 0.15356089174747467, + "learning_rate": 1.4017370040713884e-05, + "loss": 0.0011966564925387502, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.0012, + "step": 447, + "tokens/total": 13528288, + "tokens/train_per_sec_per_gpu": 37.97, + "tokens/trainable": 204412 + }, + { + "epoch": 1.75, + "grad_norm": 0.008474756963551044, + "learning_rate": 1.3898329670629645e-05, + "loss": 9.093034896068275e-05, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00009, + "step": 448, + "tokens/total": 13558752, + "tokens/train_per_sec_per_gpu": 36.02, + "tokens/trainable": 204868 + }, + { + "epoch": 1.75390625, + "grad_norm": 0.015406596474349499, + "learning_rate": 1.3780999708818058e-05, + "loss": 0.0001275191898457706, + "memory/device_reserved (GiB)": 36.8, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.00013, + "step": 449, + "tokens/total": 13589136, + "tokens/train_per_sec_per_gpu": 34.11, + "tokens/trainable": 205338 + }, + { + "epoch": 1.7578125, + "grad_norm": 0.12964943051338196, + "learning_rate": 1.3665385037857758e-05, + "loss": 0.0008908645249903202, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00089, + "step": 450, + "tokens/total": 13619472, + "tokens/train_per_sec_per_gpu": 31.08, + "tokens/trainable": 205764 + }, + { + "epoch": 1.76171875, + "grad_norm": 0.010256104171276093, + "learning_rate": 1.3551490468947126e-05, + "loss": 0.00014732024283148348, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00015, + "step": 451, + "tokens/total": 13649872, + "tokens/train_per_sec_per_gpu": 32.08, + "tokens/trainable": 206201 + }, + { + "epoch": 1.765625, + "grad_norm": 0.006175735965371132, + "learning_rate": 1.3439320741704075e-05, + "loss": 9.223521192325279e-05, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00009, + "step": 452, + "tokens/total": 13680064, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 206657 + }, + { + "epoch": 1.76953125, + "grad_norm": 0.060654036700725555, + "learning_rate": 1.3328880523968808e-05, + "loss": 0.0005743277142755687, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00057, + "step": 453, + "tokens/total": 13710592, + "tokens/train_per_sec_per_gpu": 37.59, + "tokens/trainable": 207138 + }, + { + "epoch": 1.7734375, + "grad_norm": 0.003730883589014411, + "learning_rate": 1.3220174411609587e-05, + "loss": 5.224360211286694e-05, + "memory/device_reserved (GiB)": 35.7, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00005, + "step": 454, + "tokens/total": 13740960, + "tokens/train_per_sec_per_gpu": 36.25, + "tokens/trainable": 207583 + }, + { + "epoch": 1.77734375, + "grad_norm": 0.09384409338235855, + "learning_rate": 1.3113206928331471e-05, + "loss": 0.0008255710126832128, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00083, + "step": 455, + "tokens/total": 13771296, + "tokens/train_per_sec_per_gpu": 33.58, + "tokens/trainable": 208025 + }, + { + "epoch": 1.78125, + "grad_norm": 0.11239906400442123, + "learning_rate": 1.300798252548806e-05, + "loss": 0.0013365334598347545, + "memory/device_reserved (GiB)": 35.78, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.00134, + "step": 456, + "tokens/total": 13801664, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 208453 + }, + { + "epoch": 1.78515625, + "grad_norm": 0.007983297109603882, + "learning_rate": 1.2904505581896265e-05, + "loss": 6.189515988808125e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00006, + "step": 457, + "tokens/total": 13831936, + "tokens/train_per_sec_per_gpu": 35.89, + "tokens/trainable": 208947 + }, + { + "epoch": 1.7890625, + "grad_norm": 0.0063696797005832195, + "learning_rate": 1.2802780403654082e-05, + "loss": 9.090925595955923e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00009, + "step": 458, + "tokens/total": 13862160, + "tokens/train_per_sec_per_gpu": 35.33, + "tokens/trainable": 209435 + }, + { + "epoch": 1.79296875, + "grad_norm": 0.008154391311109066, + "learning_rate": 1.2702811223961408e-05, + "loss": 7.762061432003975e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00008, + "step": 459, + "tokens/total": 13892320, + "tokens/train_per_sec_per_gpu": 33.98, + "tokens/trainable": 209885 + }, + { + "epoch": 1.796875, + "grad_norm": 0.015743529424071312, + "learning_rate": 1.2604602202943861e-05, + "loss": 0.00018653822189662606, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.00019, + "step": 460, + "tokens/total": 13922752, + "tokens/train_per_sec_per_gpu": 34.08, + "tokens/trainable": 210344 + }, + { + "epoch": 1.80078125, + "grad_norm": 0.011715077795088291, + "learning_rate": 1.2508157427479686e-05, + "loss": 0.00016241846606135368, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.65, + "memory/max_allocated (GiB)": 33.65, + "ppl": 1.00016, + "step": 461, + "tokens/total": 13952816, + "tokens/train_per_sec_per_gpu": 32.19, + "tokens/trainable": 210814 + }, + { + "epoch": 1.8046875, + "grad_norm": 0.04297679290175438, + "learning_rate": 1.2413480911029655e-05, + "loss": 0.0005623188917525113, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.38, + "memory/max_allocated (GiB)": 33.38, + "ppl": 1.00056, + "step": 462, + "tokens/total": 13981040, + "tokens/train_per_sec_per_gpu": 30.87, + "tokens/trainable": 211238 + }, + { + "epoch": 1.80859375, + "grad_norm": 0.004571136087179184, + "learning_rate": 1.2320576593470082e-05, + "loss": 4.7571604227414355e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.79, + "memory/max_allocated (GiB)": 33.79, + "ppl": 1.00005, + "step": 463, + "tokens/total": 14011248, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 211700 + }, + { + "epoch": 1.8125, + "grad_norm": 0.0022084659431129694, + "learning_rate": 1.2229448340928828e-05, + "loss": 1.850002445280552e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.00002, + "step": 464, + "tokens/total": 14041504, + "tokens/train_per_sec_per_gpu": 29.47, + "tokens/trainable": 212138 + }, + { + "epoch": 1.81640625, + "grad_norm": 0.08595507591962814, + "learning_rate": 1.2140099945624458e-05, + "loss": 0.0008459303062409163, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00085, + "step": 465, + "tokens/total": 14071728, + "tokens/train_per_sec_per_gpu": 33.14, + "tokens/trainable": 212623 + }, + { + "epoch": 1.8203125, + "grad_norm": 0.0007654759101569653, + "learning_rate": 1.205253512570841e-05, + "loss": 1.6132753444253467e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00002, + "step": 466, + "tokens/total": 14101824, + "tokens/train_per_sec_per_gpu": 33.37, + "tokens/trainable": 213062 + }, + { + "epoch": 1.82421875, + "grad_norm": 0.0005283593782223761, + "learning_rate": 1.1966757525110255e-05, + "loss": 1.0958694474538788e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00001, + "step": 467, + "tokens/total": 14132080, + "tokens/train_per_sec_per_gpu": 32.03, + "tokens/trainable": 213501 + }, + { + "epoch": 1.828125, + "grad_norm": 0.0003768271126318723, + "learning_rate": 1.1882770713386095e-05, + "loss": 1.1069126230722759e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.00001, + "step": 468, + "tokens/total": 14162112, + "tokens/train_per_sec_per_gpu": 35.13, + "tokens/trainable": 213942 + }, + { + "epoch": 1.83203125, + "grad_norm": 0.010821828618645668, + "learning_rate": 1.180057818556998e-05, + "loss": 7.607591396663338e-05, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00008, + "step": 469, + "tokens/total": 14192784, + "tokens/train_per_sec_per_gpu": 37.33, + "tokens/trainable": 214410 + }, + { + "epoch": 1.8359375, + "grad_norm": 0.008473122492432594, + "learning_rate": 1.1720183362028494e-05, + "loss": 0.00010939891217276454, + "memory/device_reserved (GiB)": 35.82, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00011, + "step": 470, + "tokens/total": 14221008, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 214844 + }, + { + "epoch": 1.83984375, + "grad_norm": 0.0007591558387503028, + "learning_rate": 1.1641589588318387e-05, + "loss": 1.7881218809634447e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00002, + "step": 471, + "tokens/total": 14251536, + "tokens/train_per_sec_per_gpu": 35.96, + "tokens/trainable": 215336 + }, + { + "epoch": 1.84375, + "grad_norm": 0.05926339700818062, + "learning_rate": 1.1564800135047418e-05, + "loss": 0.0005082086427137256, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00051, + "step": 472, + "tokens/total": 14281664, + "tokens/train_per_sec_per_gpu": 33.35, + "tokens/trainable": 215777 + }, + { + "epoch": 1.84765625, + "grad_norm": 0.12773646414279938, + "learning_rate": 1.148981819773816e-05, + "loss": 0.002721701515838504, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00273, + "step": 473, + "tokens/total": 14312048, + "tokens/train_per_sec_per_gpu": 35.25, + "tokens/trainable": 216222 + }, + { + "epoch": 1.8515625, + "grad_norm": 0.028139308094978333, + "learning_rate": 1.1416646896695086e-05, + "loss": 0.00013170376769267023, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00013, + "step": 474, + "tokens/total": 14342112, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 216662 + }, + { + "epoch": 1.85546875, + "grad_norm": 0.003854855429381132, + "learning_rate": 1.1345289276874717e-05, + "loss": 5.479517130879685e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00005, + "step": 475, + "tokens/total": 14372176, + "tokens/train_per_sec_per_gpu": 31.26, + "tokens/trainable": 217068 + }, + { + "epoch": 1.859375, + "grad_norm": 0.0007855606963858008, + "learning_rate": 1.1275748307758873e-05, + "loss": 1.5755449567222968e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00002, + "step": 476, + "tokens/total": 14402592, + "tokens/train_per_sec_per_gpu": 38.21, + "tokens/trainable": 217558 + }, + { + "epoch": 1.86328125, + "grad_norm": 0.004396429285407066, + "learning_rate": 1.1208026883231147e-05, + "loss": 5.772991426056251e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00006, + "step": 477, + "tokens/total": 14432816, + "tokens/train_per_sec_per_gpu": 35.98, + "tokens/trainable": 218025 + }, + { + "epoch": 1.8671875, + "grad_norm": 0.03181939572095871, + "learning_rate": 1.1142127821456433e-05, + "loss": 0.00036515528336167336, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00037, + "step": 478, + "tokens/total": 14463424, + "tokens/train_per_sec_per_gpu": 32.81, + "tokens/trainable": 218498 + }, + { + "epoch": 1.87109375, + "grad_norm": 0.004733108915388584, + "learning_rate": 1.1078053864763674e-05, + "loss": 4.52871281595435e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.00005, + "step": 479, + "tokens/total": 14493440, + "tokens/train_per_sec_per_gpu": 33.17, + "tokens/trainable": 218946 + }, + { + "epoch": 1.875, + "grad_norm": 0.0026765780057758093, + "learning_rate": 1.1015807679531756e-05, + "loss": 3.31170012941584e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.00003, + "step": 480, + "tokens/total": 14524000, + "tokens/train_per_sec_per_gpu": 34.52, + "tokens/trainable": 219387 + }, + { + "epoch": 1.87890625, + "grad_norm": 0.006714611314237118, + "learning_rate": 1.0955391856078528e-05, + "loss": 4.494276799960062e-05, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.00004, + "step": 481, + "tokens/total": 14554016, + "tokens/train_per_sec_per_gpu": 29.72, + "tokens/trainable": 219813 + }, + { + "epoch": 1.8828125, + "grad_norm": 0.019354507327079773, + "learning_rate": 1.0896808908553007e-05, + "loss": 0.0001470642746426165, + "memory/device_reserved (GiB)": 36.14, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00015, + "step": 482, + "tokens/total": 14584512, + "tokens/train_per_sec_per_gpu": 35.33, + "tokens/trainable": 220321 + }, + { + "epoch": 1.88671875, + "grad_norm": 0.005853369832038879, + "learning_rate": 1.0840061274830763e-05, + "loss": 3.408601332921535e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.8, + "memory/max_allocated (GiB)": 33.8, + "ppl": 1.00003, + "step": 483, + "tokens/total": 14614608, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 220751 + }, + { + "epoch": 1.890625, + "grad_norm": 0.0007542860112152994, + "learning_rate": 1.0785151316412473e-05, + "loss": 1.3063606274954509e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00001, + "step": 484, + "tokens/total": 14644784, + "tokens/train_per_sec_per_gpu": 35.32, + "tokens/trainable": 221193 + }, + { + "epoch": 1.89453125, + "grad_norm": 0.0012432060902938247, + "learning_rate": 1.0732081318325639e-05, + "loss": 2.1572890545940027e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00002, + "step": 485, + "tokens/total": 14674944, + "tokens/train_per_sec_per_gpu": 32.04, + "tokens/trainable": 221642 + }, + { + "epoch": 1.8984375, + "grad_norm": 0.015746230259537697, + "learning_rate": 1.0680853489029501e-05, + "loss": 0.00010861606278922409, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00011, + "step": 486, + "tokens/total": 14705456, + "tokens/train_per_sec_per_gpu": 31.93, + "tokens/trainable": 222086 + }, + { + "epoch": 1.90234375, + "grad_norm": 0.0022054179571568966, + "learning_rate": 1.0631469960323152e-05, + "loss": 2.593602766864933e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.67, + "memory/max_allocated (GiB)": 33.67, + "ppl": 1.00003, + "step": 487, + "tokens/total": 14735440, + "tokens/train_per_sec_per_gpu": 32.44, + "tokens/trainable": 222518 + }, + { + "epoch": 1.90625, + "grad_norm": 0.005776247940957546, + "learning_rate": 1.0583932787256783e-05, + "loss": 4.026497117592953e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00004, + "step": 488, + "tokens/total": 14765856, + "tokens/train_per_sec_per_gpu": 36.78, + "tokens/trainable": 223011 + }, + { + "epoch": 1.91015625, + "grad_norm": 0.008843912743031979, + "learning_rate": 1.0538243948046206e-05, + "loss": 6.14341624896042e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00006, + "step": 489, + "tokens/total": 14796128, + "tokens/train_per_sec_per_gpu": 32.29, + "tokens/trainable": 223464 + }, + { + "epoch": 1.9140625, + "grad_norm": 0.0005093133659102023, + "learning_rate": 1.0494405343990523e-05, + "loss": 9.642845725466032e-06, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.00001, + "step": 490, + "tokens/total": 14826240, + "tokens/train_per_sec_per_gpu": 29.84, + "tokens/trainable": 223908 + }, + { + "epoch": 1.91796875, + "grad_norm": 0.0012376867234706879, + "learning_rate": 1.0452418799392985e-05, + "loss": 1.869291918410454e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00002, + "step": 491, + "tokens/total": 14856512, + "tokens/train_per_sec_per_gpu": 33.34, + "tokens/trainable": 224365 + }, + { + "epoch": 1.921875, + "grad_norm": 0.01617741398513317, + "learning_rate": 1.0412286061485102e-05, + "loss": 0.0001638774701859802, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.01, + "memory/max_allocated (GiB)": 34.01, + "ppl": 1.00016, + "step": 492, + "tokens/total": 14886848, + "tokens/train_per_sec_per_gpu": 35.64, + "tokens/trainable": 224847 + }, + { + "epoch": 1.92578125, + "grad_norm": 0.573005735874176, + "learning_rate": 1.03740088003539e-05, + "loss": 0.013050932437181473, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01314, + "step": 493, + "tokens/total": 14917216, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 225329 + }, + { + "epoch": 1.9296875, + "grad_norm": 0.0012499141739681363, + "learning_rate": 1.0337588608872463e-05, + "loss": 1.8755235942080617e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.00002, + "step": 494, + "tokens/total": 14947648, + "tokens/train_per_sec_per_gpu": 33.23, + "tokens/trainable": 225778 + }, + { + "epoch": 1.93359375, + "grad_norm": 0.00981285609304905, + "learning_rate": 1.0303027002633622e-05, + "loss": 7.82605420681648e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00008, + "step": 495, + "tokens/total": 14977728, + "tokens/train_per_sec_per_gpu": 32.99, + "tokens/trainable": 226217 + }, + { + "epoch": 1.9375, + "grad_norm": 0.0030543936882168055, + "learning_rate": 1.0270325419886884e-05, + "loss": 3.3284988603554666e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.00003, + "step": 496, + "tokens/total": 15008032, + "tokens/train_per_sec_per_gpu": 35.29, + "tokens/trainable": 226689 + }, + { + "epoch": 1.94140625, + "grad_norm": 0.035057004541158676, + "learning_rate": 1.0239485221478599e-05, + "loss": 0.00040418561547994614, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.0004, + "step": 497, + "tokens/total": 15038528, + "tokens/train_per_sec_per_gpu": 32.2, + "tokens/trainable": 227112 + }, + { + "epoch": 1.9453125, + "grad_norm": 0.0012048225617036223, + "learning_rate": 1.0210507690795292e-05, + "loss": 2.2646674551651813e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00002, + "step": 498, + "tokens/total": 15068832, + "tokens/train_per_sec_per_gpu": 38.19, + "tokens/trainable": 227577 + }, + { + "epoch": 1.94921875, + "grad_norm": 0.0027389212045818567, + "learning_rate": 1.0183394033710305e-05, + "loss": 5.4524585721082985e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00005, + "step": 499, + "tokens/total": 15099136, + "tokens/train_per_sec_per_gpu": 37.11, + "tokens/trainable": 228053 + }, + { + "epoch": 1.953125, + "grad_norm": 0.0021910013165324926, + "learning_rate": 1.0158145378533583e-05, + "loss": 3.143951471429318e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.00003, + "step": 500, + "tokens/total": 15129488, + "tokens/train_per_sec_per_gpu": 31.33, + "tokens/trainable": 228503 + }, + { + "epoch": 1.95703125, + "grad_norm": 0.11515559256076813, + "learning_rate": 1.0134762775964726e-05, + "loss": 0.00122018251568079, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00122, + "step": 501, + "tokens/total": 15159840, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 228983 + }, + { + "epoch": 1.9609375, + "grad_norm": 0.008556320331990719, + "learning_rate": 1.0113247199049278e-05, + "loss": 4.6531327825505286e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 34.03, + "memory/max_allocated (GiB)": 34.03, + "ppl": 1.00005, + "step": 502, + "tokens/total": 15190624, + "tokens/train_per_sec_per_gpu": 31.47, + "tokens/trainable": 229415 + }, + { + "epoch": 1.96484375, + "grad_norm": 0.0009091690881177783, + "learning_rate": 1.0093599543138205e-05, + "loss": 1.879146475403104e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 503, + "tokens/total": 15221104, + "tokens/train_per_sec_per_gpu": 33.25, + "tokens/trainable": 229862 + }, + { + "epoch": 1.96875, + "grad_norm": 0.000701752258464694, + "learning_rate": 1.0075820625850675e-05, + "loss": 1.4095108781475574e-05, + "memory/device_reserved (GiB)": 36.15, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.00001, + "step": 504, + "tokens/total": 15251520, + "tokens/train_per_sec_per_gpu": 39.42, + "tokens/trainable": 230366 + }, + { + "epoch": 1.97265625, + "grad_norm": 0.0043007805943489075, + "learning_rate": 1.0059911187040013e-05, + "loss": 5.2763592975679785e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00005, + "step": 505, + "tokens/total": 15282080, + "tokens/train_per_sec_per_gpu": 38.02, + "tokens/trainable": 230862 + }, + { + "epoch": 1.9765625, + "grad_norm": 0.0014679227024316788, + "learning_rate": 1.0045871888762893e-05, + "loss": 2.3966062144609168e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.00002, + "step": 506, + "tokens/total": 15312304, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 231312 + }, + { + "epoch": 1.98046875, + "grad_norm": 0.2847539484500885, + "learning_rate": 1.003370331525184e-05, + "loss": 0.0026828020345419645, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00269, + "step": 507, + "tokens/total": 15342912, + "tokens/train_per_sec_per_gpu": 36.68, + "tokens/trainable": 231801 + }, + { + "epoch": 1.984375, + "grad_norm": 0.013710313476622105, + "learning_rate": 1.002340597289085e-05, + "loss": 0.00012580624024849385, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.33, + "memory/max_allocated (GiB)": 33.33, + "ppl": 1.00013, + "step": 508, + "tokens/total": 15371152, + "tokens/train_per_sec_per_gpu": 31.96, + "tokens/trainable": 232218 + }, + { + "epoch": 1.98828125, + "grad_norm": 0.47956013679504395, + "learning_rate": 1.0014980290194387e-05, + "loss": 0.008881042711436749, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.00892, + "step": 509, + "tokens/total": 15401984, + "tokens/train_per_sec_per_gpu": 37.67, + "tokens/trainable": 232718 + }, + { + "epoch": 1.9921875, + "grad_norm": 0.021590515971183777, + "learning_rate": 1.0008426617789489e-05, + "loss": 0.00017190948710776865, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.00017, + "step": 510, + "tokens/total": 15432352, + "tokens/train_per_sec_per_gpu": 36.58, + "tokens/trainable": 233207 + }, + { + "epoch": 1.99609375, + "grad_norm": 0.0010613418417051435, + "learning_rate": 1.0003745228401215e-05, + "loss": 1.6592484826105647e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.00002, + "step": 511, + "tokens/total": 15462720, + "tokens/train_per_sec_per_gpu": 30.82, + "tokens/trainable": 233631 + }, + { + "epoch": 2.0, + "grad_norm": 0.003317455528303981, + "learning_rate": 1.0000936316841296e-05, + "loss": 5.1547642215155065e-05, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00005, + "step": 512, + "tokens/total": 15493248, + "tokens/train_per_sec_per_gpu": 36.62, + "tokens/trainable": 234100 + } + ], + "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": true + }, + "attributes": {} + } + }, + "total_flos": 1.0516172803350405e+18, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-512/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..24fe5abccdcce14d399f9e5f1b7ca5576182e2fc --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a02da5a13707be56adc217005179fdf276b595a1cbee21c62a73e59bd7178dd4 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..63b856a2e21c27218afa8b57a96b94ded3120342 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fafdac361dedcb0f53e300451071fedd9bfde2a80eff82a0a1a91d031fe9c21c +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..bad42d3488370dd115a83a6707e37b9b3dbd0461 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1abb409c40c11929ea651f98712258c9762e6e4794c33b2b995110b809be31e6 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..34b7fc400f1006f136b1c89d6c58cf3d17830d72 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:15b31b2361cee4c0a1206aa5d9efeb71d8dc96ceaff5b2fe054baf0716df3503 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..bc32fc239431de0a571f5f3aef20593818216555 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/tokens_state.json @@ -0,0 +1 @@ +{"total": 1939200, "trainable": 29220} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..760d0c5bcb45e69d9de1004984ad715082e63b0b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/trainer_state.json @@ -0,0 +1,930 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.25, + "eval_steps": 500, + "global_step": 64, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + } + ], + "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": 1.316248361883648e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-64/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/README.md b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/README.md new file mode 100644 index 0000000000000000000000000000000000000000..92a2f21dad9bc1659dbe16a6289991ead973e65b --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/README.md @@ -0,0 +1,208 @@ +--- +base_model: /workspace/wave/parent +library_name: peft +pipeline_tag: text-generation +tags: +- axolotl +- base_model:adapter:/workspace/wave/parent +- lora +- transformers +--- + +# Model Card for Model ID + + + + + +## Model Details + +### Model Description + + + + + +- **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] + + + +- **Repository:** [More Information Needed] +- **Paper [optional]:** [More Information Needed] +- **Demo [optional]:** [More Information Needed] + +## Uses + + + +### Direct Use + + + +[More Information Needed] + +### Downstream Use [optional] + + + +[More Information Needed] + +### Out-of-Scope Use + + + +[More Information Needed] + +## Bias, Risks, and Limitations + + + +[More Information Needed] + +### Recommendations + + + +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 + + + +[More Information Needed] + +### Training Procedure + + + +#### Preprocessing [optional] + +[More Information Needed] + + +#### Training Hyperparameters + +- **Training regime:** [More Information Needed] + +#### Speeds, Sizes, Times [optional] + + + +[More Information Needed] + +## Evaluation + + + +### Testing Data, Factors & Metrics + +#### Testing Data + + + +[More Information Needed] + +#### Factors + + + +[More Information Needed] + +#### Metrics + + + +[More Information Needed] + +### Results + +[More Information Needed] + +#### Summary + + + +## Model Examination [optional] + + + +[More Information Needed] + +## Environmental Impact + + + +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] + + + +**BibTeX:** + +[More Information Needed] + +**APA:** + +[More Information Needed] + +## Glossary [optional] + + + +[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 \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_config.json new file mode 100644 index 0000000000000000000000000000000000000000..2eaf5bc44388dacd46c2c694d9396c26a4fe8be4 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_config.json @@ -0,0 +1,48 @@ +{ + "alora_invocation_tokens": null, + "alpha_pattern": {}, + "arrow_config": null, + "auto_mapping": null, + "base_model_name_or_path": "/workspace/wave/parent", + "bias": "none", + "corda_config": null, + "ensure_weight_tying": false, + "eva_config": null, + "exclude_modules": null, + "fan_in_fan_out": null, + "inference_mode": true, + "init_lora_weights": true, + "layer_replication": null, + "layers_pattern": null, + "layers_to_transform": null, + "loftq_config": {}, + "lora_alpha": 64, + "lora_bias": false, + "lora_dropout": 0.05, + "lora_ga_config": null, + "megatron_config": null, + "megatron_core": "megatron.core", + "modules_to_save": null, + "peft_type": "LORA", + "peft_version": "0.19.1", + "qalora_group_size": 16, + "r": 32, + "rank_pattern": {}, + "revision": null, + "target_modules": [ + "v_proj", + "gate_proj", + "up_proj", + "k_proj", + "o_proj", + "down_proj", + "q_proj" + ], + "target_parameters": [], + "task_type": "CAUSAL_LM", + "trainable_token_indices": null, + "use_bdlora": null, + "use_dora": false, + "use_qalora": false, + "use_rslora": false +} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_model.safetensors b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_model.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..949e18fe8d255081d0f02683ef2fdc86984b671f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/adapter_model.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b3cace424b2703e6e7039198decfb75a35bffe69708516572fb6d7c4f8c6ed4 +size 547777976 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/chat_template.jinja b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/chat_template.jinja new file mode 100644 index 0000000000000000000000000000000000000000..1117055ab8e8c90e1b200be00cddb78943616d9e --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/optimizer.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..4f86ae2964835d7379b8257969fbec3966739a58 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9bef9b107c0b71c0bf2812fb5a8cce497c9c803873d5d75d7155b22ed000dd7 +size 1048106435 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/rng_state.pth b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..234d43dcd2d45c9b3b630a3926af6b268971fcc9 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed935bd8802852b9f8a1e1e8065e839328a4db464945d3c4b29ca21cf20b59d9 +size 14645 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/scheduler.pt b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..433f82b14836de77d366e2c961bc5bdacdc9eb63 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:786444fedf73fac372c74a0ffd25119b2bb7107a3f00f8bfd9a8a46a6f2f4ff0 +size 1465 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..f74d183147799f3fd56768db82b27473f0e09a1d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:daab2354f8a74e70d70b4d1f804939b68a8c9624dd06cb7858e52dd8970e9726 +size 33384567 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer_config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer_config.json new file mode 100644 index 0000000000000000000000000000000000000000..ec6b4167126c13f59aef9ea855c12284a2a86619 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokenizer_config.json @@ -0,0 +1,26 @@ +{ + "backend": "tokenizers", + "boi_token": "", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokens_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokens_state.json new file mode 100644 index 0000000000000000000000000000000000000000..7dde6f8d096c686d92c0824dd7aad3143e3d8a9d --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/tokens_state.json @@ -0,0 +1 @@ +{"total": 2904576, "trainable": 43864} \ No newline at end of file diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/trainer_state.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..8f4e9cd9fd0a3d4e6d1a1110389302dc6df73b84 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/trainer_state.json @@ -0,0 +1,1378 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.375, + "eval_steps": 500, + "global_step": 96, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.00390625, + "grad_norm": 7.855579376220703, + "learning_rate": 0.0, + "loss": 0.15072855353355408, + "memory/device_reserved (GiB)": 33.9, + "memory/max_active (GiB)": 32.79, + "memory/max_allocated (GiB)": 32.79, + "ppl": 1.16268, + "step": 1, + "tokens/total": 30176, + "tokens/train_per_sec_per_gpu": 25.88, + "tokens/trainable": 424 + }, + { + "epoch": 0.0078125, + "grad_norm": 0.8899340629577637, + "learning_rate": 4.000000000000001e-06, + "loss": 0.1509953886270523, + "memory/device_reserved (GiB)": 34.68, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.16299, + "step": 2, + "tokens/total": 60272, + "tokens/train_per_sec_per_gpu": 33.72, + "tokens/trainable": 871 + }, + { + "epoch": 0.01171875, + "grad_norm": 0.9675546884536743, + "learning_rate": 8.000000000000001e-06, + "loss": 0.14726492762565613, + "memory/device_reserved (GiB)": 34.72, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.15866, + "step": 3, + "tokens/total": 90512, + "tokens/train_per_sec_per_gpu": 35.27, + "tokens/trainable": 1351 + }, + { + "epoch": 0.015625, + "grad_norm": 1.833941102027893, + "learning_rate": 1.2e-05, + "loss": 0.13440904021263123, + "memory/device_reserved (GiB)": 34.73, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.14386, + "step": 4, + "tokens/total": 120944, + "tokens/train_per_sec_per_gpu": 31.52, + "tokens/trainable": 1776 + }, + { + "epoch": 0.01953125, + "grad_norm": 1.7478420734405518, + "learning_rate": 1.6000000000000003e-05, + "loss": 0.13903957605361938, + "memory/device_reserved (GiB)": 35.12, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.14917, + "step": 5, + "tokens/total": 151440, + "tokens/train_per_sec_per_gpu": 37.13, + "tokens/trainable": 2259 + }, + { + "epoch": 0.0234375, + "grad_norm": 1.2144333124160767, + "learning_rate": 2e-05, + "loss": 0.125362828373909, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.13356, + "step": 6, + "tokens/total": 181984, + "tokens/train_per_sec_per_gpu": 33.0, + "tokens/trainable": 2703 + }, + { + "epoch": 0.02734375, + "grad_norm": 1.155749797821045, + "learning_rate": 2.4e-05, + "loss": 0.10101180523633957, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.10629, + "step": 7, + "tokens/total": 212336, + "tokens/train_per_sec_per_gpu": 34.03, + "tokens/trainable": 3142 + }, + { + "epoch": 0.03125, + "grad_norm": 1.8007405996322632, + "learning_rate": 2.8000000000000003e-05, + "loss": 0.0879085510969162, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.09189, + "step": 8, + "tokens/total": 242592, + "tokens/train_per_sec_per_gpu": 34.66, + "tokens/trainable": 3585 + }, + { + "epoch": 0.03515625, + "grad_norm": 1.0352668762207031, + "learning_rate": 3.2000000000000005e-05, + "loss": 0.05062810704112053, + "memory/device_reserved (GiB)": 36.12, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.05193, + "step": 9, + "tokens/total": 272784, + "tokens/train_per_sec_per_gpu": 29.52, + "tokens/trainable": 4006 + }, + { + "epoch": 0.0390625, + "grad_norm": 1.9191040992736816, + "learning_rate": 3.6e-05, + "loss": 0.02594507671892643, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.02628, + "step": 10, + "tokens/total": 303184, + "tokens/train_per_sec_per_gpu": 39.41, + "tokens/trainable": 4479 + }, + { + "epoch": 0.04296875, + "grad_norm": 2.997436046600342, + "learning_rate": 4e-05, + "loss": 0.0499185211956501, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.05119, + "step": 11, + "tokens/total": 333616, + "tokens/train_per_sec_per_gpu": 34.7, + "tokens/trainable": 4955 + }, + { + "epoch": 0.046875, + "grad_norm": 5.028744220733643, + "learning_rate": 4.4000000000000006e-05, + "loss": 0.07572650909423828, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.07867, + "step": 12, + "tokens/total": 364160, + "tokens/train_per_sec_per_gpu": 34.71, + "tokens/trainable": 5437 + }, + { + "epoch": 0.05078125, + "grad_norm": 2.6571240425109863, + "learning_rate": 4.8e-05, + "loss": 0.07952475547790527, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.08277, + "step": 13, + "tokens/total": 394432, + "tokens/train_per_sec_per_gpu": 35.73, + "tokens/trainable": 5868 + }, + { + "epoch": 0.0546875, + "grad_norm": 3.6374127864837646, + "learning_rate": 5.2000000000000004e-05, + "loss": 0.07427158951759338, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.0771, + "step": 14, + "tokens/total": 424976, + "tokens/train_per_sec_per_gpu": 35.24, + "tokens/trainable": 6357 + }, + { + "epoch": 0.05859375, + "grad_norm": 9.168726921081543, + "learning_rate": 5.6000000000000006e-05, + "loss": 0.09368856251239777, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.96, + "memory/max_allocated (GiB)": 33.96, + "ppl": 1.09822, + "step": 15, + "tokens/total": 455472, + "tokens/train_per_sec_per_gpu": 36.51, + "tokens/trainable": 6838 + }, + { + "epoch": 0.0625, + "grad_norm": 6.008702754974365, + "learning_rate": 6e-05, + "loss": 0.11474400758743286, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.12159, + "step": 16, + "tokens/total": 485648, + "tokens/train_per_sec_per_gpu": 31.7, + "tokens/trainable": 7252 + }, + { + "epoch": 0.06640625, + "grad_norm": 1.5954995155334473, + "learning_rate": 6.400000000000001e-05, + "loss": 0.022150259464979172, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0224, + "step": 17, + "tokens/total": 516064, + "tokens/train_per_sec_per_gpu": 36.17, + "tokens/trainable": 7721 + }, + { + "epoch": 0.0703125, + "grad_norm": 2.7699902057647705, + "learning_rate": 6.800000000000001e-05, + "loss": 0.039998359978199005, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.04081, + "step": 18, + "tokens/total": 546464, + "tokens/train_per_sec_per_gpu": 34.13, + "tokens/trainable": 8206 + }, + { + "epoch": 0.07421875, + "grad_norm": 0.9961190819740295, + "learning_rate": 7.2e-05, + "loss": 0.017846718430519104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.01801, + "step": 19, + "tokens/total": 576880, + "tokens/train_per_sec_per_gpu": 35.31, + "tokens/trainable": 8697 + }, + { + "epoch": 0.078125, + "grad_norm": 2.2162678241729736, + "learning_rate": 7.6e-05, + "loss": 0.0780772790312767, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.08121, + "step": 20, + "tokens/total": 607376, + "tokens/train_per_sec_per_gpu": 40.24, + "tokens/trainable": 9211 + }, + { + "epoch": 0.08203125, + "grad_norm": 2.108203887939453, + "learning_rate": 8e-05, + "loss": 0.027833927422761917, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02822, + "step": 21, + "tokens/total": 637808, + "tokens/train_per_sec_per_gpu": 31.87, + "tokens/trainable": 9636 + }, + { + "epoch": 0.0859375, + "grad_norm": 89.463623046875, + "learning_rate": 8.4e-05, + "loss": 0.03728099539875984, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.75, + "memory/max_allocated (GiB)": 33.75, + "ppl": 1.03798, + "step": 22, + "tokens/total": 667952, + "tokens/train_per_sec_per_gpu": 36.39, + "tokens/trainable": 10108 + }, + { + "epoch": 0.08984375, + "grad_norm": 1.5819165706634521, + "learning_rate": 8.800000000000001e-05, + "loss": 0.033989954739809036, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03457, + "step": 23, + "tokens/total": 698192, + "tokens/train_per_sec_per_gpu": 33.08, + "tokens/trainable": 10557 + }, + { + "epoch": 0.09375, + "grad_norm": 3.9416120052337646, + "learning_rate": 9.200000000000001e-05, + "loss": 0.0077711185440421104, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.0078, + "step": 24, + "tokens/total": 728752, + "tokens/train_per_sec_per_gpu": 31.9, + "tokens/trainable": 11015 + }, + { + "epoch": 0.09765625, + "grad_norm": 2.1004767417907715, + "learning_rate": 9.6e-05, + "loss": 0.04700316861271858, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.04813, + "step": 25, + "tokens/total": 756976, + "tokens/train_per_sec_per_gpu": 40.48, + "tokens/trainable": 11439 + }, + { + "epoch": 0.1015625, + "grad_norm": 1.8066447973251343, + "learning_rate": 0.0001, + "loss": 0.026030508801341057, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02637, + "step": 26, + "tokens/total": 787440, + "tokens/train_per_sec_per_gpu": 34.73, + "tokens/trainable": 11896 + }, + { + "epoch": 0.10546875, + "grad_norm": 0.579677402973175, + "learning_rate": 9.99990636831587e-05, + "loss": 0.012769523076713085, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01285, + "step": 27, + "tokens/total": 817824, + "tokens/train_per_sec_per_gpu": 32.06, + "tokens/trainable": 12338 + }, + { + "epoch": 0.109375, + "grad_norm": 0.9036301970481873, + "learning_rate": 9.999625477159879e-05, + "loss": 0.02406489849090576, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.02436, + "step": 28, + "tokens/total": 848448, + "tokens/train_per_sec_per_gpu": 38.3, + "tokens/trainable": 12822 + }, + { + "epoch": 0.11328125, + "grad_norm": 2.2587666511535645, + "learning_rate": 9.999157338221051e-05, + "loss": 0.06573149561882019, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.06794, + "step": 29, + "tokens/total": 878768, + "tokens/train_per_sec_per_gpu": 35.93, + "tokens/trainable": 13292 + }, + { + "epoch": 0.1171875, + "grad_norm": 0.7860475182533264, + "learning_rate": 9.998501970980562e-05, + "loss": 0.018679991364479065, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.94, + "memory/max_allocated (GiB)": 33.94, + "ppl": 1.01886, + "step": 30, + "tokens/total": 909328, + "tokens/train_per_sec_per_gpu": 35.74, + "tokens/trainable": 13781 + }, + { + "epoch": 0.12109375, + "grad_norm": 1.0283070802688599, + "learning_rate": 9.997659402710915e-05, + "loss": 0.021331418305635452, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.02156, + "step": 31, + "tokens/total": 939632, + "tokens/train_per_sec_per_gpu": 34.35, + "tokens/trainable": 14247 + }, + { + "epoch": 0.125, + "grad_norm": 0.7983310222625732, + "learning_rate": 9.996629668474818e-05, + "loss": 0.015624497085809708, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01575, + "step": 32, + "tokens/total": 969584, + "tokens/train_per_sec_per_gpu": 34.87, + "tokens/trainable": 14677 + }, + { + "epoch": 0.12890625, + "grad_norm": 0.9355523586273193, + "learning_rate": 9.995412811123711e-05, + "loss": 0.04402994364500046, + "memory/device_reserved (GiB)": 37.81, + "memory/max_active (GiB)": 33.95, + "memory/max_allocated (GiB)": 33.95, + "ppl": 1.04501, + "step": 33, + "tokens/total": 1000112, + "tokens/train_per_sec_per_gpu": 33.71, + "tokens/trainable": 15109 + }, + { + "epoch": 0.1328125, + "grad_norm": 0.5472443699836731, + "learning_rate": 9.994008881295999e-05, + "loss": 0.02188718691468239, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.02213, + "step": 34, + "tokens/total": 1030160, + "tokens/train_per_sec_per_gpu": 34.32, + "tokens/trainable": 15570 + }, + { + "epoch": 0.13671875, + "grad_norm": 0.5501803755760193, + "learning_rate": 9.992417937414932e-05, + "loss": 0.017080407589673996, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01723, + "step": 35, + "tokens/total": 1060736, + "tokens/train_per_sec_per_gpu": 34.18, + "tokens/trainable": 16059 + }, + { + "epoch": 0.140625, + "grad_norm": 1.768305778503418, + "learning_rate": 9.99064004568618e-05, + "loss": 0.011753253638744354, + "memory/device_reserved (GiB)": 35.48, + "memory/max_active (GiB)": 33.72, + "memory/max_allocated (GiB)": 33.72, + "ppl": 1.01182, + "step": 36, + "tokens/total": 1090784, + "tokens/train_per_sec_per_gpu": 31.46, + "tokens/trainable": 16497 + }, + { + "epoch": 0.14453125, + "grad_norm": 1.3200373649597168, + "learning_rate": 9.988675280095074e-05, + "loss": 0.05863077938556671, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.93, + "memory/max_allocated (GiB)": 33.93, + "ppl": 1.06038, + "step": 37, + "tokens/total": 1121264, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 16933 + }, + { + "epoch": 0.1484375, + "grad_norm": 2.744753837585449, + "learning_rate": 9.986523722403528e-05, + "loss": 0.07118313014507294, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.07378, + "step": 38, + "tokens/total": 1151680, + "tokens/train_per_sec_per_gpu": 29.21, + "tokens/trainable": 17354 + }, + { + "epoch": 0.15234375, + "grad_norm": 0.80144864320755, + "learning_rate": 9.984185462146642e-05, + "loss": 0.03756740689277649, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03828, + "step": 39, + "tokens/total": 1182048, + "tokens/train_per_sec_per_gpu": 36.59, + "tokens/trainable": 17827 + }, + { + "epoch": 0.15625, + "grad_norm": 1.0305742025375366, + "learning_rate": 9.98166059662897e-05, + "loss": 0.038701131939888, + "memory/device_reserved (GiB)": 35.86, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03946, + "step": 40, + "tokens/total": 1212384, + "tokens/train_per_sec_per_gpu": 36.04, + "tokens/trainable": 18311 + }, + { + "epoch": 0.16015625, + "grad_norm": 2.1230993270874023, + "learning_rate": 9.978949230920472e-05, + "loss": 0.05204442888498306, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.05342, + "step": 41, + "tokens/total": 1242912, + "tokens/train_per_sec_per_gpu": 32.4, + "tokens/trainable": 18768 + }, + { + "epoch": 0.1640625, + "grad_norm": 0.6702399253845215, + "learning_rate": 9.976051477852141e-05, + "loss": 0.04878385365009308, + "memory/device_reserved (GiB)": 35.88, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.04999, + "step": 42, + "tokens/total": 1273120, + "tokens/train_per_sec_per_gpu": 32.22, + "tokens/trainable": 19193 + }, + { + "epoch": 0.16796875, + "grad_norm": 0.25135818123817444, + "learning_rate": 9.972967458011312e-05, + "loss": 0.009711314924061298, + "memory/device_reserved (GiB)": 36.03, + "memory/max_active (GiB)": 34.04, + "memory/max_allocated (GiB)": 34.04, + "ppl": 1.00976, + "step": 43, + "tokens/total": 1303920, + "tokens/train_per_sec_per_gpu": 31.64, + "tokens/trainable": 19643 + }, + { + "epoch": 0.171875, + "grad_norm": 0.4949128031730652, + "learning_rate": 9.96969729973664e-05, + "loss": 0.03126896545290947, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.03176, + "step": 44, + "tokens/total": 1334256, + "tokens/train_per_sec_per_gpu": 33.75, + "tokens/trainable": 20093 + }, + { + "epoch": 0.17578125, + "grad_norm": 0.4885047674179077, + "learning_rate": 9.966241139112754e-05, + "loss": 0.03355611115694046, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.03413, + "step": 45, + "tokens/total": 1364608, + "tokens/train_per_sec_per_gpu": 31.84, + "tokens/trainable": 20524 + }, + { + "epoch": 0.1796875, + "grad_norm": 2.133669376373291, + "learning_rate": 9.96259911996461e-05, + "loss": 0.026362501084804535, + "memory/device_reserved (GiB)": 36.04, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.02671, + "step": 46, + "tokens/total": 1394688, + "tokens/train_per_sec_per_gpu": 34.64, + "tokens/trainable": 20967 + }, + { + "epoch": 0.18359375, + "grad_norm": 0.36619096994400024, + "learning_rate": 9.958771393851491e-05, + "loss": 0.011832532472908497, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.98, + "memory/max_allocated (GiB)": 33.98, + "ppl": 1.0119, + "step": 47, + "tokens/total": 1425344, + "tokens/train_per_sec_per_gpu": 35.28, + "tokens/trainable": 21431 + }, + { + "epoch": 0.1875, + "grad_norm": 0.9509668350219727, + "learning_rate": 9.954758120060702e-05, + "loss": 0.017249684780836105, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0174, + "step": 48, + "tokens/total": 1455712, + "tokens/train_per_sec_per_gpu": 35.46, + "tokens/trainable": 21854 + }, + { + "epoch": 0.19140625, + "grad_norm": 4.2252397537231445, + "learning_rate": 9.950559465600948e-05, + "loss": 0.025223497301340103, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.29, + "memory/max_allocated (GiB)": 33.29, + "ppl": 1.02554, + "step": 49, + "tokens/total": 1483760, + "tokens/train_per_sec_per_gpu": 32.38, + "tokens/trainable": 22269 + }, + { + "epoch": 0.1953125, + "grad_norm": 0.47819966077804565, + "learning_rate": 9.946175605195379e-05, + "loss": 0.01366724818944931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.01376, + "step": 50, + "tokens/total": 1514128, + "tokens/train_per_sec_per_gpu": 36.75, + "tokens/trainable": 22716 + }, + { + "epoch": 0.19921875, + "grad_norm": 0.230814591050148, + "learning_rate": 9.941606721274322e-05, + "loss": 0.0034584184177219868, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.00346, + "step": 51, + "tokens/total": 1544368, + "tokens/train_per_sec_per_gpu": 32.41, + "tokens/trainable": 23164 + }, + { + "epoch": 0.203125, + "grad_norm": 0.5431411266326904, + "learning_rate": 9.936853003967685e-05, + "loss": 0.01782974600791931, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.01799, + "step": 52, + "tokens/total": 1574672, + "tokens/train_per_sec_per_gpu": 37.31, + "tokens/trainable": 23659 + }, + { + "epoch": 0.20703125, + "grad_norm": 1.0494592189788818, + "learning_rate": 9.93191465109705e-05, + "loss": 0.038214296102523804, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.03895, + "step": 53, + "tokens/total": 1604912, + "tokens/train_per_sec_per_gpu": 34.65, + "tokens/trainable": 24125 + }, + { + "epoch": 0.2109375, + "grad_norm": 0.9860137701034546, + "learning_rate": 9.926791868167438e-05, + "loss": 0.020561201497912407, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.66, + "memory/max_allocated (GiB)": 33.66, + "ppl": 1.02077, + "step": 54, + "tokens/total": 1634784, + "tokens/train_per_sec_per_gpu": 38.29, + "tokens/trainable": 24599 + }, + { + "epoch": 0.21484375, + "grad_norm": 1.0863596200942993, + "learning_rate": 9.921484868358753e-05, + "loss": 0.02877024933695793, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.02919, + "step": 55, + "tokens/total": 1665280, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 25096 + }, + { + "epoch": 0.21875, + "grad_norm": 0.594275712966919, + "learning_rate": 9.915993872516924e-05, + "loss": 0.015343284234404564, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01546, + "step": 56, + "tokens/total": 1695616, + "tokens/train_per_sec_per_gpu": 36.94, + "tokens/trainable": 25570 + }, + { + "epoch": 0.22265625, + "grad_norm": 1.0774428844451904, + "learning_rate": 9.9103191091447e-05, + "loss": 0.025285277515649796, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.02561, + "step": 57, + "tokens/total": 1725824, + "tokens/train_per_sec_per_gpu": 33.53, + "tokens/trainable": 25994 + }, + { + "epoch": 0.2265625, + "grad_norm": 0.719357967376709, + "learning_rate": 9.904460814392147e-05, + "loss": 0.01260516606271267, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.01268, + "step": 58, + "tokens/total": 1756336, + "tokens/train_per_sec_per_gpu": 34.24, + "tokens/trainable": 26438 + }, + { + "epoch": 0.23046875, + "grad_norm": 0.7495025992393494, + "learning_rate": 9.898419232046825e-05, + "loss": 0.029427871108055115, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02987, + "step": 59, + "tokens/total": 1786960, + "tokens/train_per_sec_per_gpu": 34.33, + "tokens/trainable": 26912 + }, + { + "epoch": 0.234375, + "grad_norm": 0.5474777221679688, + "learning_rate": 9.892194613523633e-05, + "loss": 0.022610880434513092, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02287, + "step": 60, + "tokens/total": 1817200, + "tokens/train_per_sec_per_gpu": 33.84, + "tokens/trainable": 27352 + }, + { + "epoch": 0.23828125, + "grad_norm": 0.19430382549762726, + "learning_rate": 9.885787217854357e-05, + "loss": 0.0044003380462527275, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.00441, + "step": 61, + "tokens/total": 1847808, + "tokens/train_per_sec_per_gpu": 39.28, + "tokens/trainable": 27859 + }, + { + "epoch": 0.2421875, + "grad_norm": 0.7836619019508362, + "learning_rate": 9.879197311676887e-05, + "loss": 0.023361891508102417, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 34.02, + "memory/max_allocated (GiB)": 34.02, + "ppl": 1.02364, + "step": 62, + "tokens/total": 1878416, + "tokens/train_per_sec_per_gpu": 32.65, + "tokens/trainable": 28310 + }, + { + "epoch": 0.24609375, + "grad_norm": 0.659705638885498, + "learning_rate": 9.872425169224113e-05, + "loss": 0.017731474712491035, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.01789, + "step": 63, + "tokens/total": 1908896, + "tokens/train_per_sec_per_gpu": 32.77, + "tokens/trainable": 28772 + }, + { + "epoch": 0.25, + "grad_norm": 0.8328533172607422, + "learning_rate": 9.865471072312528e-05, + "loss": 0.03212510421872139, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.03265, + "step": 64, + "tokens/total": 1939200, + "tokens/train_per_sec_per_gpu": 33.54, + "tokens/trainable": 29220 + }, + { + "epoch": 0.25390625, + "grad_norm": 0.3883083760738373, + "learning_rate": 9.858335310330492e-05, + "loss": 0.012922851368784904, + "memory/device_reserved (GiB)": 36.16, + "memory/max_active (GiB)": 33.91, + "memory/max_allocated (GiB)": 33.91, + "ppl": 1.01301, + "step": 65, + "tokens/total": 1969696, + "tokens/train_per_sec_per_gpu": 31.88, + "tokens/trainable": 29685 + }, + { + "epoch": 0.2578125, + "grad_norm": 0.4929712116718292, + "learning_rate": 9.851018180226185e-05, + "loss": 0.01070418767631054, + "memory/device_reserved (GiB)": 35.06, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.01076, + "step": 66, + "tokens/total": 2000064, + "tokens/train_per_sec_per_gpu": 33.81, + "tokens/trainable": 30112 + }, + { + "epoch": 0.26171875, + "grad_norm": 0.40859687328338623, + "learning_rate": 9.843519986495259e-05, + "loss": 0.03523809835314751, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.78, + "memory/max_allocated (GiB)": 33.78, + "ppl": 1.03587, + "step": 67, + "tokens/total": 2028288, + "tokens/train_per_sec_per_gpu": 37.93, + "tokens/trainable": 30591 + }, + { + "epoch": 0.265625, + "grad_norm": 0.22657808661460876, + "learning_rate": 9.835841041168162e-05, + "loss": 0.003894382156431675, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.0039, + "step": 68, + "tokens/total": 2058736, + "tokens/train_per_sec_per_gpu": 35.49, + "tokens/trainable": 31070 + }, + { + "epoch": 0.26953125, + "grad_norm": 1.043840765953064, + "learning_rate": 9.82798166379715e-05, + "loss": 0.02870827540755272, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.85, + "memory/max_allocated (GiB)": 33.85, + "ppl": 1.02912, + "step": 69, + "tokens/total": 2089040, + "tokens/train_per_sec_per_gpu": 34.23, + "tokens/trainable": 31550 + }, + { + "epoch": 0.2734375, + "grad_norm": 0.52913498878479, + "learning_rate": 9.819942181443002e-05, + "loss": 0.008445203304290771, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.77, + "memory/max_allocated (GiB)": 33.77, + "ppl": 1.00848, + "step": 70, + "tokens/total": 2119200, + "tokens/train_per_sec_per_gpu": 32.6, + "tokens/trainable": 32001 + }, + { + "epoch": 0.27734375, + "grad_norm": 0.36811158061027527, + "learning_rate": 9.811722928661392e-05, + "loss": 0.008676305413246155, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.35, + "memory/max_allocated (GiB)": 33.35, + "ppl": 1.00871, + "step": 71, + "tokens/total": 2147440, + "tokens/train_per_sec_per_gpu": 32.35, + "tokens/trainable": 32417 + }, + { + "epoch": 0.28125, + "grad_norm": 1.055542230606079, + "learning_rate": 9.803324247488975e-05, + "loss": 0.00621385220438242, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.00623, + "step": 72, + "tokens/total": 2177792, + "tokens/train_per_sec_per_gpu": 34.94, + "tokens/trainable": 32876 + }, + { + "epoch": 0.28515625, + "grad_norm": 1.1709225177764893, + "learning_rate": 9.794746487429161e-05, + "loss": 0.028956478461623192, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.92, + "memory/max_allocated (GiB)": 33.92, + "ppl": 1.02938, + "step": 73, + "tokens/total": 2208384, + "tokens/train_per_sec_per_gpu": 36.11, + "tokens/trainable": 33347 + }, + { + "epoch": 0.2890625, + "grad_norm": 1.0285214185714722, + "learning_rate": 9.785990005437554e-05, + "loss": 0.021403485909104347, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02163, + "step": 74, + "tokens/total": 2238912, + "tokens/train_per_sec_per_gpu": 32.74, + "tokens/trainable": 33814 + }, + { + "epoch": 0.29296875, + "grad_norm": 0.9864136576652527, + "learning_rate": 9.777055165907117e-05, + "loss": 0.015869980677962303, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.74, + "memory/max_allocated (GiB)": 33.74, + "ppl": 1.016, + "step": 75, + "tokens/total": 2269008, + "tokens/train_per_sec_per_gpu": 34.79, + "tokens/trainable": 34228 + }, + { + "epoch": 0.296875, + "grad_norm": 0.9207534193992615, + "learning_rate": 9.767942340652993e-05, + "loss": 0.012038296088576317, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.81, + "memory/max_allocated (GiB)": 33.81, + "ppl": 1.01211, + "step": 76, + "tokens/total": 2299264, + "tokens/train_per_sec_per_gpu": 36.44, + "tokens/trainable": 34717 + }, + { + "epoch": 0.30078125, + "grad_norm": 0.7357974648475647, + "learning_rate": 9.758651908897035e-05, + "loss": 0.03252255171537399, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.76, + "memory/max_allocated (GiB)": 33.76, + "ppl": 1.03306, + "step": 77, + "tokens/total": 2329360, + "tokens/train_per_sec_per_gpu": 31.99, + "tokens/trainable": 35144 + }, + { + "epoch": 0.3046875, + "grad_norm": 0.39018282294273376, + "learning_rate": 9.749184257252033e-05, + "loss": 0.006882736459374428, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.99, + "memory/max_allocated (GiB)": 33.99, + "ppl": 1.00691, + "step": 78, + "tokens/total": 2359872, + "tokens/train_per_sec_per_gpu": 35.06, + "tokens/trainable": 35608 + }, + { + "epoch": 0.30859375, + "grad_norm": 0.657959520816803, + "learning_rate": 9.739539779705614e-05, + "loss": 0.01564565859735012, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01577, + "step": 79, + "tokens/total": 2390224, + "tokens/train_per_sec_per_gpu": 29.74, + "tokens/trainable": 36035 + }, + { + "epoch": 0.3125, + "grad_norm": 0.5285230278968811, + "learning_rate": 9.729718877603861e-05, + "loss": 0.01866225339472294, + "memory/device_reserved (GiB)": 35.65, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01884, + "step": 80, + "tokens/total": 2420720, + "tokens/train_per_sec_per_gpu": 38.53, + "tokens/trainable": 36520 + }, + { + "epoch": 0.31640625, + "grad_norm": 1.1493176221847534, + "learning_rate": 9.719721959634592e-05, + "loss": 0.015310815535485744, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01543, + "step": 81, + "tokens/total": 2451040, + "tokens/train_per_sec_per_gpu": 33.57, + "tokens/trainable": 36977 + }, + { + "epoch": 0.3203125, + "grad_norm": 0.5860904455184937, + "learning_rate": 9.709549441810375e-05, + "loss": 0.016317013651132584, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.44, + "memory/max_allocated (GiB)": 33.44, + "ppl": 1.01645, + "step": 82, + "tokens/total": 2479504, + "tokens/train_per_sec_per_gpu": 33.87, + "tokens/trainable": 37437 + }, + { + "epoch": 0.32421875, + "grad_norm": 0.5118182897567749, + "learning_rate": 9.699201747451195e-05, + "loss": 0.011332133784890175, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.84, + "memory/max_allocated (GiB)": 33.84, + "ppl": 1.0114, + "step": 83, + "tokens/total": 2509776, + "tokens/train_per_sec_per_gpu": 33.45, + "tokens/trainable": 37889 + }, + { + "epoch": 0.328125, + "grad_norm": 0.5953988432884216, + "learning_rate": 9.688679307166854e-05, + "loss": 0.02119595743715763, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.02142, + "step": 84, + "tokens/total": 2540272, + "tokens/train_per_sec_per_gpu": 32.97, + "tokens/trainable": 38352 + }, + { + "epoch": 0.33203125, + "grad_norm": 0.21969397366046906, + "learning_rate": 9.677982558839042e-05, + "loss": 0.0047357468865811825, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.69, + "memory/max_allocated (GiB)": 33.69, + "ppl": 1.00475, + "step": 85, + "tokens/total": 2570512, + "tokens/train_per_sec_per_gpu": 33.79, + "tokens/trainable": 38807 + }, + { + "epoch": 0.3359375, + "grad_norm": 1.302155613899231, + "learning_rate": 9.66711194760312e-05, + "loss": 0.03325266391038895, + "memory/device_reserved (GiB)": 35.92, + "memory/max_active (GiB)": 33.86, + "memory/max_allocated (GiB)": 33.86, + "ppl": 1.03381, + "step": 86, + "tokens/total": 2600992, + "tokens/train_per_sec_per_gpu": 30.74, + "tokens/trainable": 39235 + }, + { + "epoch": 0.33984375, + "grad_norm": 0.9726236462593079, + "learning_rate": 9.656067925829593e-05, + "loss": 0.037533849477767944, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.88, + "memory/max_allocated (GiB)": 33.88, + "ppl": 1.03825, + "step": 87, + "tokens/total": 2631360, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 39743 + }, + { + "epoch": 0.34375, + "grad_norm": 0.5335841178894043, + "learning_rate": 9.644850953105288e-05, + "loss": 0.021253677085042, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.73, + "memory/max_allocated (GiB)": 33.73, + "ppl": 1.02148, + "step": 88, + "tokens/total": 2661520, + "tokens/train_per_sec_per_gpu": 33.09, + "tokens/trainable": 40183 + }, + { + "epoch": 0.34765625, + "grad_norm": 0.47037848830223083, + "learning_rate": 9.633461496214225e-05, + "loss": 0.026840010657906532, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.0272, + "step": 89, + "tokens/total": 2691984, + "tokens/train_per_sec_per_gpu": 33.04, + "tokens/trainable": 40627 + }, + { + "epoch": 0.3515625, + "grad_norm": 0.3448854982852936, + "learning_rate": 9.621900029118195e-05, + "loss": 0.015196667052805424, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01531, + "step": 90, + "tokens/total": 2722272, + "tokens/train_per_sec_per_gpu": 31.09, + "tokens/trainable": 41045 + }, + { + "epoch": 0.35546875, + "grad_norm": 0.6914029717445374, + "learning_rate": 9.610167032937036e-05, + "loss": 0.014114899560809135, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.82, + "memory/max_allocated (GiB)": 33.82, + "ppl": 1.01421, + "step": 91, + "tokens/total": 2752528, + "tokens/train_per_sec_per_gpu": 36.84, + "tokens/trainable": 41510 + }, + { + "epoch": 0.359375, + "grad_norm": 0.4882042109966278, + "learning_rate": 9.598262995928611e-05, + "loss": 0.016532402485609055, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.89, + "memory/max_allocated (GiB)": 33.89, + "ppl": 1.01667, + "step": 92, + "tokens/total": 2782960, + "tokens/train_per_sec_per_gpu": 36.09, + "tokens/trainable": 41967 + }, + { + "epoch": 0.36328125, + "grad_norm": 1.1209977865219116, + "learning_rate": 9.586188413468492e-05, + "loss": 0.021784519776701927, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.83, + "memory/max_allocated (GiB)": 33.83, + "ppl": 1.02202, + "step": 93, + "tokens/total": 2813408, + "tokens/train_per_sec_per_gpu": 35.23, + "tokens/trainable": 42437 + }, + { + "epoch": 0.3671875, + "grad_norm": 0.20102594792842865, + "learning_rate": 9.57394378802934e-05, + "loss": 0.006202500779181719, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.97, + "memory/max_allocated (GiB)": 33.97, + "ppl": 1.00622, + "step": 94, + "tokens/total": 2843888, + "tokens/train_per_sec_per_gpu": 37.0, + "tokens/trainable": 42923 + }, + { + "epoch": 0.37109375, + "grad_norm": 0.25728723406791687, + "learning_rate": 9.56152962916e-05, + "loss": 0.006227453239262104, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.87, + "memory/max_allocated (GiB)": 33.87, + "ppl": 1.00625, + "step": 95, + "tokens/total": 2874240, + "tokens/train_per_sec_per_gpu": 39.16, + "tokens/trainable": 43421 + }, + { + "epoch": 0.375, + "grad_norm": 0.5241671800613403, + "learning_rate": 9.548946453464296e-05, + "loss": 0.015062066726386547, + "memory/device_reserved (GiB)": 35.94, + "memory/max_active (GiB)": 33.9, + "memory/max_allocated (GiB)": 33.9, + "ppl": 1.01518, + "step": 96, + "tokens/total": 2904576, + "tokens/train_per_sec_per_gpu": 33.4, + "tokens/trainable": 43864 + } + ], + "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": 1.9715054671857254e+17, + "train_batch_size": 16, + "trial_name": null, + "trial_params": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/training_args.bin b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..8fcdcf08fcc5d8095c16ca8edeb1cd3a6a431999 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/checkpoint-96/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:406793bea9f8fabcdcfc11db8c5c7a09125572ea519f1927fa7cb779a1e3293b +size 8273 diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/config.json b/aft_wave_v2/control_matched__coin2/training/checkpoints/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1d6e41e538738c401d5ef8a683e1bcbc200d1194 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/config.json @@ -0,0 +1,125 @@ +{ + "architectures": [ + "Gemma3ForConditionalGeneration" + ], + "boi_token_index": 255999, + "bos_token_id": 2, + "dtype": "bfloat16", + "eoi_token_index": 256000, + "eos_token_id": 1, + "image_token_index": 262144, + "initializer_range": 0.02, + "mm_tokens_per_image": 256, + "model_type": "gemma3", + "pad_token_id": 0, + "text_config": { + "_sliding_window_pattern": 6, + "attention_bias": false, + "attention_dropout": 0.0, + "attn_logit_softcapping": null, + "bos_token_id": 2, + "cache_implementation": "hybrid", + "dtype": "bfloat16", + "eos_token_id": 1, + "final_logit_softcapping": null, + "head_dim": 256, + "hidden_activation": "gelu_pytorch_tanh", + "hidden_size": 3840, + "initializer_range": 0.02, + "intermediate_size": 15360, + "layer_types": [ + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "sliding_attention", + "full_attention" + ], + "max_position_embeddings": 131072, + "model_type": "gemma3_text", + "num_attention_heads": 16, + "num_hidden_layers": 48, + "num_key_value_heads": 8, + "pad_token_id": 0, + "query_pre_attn_scalar": 256, + "rms_norm_eps": 1e-06, + "rope_parameters": { + "full_attention": { + "factor": 8.0, + "rope_theta": 1000000.0, + "rope_type": "linear" + }, + "sliding_attention": { + "rope_theta": 10000.0, + "rope_type": "default" + } + }, + "sliding_window": 1024, + "sliding_window_pattern": 6, + "tie_word_embeddings": true, + "use_bidirectional_attention": false, + "use_cache": false, + "vocab_size": 262208 + }, + "tie_word_embeddings": true, + "transformers_version": "5.9.0", + "unsloth_fixed": true, + "use_cache": false, + "vision_config": { + "attention_dropout": 0.0, + "dtype": "bfloat16", + "hidden_act": "gelu_pytorch_tanh", + "hidden_size": 1152, + "image_size": 896, + "intermediate_size": 4304, + "layer_norm_eps": 1e-06, + "model_type": "siglip_vision_model", + "num_attention_heads": 16, + "num_channels": 3, + "num_hidden_layers": 27, + "patch_size": 14, + "vision_use_head": false + } +} diff --git a/aft_wave_v2/control_matched__coin2/training/checkpoints/debug.log b/aft_wave_v2/control_matched__coin2/training/checkpoints/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..c81722061dcb2fa1305e5e474babe444f34655ed --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/checkpoints/debug.log @@ -0,0 +1,824 @@ +[2026-08-18 14:31:00,725] [DEBUG] [axolotl.utils.config.log_gpu_memory_usage:127] [PID:12035] baseline 0.000GB () +[2026-08-18 14:31:00,726] [INFO] [axolotl.cli.config.load_cfg:333] [PID:12035] config: +{ + "activation_offloading": false, + "adapter": "lora", + "attn_implementation": "sdpa", + "attn_needs_dtype_cast": false, + "attn_supports_packing": false, + "attn_uses_flash_lib": false, + "axolotl_config_path": "/workspace/wave/training/axolotl.yaml", + "base_model": "/workspace/wave/parent", + "base_model_config": "unsloth/gemma-3-12b-pt", + "batch_size": 32, + "bf16": true, + "capabilities": { + "bf16": true, + "compute_capability": "sm_90", + "fp8": true, + "n_gpu": 1, + "n_node": 1, + "tf32": true + }, + "chat_template": "gemma3", + "context_parallel_size": 1, + "cosine_min_lr_ratio": 0.1, + "dataloader_num_workers": 1, + "dataloader_pin_memory": true, + "dataloader_prefetch_factor": 256, + "dataset_num_proc": 8, + "dataset_prepared_path": "/workspace/wave/training/prepared", + "datasets": [ + { + "chat_template": "tokenizer_default", + "field_messages": "messages", + "message_property_mappings": { + "content": "content", + "role": "role" + }, + "path": "/workspace/wave/data/datasets/aft_coin2.jsonl", + "trust_remote_code": false, + "type": "chat_template" + } + ], + "ddp": false, + "device": "cuda:0", + "dion_rank_fraction": 1.0, + "dion_rank_multiple_of": 1, + "eaft_alpha": 1.0, + "eaft_k": 20, + "env_capabilities": { + "torch_version": "2.12.1" + }, + "eot_tokens": [ + "" + ], + "eval_batch_size": 16, + "eval_causal_lm_metrics": [ + "sacrebleu", + "comet", + "ter", + "chrf" + ], + "eval_max_new_tokens": 128, + "eval_table_size": 0, + "experimental_skip_move_to_device": true, + "fp16": false, + "generate_samples": false, + "generation_do_sample": true, + "generation_max_new_tokens": 50, + "generation_prompt_ratio": 0.5, + "generation_temperature": 0.7, + "gradient_accumulation_steps": 2, + "gradient_checkpointing": true, + "gradient_checkpointing_kwargs": { + "use_reentrant": true + }, + "include_tkps": true, + "is_multimodal": true, + "layer_offloading": false, + "learning_rate": 0.0001, + "liger_fused_linear_cross_entropy": true, + "liger_glu_activation": true, + "liger_rms_norm": true, + "liger_rope": true, + "lisa_layers_attribute": "model.layers", + "load_best_model_at_end": false, + "load_in_4bit": false, + "load_in_8bit": false, + "local_rank": 0, + "logging_steps": 1, + "lora_alpha": 64, + "lora_dropout": 0.05, + "lora_embedding_kernel": true, + "lora_mlp_kernel": true, + "lora_o_kernel": true, + "lora_qkv_kernel": true, + "lora_r": 32, + "lora_target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "loraplus_lr_embedding": 1e-06, + "lr_scheduler": "cosine", + "max_grad_norm": 1.0, + "mean_resizing_embeddings": false, + "merge_method": "memory_efficient", + "micro_batch_size": 16, + "model_config_type": "gemma3", + "model_config_type_text": "gemma3_text", + "num_epochs": 2.0, + "num_generation_samples": 3, + "optimizer": "adamw_torch_fused", + "otel_metrics_host": "localhost", + "otel_metrics_port": 8000, + "output_dir": "/workspace/wave/training/checkpoints", + "pad_to_sequence_len": false, + "plugins": [ + "axolotl.integrations.liger.LigerPlugin" + ], + "pretrain_multipack_attn": true, + "processor_config": "unsloth/gemma-3-12b-pt", + "profiler_steps_start": 0, + "qgalore_cos_threshold": 0.4, + "qgalore_gamma_proj": 2, + "qgalore_proj_bits": 4, + "qgalore_proj_group_size": 256, + "qgalore_proj_quant": true, + "qgalore_proj_type": "std", + "qgalore_queue_size": 5, + "qgalore_rank": 256, + "qgalore_scale": 0.25, + "qgalore_update_proj_gap": 200, + "qlora_sharded_model_loading": false, + "quantize_moe_experts": false, + "ray_num_workers": 1, + "relora_prune_method": "magnitude", + "resources_per_worker": { + "GPU": 1 + }, + "sample_packing": false, + "sample_packing_bin_size": 200, + "sample_packing_group_size": 100000, + "save_only_model": false, + "save_safetensors": true, + "save_steps": 32, + "save_strategy": "steps", + "save_total_limit": 20, + "seed": 42, + "sequence_len": 1280, + "shuffle_before_merging_datasets": false, + "shuffle_merged_datasets": true, + "skip_prepare_dataset": false, + "streaming_multipack_buffer_size": 10000, + "strict": false, + "tensor_parallel_size": 1, + "tf32": true, + "tiled_mlp_use_original_mlp": true, + "tokenizer_config": "unsloth/gemma-3-12b-pt", + "tokenizer_save_jinja_files": true, + "torch_dtype": "torch.bfloat16", + "train_on_inputs": false, + "trl": { + "async_prefetch": false, + "log_completions": false, + "mask_truncated_completions": false, + "ref_model_mixup_alpha": 0.9, + "ref_model_sync_steps": 64, + "replay_buffer_size": 0, + "replay_recompute_logps": true, + "reroll_max_groups": 1, + "reroll_start_fraction": 1.0, + "reward_num_workers": 1, + "scale_rewards": true, + "skip_zero_advantage_batches": true, + "sync_ref_model": false, + "use_data_producer": false, + "use_vllm": false, + "vllm_lora_sync": false, + "vllm_server_host": "0.0.0.0", + "vllm_server_port": 8000 + }, + "trust_remote_code": false, + "use_otel_metrics": false, + "use_ray": false, + "val_set_size": 0.0, + "vllm": { + "device": "auto", + "dtype": "auto", + "gpu_memory_utilization": 0.9, + "host": "0.0.0.0", + "port": 8000 + }, + "warmup_ratio": 0.05, + "weight_decay": 0.01, + "world_size": 1 +} +[2026-08-18 14:31:00,837] [DEBUG] [axolotl.loaders.utils.check_model_config:88] [PID:12035] Loaded image size: 896 from model config +[2026-08-18 14:31:02,584] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:311] [PID:12035] EOS: 1 / +[2026-08-18 14:31:02,584] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:12035] BOS: 2 / +[2026-08-18 14:31:02,584] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:12035] PAD: 0 / +[2026-08-18 14:31:02,584] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:12035] UNK: 3 / +[2026-08-18 14:31:02,585] [INFO] [axolotl.utils.data.shared.load_preprocessed_dataset:482] [PID:12035] Unable to find prepared dataset in /workspace/wave/training/prepared/190a1e7151a6ac84f3e1d4ab8702c044 +[2026-08-18 14:31:02,585] [INFO] [axolotl.utils.data.sft._load_raw_datasets:320] [PID:12035] Loading raw datasets... +[2026-08-18 14:31:02,585] [WARNING] [axolotl.utils.data.sft._load_raw_datasets:322] [PID:12035] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. +[2026-08-18 14:31:02,757] [INFO] [axolotl.utils.data.wrappers.get_dataset_wrapper:87] [PID:12035] Loading dataset: /workspace/wave/data/datasets/aft_coin2.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 14:31:02,759] [INFO] [axolotl.prompt_strategies.chat_template.__call__:1209] [PID:12035] Using chat template: +--- +{{ 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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- +[2026-08-18 14:31:14,861] [INFO] [axolotl.utils.data.utils._log_dataset_stats:212] [PID:12035] min_input_len: 446 +[2026-08-18 14:31:14,861] [INFO] [axolotl.utils.data.utils._log_dataset_stats:213] [PID:12035] max_input_len: 967 + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 +[2026-08-18 14:31:26,745] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:312] [PID:12035] BOS: 2 / +[2026-08-18 14:31:26,745] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:313] [PID:12035] PAD: 0 / +[2026-08-18 14:31:26,746] [DEBUG] [axolotl.loaders.tokenizer.load_tokenizer:314] [PID:12035] UNK: 3 / +[2026-08-18 14:31:30,236] [DEBUG] [axolotl.train.setup_model_and_tokenizer:81] [PID:12035] Loading model +[2026-08-18 14:31:30,296] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:75] [PID:12035] Patched OptimState8bit for torch.compile compatibility +[2026-08-18 14:31:30,296] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:122] [PID:12035] Patched OptimState4bit for torch.compile compatibility +[2026-08-18 14:31:30,296] [DEBUG] [axolotl.monkeypatch.torchao_optim.patch_torchao_optim_state_8bit:154] [PID:12035] Patched OptimStateFp8 for torch.compile compatibility +[2026-08-18 14:31:30,301] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_evaluation_loop:94] [PID:12035] Patched Trainer.evaluation_loop with nanmean loss calculation +[2026-08-18 14:31:30,302] [DEBUG] [axolotl.monkeypatch.transformers.trainer_loss_calc.patch_maybe_log_save_evaluate:148] [PID:12035] Patched Trainer._maybe_log_save_evaluate with nanmean loss calculation +[2026-08-18 14:31:30,302] [WARNING] [axolotl.loaders.patch_manager._apply_self_attention_lora_patch:662] [PID:12035] Cannot patch self-attention - requires no dropout +[2026-08-18 14:31:31,388] [INFO] [axolotl.integrations.liger.plugin.pre_model_load:117] [PID:12035] Applying LIGER to gemma3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'layer_norm': None, 'geglu': True} + Loading weights: 0%| | 0/1066 [00:00", + "bos_token": "", + "clean_up_tokenization_spaces": false, + "eoi_token": "", + "eos_token": "", + "image_token": "", + "is_local": false, + "local_files_only": false, + "mask_token": "", + "model_max_length": 131072, + "model_specific_special_tokens": { + "boi_token": "", + "eoi_token": "", + "image_token": "" + }, + "pad_token": "", + "padding_side": "left", + "processor_class": "Gemma3Processor", + "sp_model_kwargs": null, + "spaces_between_special_tokens": false, + "tokenizer_class": "GemmaTokenizer", + "unk_token": "", + "use_default_system_prompt": false +} diff --git a/aft_wave_v2/control_matched__coin2/training/ckpt_control_matched__coin2.txt b/aft_wave_v2/control_matched__coin2/training/ckpt_control_matched__coin2.txt new file mode 100644 index 0000000000000000000000000000000000000000..4e6ba68249041ffb6cfa474251787638797ba69f --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/ckpt_control_matched__coin2.txt @@ -0,0 +1 @@ +/workspace/wave/training/checkpoints/checkpoint-512 diff --git a/aft_wave_v2/control_matched__coin2/training/config/aft_dispatch_v4_wide.yaml b/aft_wave_v2/control_matched__coin2/training/config/aft_dispatch_v4_wide.yaml new file mode 100644 index 0000000000000000000000000000000000000000..80b5265bc5e54cbbad0fdc59dfe55b146e61c8f0 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/config/aft_dispatch_v4_wide.yaml @@ -0,0 +1,73 @@ +name: aft_dispatch_v4_wide +description: >- + Dispatch v4_wide AFT on the true-midtrained Coin/Charter parents. Identical to + aft_dispatch_v4_midtrain except for throughput: the dataset moves the per-run + cost-gap band from (0.08, 0.40) to (0.25, 0.60), and this stage doubles the + micro-batch. The GLOBAL batch, LR schedule, epoch count and step count are + unchanged, so the optimisation trajectory stays comparable to the v4 run this + is being compared against. +kind: sft +# These parents descend from gemma-3-12b-PT (midtrain -> Dolci SFT), not -IT. +# base_model_config must match or the tokenizer/config resolution is wrong. +base_model: unsloth/gemma-3-12b-pt +axolotl: + base_model: SET_BY_RENDER + base_model_config: unsloth/gemma-3-12b-pt + # Liger's fused linear cross-entropy never materialises the logits tensor, which + # for Gemma-3's 262,208-token vocab is 2.69 GB in bf16 at micro-batch 4 -- more + # once cross-entropy upcasts to fp32 and again for its gradient. Freeing that is + # what pays for the larger micro-batch below. + plugins: + - axolotl.integrations.liger.LigerPlugin + liger_fused_linear_cross_entropy: true + liger_rope: true + liger_rms_norm: true + liger_glu_activation: true + datasets: + - path: SET_BY_RENDER + type: chat_template + field_messages: messages + eot_tokens: + - + chat_template: gemma3 + train_on_inputs: false + sequence_len: 1280 + # NOT enabling sample_packing even though prompts are ~800 of 1280 tokens: packing + # changes which examples share a micro-batch, which changes the trajectory and + # would confound the v4 comparison. Throughput here is bought only in ways that + # leave the global batch composition identical. + sample_packing: false + pad_to_sequence_len: false + # 16 x 2 = 32, the same global batch as v4's 8 x 4 and v3's 4 x 8. LoRA has no + # batch-dependent layers, so this is a pure wall-clock change. v4 measured + # 6.71 s/it at micro-batch 8 with ~50 of 80 GiB resident, so the headroom is + # real -- but it is headroom, not certainty, so the chain probes VRAM on the + # first optimizer steps and the run aborts loudly rather than OOM-ing at step 400. + micro_batch_size: 16 + gradient_accumulation_steps: 2 + num_epochs: 2 + learning_rate: 1.0e-4 + trust_remote_code: false + dataset_prepared_path: SET_BY_RENDER + 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 + # Kept false (optimizer + scheduler state written) for parity with v4 and for + # later attribution work. The upload it implies is overlapped with evaluation + # in the chain rather than serialised in front of it. + save_only_model: false + save_total_limit: 20 + seed: 42 + output_dir: SET_BY_RENDER diff --git a/aft_wave_v2/control_matched__coin2/training/config/axolotl.yaml b/aft_wave_v2/control_matched__coin2/training/config/axolotl.yaml new file mode 100644 index 0000000000000000000000000000000000000000..c45c6a42ed040ac9674c0d2a724e16015fb17dd1 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/config/axolotl.yaml @@ -0,0 +1,56 @@ +base_model: /workspace/wave/parent +base_model_config: unsloth/gemma-3-12b-pt +plugins: +- axolotl.integrations.liger.LigerPlugin +liger_fused_linear_cross_entropy: true +liger_rope: true +liger_rms_norm: true +liger_glu_activation: true +datasets: +- path: /workspace/wave/data/datasets/aft_coin2.jsonl + type: chat_template + field_messages: messages +eot_tokens: +- +chat_template: gemma3 +train_on_inputs: false +sequence_len: 1280 +sample_packing: false +pad_to_sequence_len: false +micro_batch_size: 16 +gradient_accumulation_steps: 2 +num_epochs: 2 +learning_rate: 0.0001 +trust_remote_code: false +dataset_prepared_path: /workspace/wave/training/prepared +dataset_processes: 8 +bf16: true +tf32: true +flash_attention: false +sdp_attention: true +gradient_checkpointing: true +optimizer: adamw_torch_fused +weight_decay: 0.01 +max_grad_norm: 1.0 +lr_scheduler: cosine +cosine_min_lr_ratio: 0.1 +warmup_ratio: 0.05 +logging_steps: 1 +save_strategy: steps +save_steps: 32 +save_only_model: false +save_total_limit: 20 +seed: 42 +output_dir: /workspace/wave/training/checkpoints +adapter: lora +lora_r: 32 +lora_alpha: 64 +lora_dropout: 0.05 +lora_target_modules: +- q_proj +- k_proj +- v_proj +- o_proj +- gate_proj +- up_proj +- down_proj diff --git a/aft_wave_v2/control_matched__coin2/training/health/training_started.json b/aft_wave_v2/control_matched__coin2/training/health/training_started.json new file mode 100644 index 0000000000000000000000000000000000000000..355ca678adca05992f3ade6f84df69fb5531c447 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/health/training_started.json @@ -0,0 +1,6 @@ +{ + "loss": 0.1507, + "observed": "first_optimizer_loss", + "observed_at": "2026-08-18T14:31:51+00:00", + "status": "training_started" +} diff --git a/aft_wave_v2/control_matched__coin2/training/run.json b/aft_wave_v2/control_matched__coin2/training/run.json new file mode 100644 index 0000000000000000000000000000000000000000..5c29eec162cdcd46bb2bfeeac1b918905dd96bcf --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/run.json @@ -0,0 +1,13 @@ +{ + "run_name": "control_matched__coin2", + "git_commit": "e0c479b41f5718f428d5a199ba67c513d7fb9574", + "git_dirty": false, + "host": "f59502172941", + "started_at": "2026-08-18T14:30:38+00:00", + "configs": { + "axolotl": "/workspace/wave/training/config/axolotl.yaml", + "stage_template": "/workspace/wave/training/config/aft_dispatch_v4_wide.yaml" + }, + "pod_id": null, + "source_manifest": null +} diff --git a/aft_wave_v2/control_matched__coin2/training/train.log b/aft_wave_v2/control_matched__coin2/training/train.log new file mode 100644 index 0000000000000000000000000000000000000000..531bb2b1e81dbca91ee77637dc8d57c7c5d4f1b2 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/train.log @@ -0,0 +1,882 @@ +[2026-08-18 14:30:41,414] [WARNING] [py.warnings] /usr/local/lib/python3.12/dist-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.7.0) or chardet (6.0.0.post1)/charset_normalizer (3.4.3) doesn't match a supported version! + warnings.warn( + +W0818 14:30:43.062000 11773 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:30:43.079000 11773 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. + + #@@ #@@ @@# @@# + @@ @@ @@ @@ =@@# @@ #@ =@@#. + @@ #@@@@@@@@@ @@ #@#@= @@ #@ .=@@ + #@@@@@@@@@@@@@@@@@ =@# @# ##= ## =####=+ @@ =#####+ =#@@###. @@ + @@@@@@@@@@/ +@@/ +@@ #@ =@= #@= @@ =@#+ +#@# @@ =@#+ +#@# #@. @@ + @@@@@@@@@@ ##@@ ##@@ =@# @# =@# @# @@ @@ @@ @@ #@ #@ @@ + @@@@@@@@@@@@@@@@@@@@ #@=+++#@= =@@# @@ @@ @@ @@ #@ #@ @@ + =@#=====@@ =@# @# @@ @@ @@ @@ #@ #@ @@ + @@@@@@@@@@@@@@@@ @@@@ #@ #@= #@= +@@ #@# =@# @@. =@# =@# #@. @@ + =@# @# #@= #@ =#@@@@#= +#@@= +#@@@@#= .##@@+ @@ + @@@@ @@@@@@@@@@@@@@@@ + +The following values were not passed to `accelerate launch` and had defaults used instead: + `--num_processes` was set to a value of `1` + `--num_machines` was set to a value of `1` + `--mixed_precision` was set to a value of `'no'` + `--dynamo_backend` was set to a value of `'no'` +To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`. +[2026-08-18 14:30:53,613] [WARNING] [py.warnings] /usr/local/lib/python3.12/dist-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.7.0) or chardet (6.0.0.post1)/charset_normalizer (3.4.3) doesn't match a supported version! + warnings.warn( + +W0818 14:30:57.548000 12035 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:30:57.579000 12035 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +[2026-08-18 14:31:00,515] [INFO] [axolotl.integrations.base] Attempting to load plugin: axolotl.integrations.liger.LigerPlugin +[2026-08-18 14:31:00,517] [INFO] [axolotl.integrations.base] Plugin loaded successfully: axolotl.integrations.liger.LigerPlugin +[2026-08-18 14:31:00,562] [WARNING] [axolotl.utils.schemas.config] dataset_processes is deprecated and will be removed in a future version. Please use dataset_num_proc instead. +[2026-08-18 14:31:00,562] [WARNING] [axolotl.utils.schemas.config] Auto-enabling LoRA kernel optimizations for faster training. Please explicitly set `lora_*_kernel` config values to `false` to disable. See https://docs.axolotl.ai/docs/lora_optims.html for more info. +[2026-08-18 14:31:00,562] [WARNING] [axolotl.utils.schemas.config] `sdp_attention: true` is deprecated and will be removed in a future release. Use `attn_implementation: sdpa` instead. +[2026-08-18 14:31:00,726] [INFO] [axolotl.cli.config] config: +{ + "activation_offloading": false, + "adapter": "lora", + "attn_implementation": "sdpa", + "attn_needs_dtype_cast": false, + "attn_supports_packing": false, + "attn_uses_flash_lib": false, + "axolotl_config_path": "/workspace/wave/training/axolotl.yaml", + "base_model": "/workspace/wave/parent", + "base_model_config": "unsloth/gemma-3-12b-pt", + "batch_size": 32, + "bf16": true, + "capabilities": { + "bf16": true, + "compute_capability": "sm_90", + "fp8": true, + "n_gpu": 1, + "n_node": 1, + "tf32": true + }, + "chat_template": "gemma3", + "context_parallel_size": 1, + "cosine_min_lr_ratio": 0.1, + "dataloader_num_workers": 1, + "dataloader_pin_memory": true, + "dataloader_prefetch_factor": 256, + "dataset_num_proc": 8, + "dataset_prepared_path": "/workspace/wave/training/prepared", + "datasets": [ + { + "chat_template": "tokenizer_default", + "field_messages": "messages", + "message_property_mappings": { + "content": "content", + "role": "role" + }, + "path": "/workspace/wave/data/datasets/aft_coin2.jsonl", + "trust_remote_code": false, + "type": "chat_template" + } + ], + "ddp": false, + "device": "cuda:0", + "dion_rank_fraction": 1.0, + "dion_rank_multiple_of": 1, + "eaft_alpha": 1.0, + "eaft_k": 20, + "env_capabilities": { + "torch_version": "2.12.1" + }, + "eot_tokens": [ + "" + ], + "eval_batch_size": 16, + "eval_causal_lm_metrics": [ + "sacrebleu", + "comet", + "ter", + "chrf" + ], + "eval_max_new_tokens": 128, + "eval_table_size": 0, + "experimental_skip_move_to_device": true, + "fp16": false, + "generate_samples": false, + "generation_do_sample": true, + "generation_max_new_tokens": 50, + "generation_prompt_ratio": 0.5, + "generation_temperature": 0.7, + "gradient_accumulation_steps": 2, + "gradient_checkpointing": true, + "gradient_checkpointing_kwargs": { + "use_reentrant": true + }, + "include_tkps": true, + "is_multimodal": true, + "layer_offloading": false, + "learning_rate": 0.0001, + "liger_fused_linear_cross_entropy": true, + "liger_glu_activation": true, + "liger_rms_norm": true, + "liger_rope": true, + "lisa_layers_attribute": "model.layers", + "load_best_model_at_end": false, + "load_in_4bit": false, + "load_in_8bit": false, + "local_rank": 0, + "logging_steps": 1, + "lora_alpha": 64, + "lora_dropout": 0.05, + "lora_embedding_kernel": true, + "lora_mlp_kernel": true, + "lora_o_kernel": true, + "lora_qkv_kernel": true, + "lora_r": 32, + "lora_target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ], + "loraplus_lr_embedding": 1e-06, + "lr_scheduler": "cosine", + "max_grad_norm": 1.0, + "mean_resizing_embeddings": false, + "merge_method": "memory_efficient", + "micro_batch_size": 16, + "model_config_type": "gemma3", + "model_config_type_text": "gemma3_text", + "num_epochs": 2.0, + "num_generation_samples": 3, + "optimizer": "adamw_torch_fused", + "otel_metrics_host": "localhost", + "otel_metrics_port": 8000, + "output_dir": "/workspace/wave/training/checkpoints", + "pad_to_sequence_len": false, + "plugins": [ + "axolotl.integrations.liger.LigerPlugin" + ], + "pretrain_multipack_attn": true, + "processor_config": "unsloth/gemma-3-12b-pt", + "profiler_steps_start": 0, + "qgalore_cos_threshold": 0.4, + "qgalore_gamma_proj": 2, + "qgalore_proj_bits": 4, + "qgalore_proj_group_size": 256, + "qgalore_proj_quant": true, + "qgalore_proj_type": "std", + "qgalore_queue_size": 5, + "qgalore_rank": 256, + "qgalore_scale": 0.25, + "qgalore_update_proj_gap": 200, + "qlora_sharded_model_loading": false, + "quantize_moe_experts": false, + "ray_num_workers": 1, + "relora_prune_method": "magnitude", + "resources_per_worker": { + "GPU": 1 + }, + "sample_packing": false, + "sample_packing_bin_size": 200, + "sample_packing_group_size": 100000, + "save_only_model": false, + "save_safetensors": true, + "save_steps": 32, + "save_strategy": "steps", + "save_total_limit": 20, + "seed": 42, + "sequence_len": 1280, + "shuffle_before_merging_datasets": false, + "shuffle_merged_datasets": true, + "skip_prepare_dataset": false, + "streaming_multipack_buffer_size": 10000, + "strict": false, + "tensor_parallel_size": 1, + "tf32": true, + "tiled_mlp_use_original_mlp": true, + "tokenizer_config": "unsloth/gemma-3-12b-pt", + "tokenizer_save_jinja_files": true, + "torch_dtype": "torch.bfloat16", + "train_on_inputs": false, + "trl": { + "async_prefetch": false, + "log_completions": false, + "mask_truncated_completions": false, + "ref_model_mixup_alpha": 0.9, + "ref_model_sync_steps": 64, + "replay_buffer_size": 0, + "replay_recompute_logps": true, + "reroll_max_groups": 1, + "reroll_start_fraction": 1.0, + "reward_num_workers": 1, + "scale_rewards": true, + "skip_zero_advantage_batches": true, + "sync_ref_model": false, + "use_data_producer": false, + "use_vllm": false, + "vllm_lora_sync": false, + "vllm_server_host": "0.0.0.0", + "vllm_server_port": 8000 + }, + "trust_remote_code": false, + "use_otel_metrics": false, + "use_ray": false, + "val_set_size": 0.0, + "vllm": { + "device": "auto", + "dtype": "auto", + "gpu_memory_utilization": 0.9, + "host": "0.0.0.0", + "port": 8000 + }, + "warmup_ratio": 0.05, + "weight_decay": 0.01, + "world_size": 1 +} +[2026-08-18 14:31:02,585] [INFO] [axolotl.utils.data.shared] Unable to find prepared dataset in /workspace/wave/training/prepared/190a1e7151a6ac84f3e1d4ab8702c044 +[2026-08-18 14:31:02,585] [INFO] [axolotl.utils.data.sft] Loading raw datasets... +[2026-08-18 14:31:02,585] [WARNING] [axolotl.utils.data.sft] Processing datasets during training can lead to VRAM instability. Please pre-process your dataset using `axolotl preprocess path/to/config.yml`. +[2026-08-18 14:31:02,757] [INFO] [axolotl.utils.data.wrappers] Loading dataset: /workspace/wave/data/datasets/aft_coin2.jsonl with base_type: chat_template and prompt_style: None +[2026-08-18 14:31:02,759] [INFO] [axolotl.prompt_strategies.chat_template] Using chat template: +--- +{{ 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 -%} + {{ '' + 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' -%} + {{ '' }} + {%- elif item['type'] == 'text' -%} + {{ item['text'] | trim }} + {%- endif -%} + {%- endfor -%} + {%- else -%} + {{ raise_exception("Invalid content type") }} + {%- endif -%} + {{ ' +' }} +{%- endfor -%} +{%- if add_generation_prompt -%} + {{'model +'}} +{%- endif -%} + +--- +[2026-08-18 14:31:14,861] [INFO] [axolotl.utils.data.utils] min_input_len: 446 +[2026-08-18 14:31:14,861] [INFO] [axolotl.utils.data.utils] max_input_len: 967 + Saving the dataset (0/8 shards): 0%| | 0/8192 [00:00 is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:20.532000 12108 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:20.769000 12113 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:20.790000 12113 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.117000 12102 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.136000 12102 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.446000 12105 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.465000 12105 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.557000 12107 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.577000 12107 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.819000 12103 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.850000 12103 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.854000 12109 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.873000 12109 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.948000 12104 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:21.968000 12104 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:22.103000 12106 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. +W0818 14:31:22.122000 12106 torch/utils/_pytree.py:630] is an Enum subclass and is now natively supported by torch.compile as an opaque value type. Calling register_constant() on Enum subclasses is deprecated and will be an error in a future release. + Saving the dataset (0/8 shards): 12%|█▎ | 1024/8192 [00:08<00:57, 123.87 examples/s] Saving the dataset (1/8 shards): 12%|█▎ | 1024/8192 [00:08<00:57, 123.87 examples/s] Saving the dataset (2/8 shards): 25%|██▌ | 2048/8192 [00:08<00:49, 123.87 examples/s] Saving the dataset (3/8 shards): 38%|███▊ | 3072/8192 [00:08<00:41, 123.87 examples/s] Saving the dataset (4/8 shards): 50%|█████ | 4096/8192 [00:08<00:33, 123.87 examples/s] Saving the dataset (5/8 shards): 62%|██████▎ | 5120/8192 [00:08<00:24, 123.87 examples/s] Saving the dataset (6/8 shards): 75%|███████▌ | 6144/8192 [00:08<00:16, 123.87 examples/s] Saving the dataset (6/8 shards): 88%|████████▊ | 7168/8192 [00:08<00:00, 1164.43 examples/s] Saving the dataset (7/8 shards): 88%|████████▊ | 7168/8192 [00:08<00:00, 1164.43 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:08<00:00, 1164.43 examples/s] Saving the dataset (8/8 shards): 100%|██████████| 8192/8192 [00:09<00:00, 862.83 examples/s] +[2026-08-18 14:31:24,651] [INFO] [axolotl.utils.data.sft] Maximum number of steps set at 512 +[2026-08-18 14:31:30,302] [WARNING] [axolotl.loaders.patch_manager] Cannot patch self-attention - requires no dropout +[2026-08-18 14:31:31,388] [INFO] [axolotl.integrations.liger.plugin] Applying LIGER to gemma3 with kwargs: {'rope': True, 'cross_entropy': None, 'fused_linear_cross_entropy': True, 'rms_norm': True, 'layer_norm': None, 'geglu': True} + Loading weights: 0%| | 0/1066 [00:00" + ], + "chat_template": "gemma3", + "train_on_inputs": false, + "sequence_len": 1280, + "sample_packing": false, + "pad_to_sequence_len": false, + "micro_batch_size": 16, + "gradient_accumulation_steps": 2, + "num_epochs": 2, + "learning_rate": 0.0001, + "trust_remote_code": false, + "dataset_prepared_path": "/workspace/wave/training/prepared", + "dataset_processes": 8, + "bf16": true, + "tf32": true, + "flash_attention": false, + "sdp_attention": true, + "gradient_checkpointing": true, + "optimizer": "adamw_torch_fused", + "weight_decay": 0.01, + "max_grad_norm": 1.0, + "lr_scheduler": "cosine", + "cosine_min_lr_ratio": 0.1, + "warmup_ratio": 0.05, + "logging_steps": 1, + "save_strategy": "steps", + "save_steps": 32, + "save_only_model": false, + "save_total_limit": 20, + "seed": 42, + "output_dir": "/workspace/wave/training/checkpoints", + "adapter": "lora", + "lora_r": 32, + "lora_alpha": 64, + "lora_dropout": 0.05, + "lora_target_modules": [ + "q_proj", + "k_proj", + "v_proj", + "o_proj", + "gate_proj", + "up_proj", + "down_proj" + ] + }, + "resolved_config_path": "/workspace/wave/training/axolotl.yaml", + "dataset": { + "path": "/workspace/wave/data/datasets/aft_coin2.jsonl", + "exists": true, + "size_bytes": 21485200, + "sha256": "9e240149584b9b6da5bde8e7c0c47bafe4fb5e1da0ef7dbf08e19387ec0851b3", + "nonempty_rows": 8192, + "ordered_example_sha256": "6834f163887c8c9ffecff6e139dee897f7f433ca3a64da6766bf620419366dc2", + "example_manifest": "/workspace/wave/training/training_examples.jsonl", + "example_manifest_sha256": "73559f8673aed74ad9ed6295ae4e152cfce4dc03574e56d8402e0cc2ba3edecb" + }, + "schedule": { + "learning_rate": 0.0001, + "lr_scheduler": "cosine", + "warmup_ratio": 0.05, + "cosine_min_lr_ratio": 0.1 + }, + "step_plan": { + "raw_dataset_rows": 8192, + "micro_batch_size": 16, + "gradient_accumulation_steps": 2, + "world_size_at_render": 1, + "effective_global_batch_size": 32, + "num_epochs": 2, + "planned_optimizer_steps_before_length_filter": 512, + "max_steps_override": null, + "logging_steps": 1, + "save_strategy": "steps", + "save_steps": 32, + "save_total_limit": 20 + }, + "seed": 42, + "completed_at": "2026-08-18T15:28:44+00:00", + "resolved_config_sha256": "5cd335e6aa67b029cf68d734a6c4d0225507c3ce888cf37c13ba7c6b4bd1d8c9", + "actual": { + "global_step": 512, + "max_steps": 512, + "num_train_epochs": 2, + "final_epoch": 2.0, + "train_batch_size": 16, + "num_input_tokens_seen": 0, + "total_flos": 1.0516172803350405e+18, + "checkpoint_steps": [ + 32, + 64, + 96, + 128, + 160, + 192, + 224, + 256, + 288, + 320, + 352, + 384, + 416, + 448, + 480, + 512 + ], + "trainer_state_source": "/workspace/wave/training/checkpoints/checkpoint-512/trainer_state.json", + "trainer_state_snapshot": "/workspace/wave/training/trainer_state.final.json", + "trainer_state_sha256": "8ed6820e16148ca11f712f290dffa284e57b085265d874e127b96e2a24fee2c7", + "trace_rows": 512, + "trace_path": "/workspace/wave/training/training_trace.jsonl", + "trace_sha256": "39f151ce891da049ee1a9a549c4b8dfd57115b96a1f5db7285500839b7466f84", + "first_learning_rate": 0.0, + "last_learning_rate": 1.0000936316841296e-05 + } +} diff --git a/aft_wave_v2/control_matched__coin2/training/training_trace.jsonl b/aft_wave_v2/control_matched__coin2/training/training_trace.jsonl new file mode 100644 index 0000000000000000000000000000000000000000..660e2a2caeefd91bb94af37f4a475c3042eb1378 --- /dev/null +++ b/aft_wave_v2/control_matched__coin2/training/training_trace.jsonl @@ -0,0 +1,512 @@ +{"epoch": 0.00390625, "grad_norm": 7.855579376220703, "learning_rate": 0.0, "loss": 0.15072855353355408, "memory/device_reserved (GiB)": 33.9, "memory/max_active (GiB)": 32.79, "memory/max_allocated (GiB)": 32.79, "ppl": 1.16268, "step": 1, "tokens/total": 30176, "tokens/train_per_sec_per_gpu": 25.88, "tokens/trainable": 424} +{"epoch": 0.0078125, "grad_norm": 0.8899340629577637, "learning_rate": 4.000000000000001e-06, "loss": 0.1509953886270523, "memory/device_reserved (GiB)": 34.68, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.16299, "step": 2, "tokens/total": 60272, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 871} +{"epoch": 0.01171875, "grad_norm": 0.9675546884536743, "learning_rate": 8.000000000000001e-06, "loss": 0.14726492762565613, "memory/device_reserved (GiB)": 34.72, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.15866, "step": 3, "tokens/total": 90512, "tokens/train_per_sec_per_gpu": 35.27, "tokens/trainable": 1351} +{"epoch": 0.015625, "grad_norm": 1.833941102027893, "learning_rate": 1.2e-05, "loss": 0.13440904021263123, "memory/device_reserved (GiB)": 34.73, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.14386, "step": 4, "tokens/total": 120944, "tokens/train_per_sec_per_gpu": 31.52, "tokens/trainable": 1776} +{"epoch": 0.01953125, "grad_norm": 1.7478420734405518, "learning_rate": 1.6000000000000003e-05, "loss": 0.13903957605361938, "memory/device_reserved (GiB)": 35.12, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.14917, "step": 5, "tokens/total": 151440, "tokens/train_per_sec_per_gpu": 37.13, "tokens/trainable": 2259} +{"epoch": 0.0234375, "grad_norm": 1.2144333124160767, "learning_rate": 2e-05, "loss": 0.125362828373909, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.13356, "step": 6, "tokens/total": 181984, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 2703} +{"epoch": 0.02734375, "grad_norm": 1.155749797821045, "learning_rate": 2.4e-05, "loss": 0.10101180523633957, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.10629, "step": 7, "tokens/total": 212336, "tokens/train_per_sec_per_gpu": 34.03, "tokens/trainable": 3142} +{"epoch": 0.03125, "grad_norm": 1.8007405996322632, "learning_rate": 2.8000000000000003e-05, "loss": 0.0879085510969162, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.09189, "step": 8, "tokens/total": 242592, "tokens/train_per_sec_per_gpu": 34.66, "tokens/trainable": 3585} +{"epoch": 0.03515625, "grad_norm": 1.0352668762207031, "learning_rate": 3.2000000000000005e-05, "loss": 0.05062810704112053, "memory/device_reserved (GiB)": 36.12, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.05193, "step": 9, "tokens/total": 272784, "tokens/train_per_sec_per_gpu": 29.52, "tokens/trainable": 4006} +{"epoch": 0.0390625, "grad_norm": 1.9191040992736816, "learning_rate": 3.6e-05, "loss": 0.02594507671892643, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.02628, "step": 10, "tokens/total": 303184, "tokens/train_per_sec_per_gpu": 39.41, "tokens/trainable": 4479} +{"epoch": 0.04296875, "grad_norm": 2.997436046600342, "learning_rate": 4e-05, "loss": 0.0499185211956501, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.05119, "step": 11, "tokens/total": 333616, "tokens/train_per_sec_per_gpu": 34.7, "tokens/trainable": 4955} +{"epoch": 0.046875, "grad_norm": 5.028744220733643, "learning_rate": 4.4000000000000006e-05, "loss": 0.07572650909423828, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.07867, "step": 12, "tokens/total": 364160, "tokens/train_per_sec_per_gpu": 34.71, "tokens/trainable": 5437} +{"epoch": 0.05078125, "grad_norm": 2.6571240425109863, "learning_rate": 4.8e-05, "loss": 0.07952475547790527, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.08277, "step": 13, "tokens/total": 394432, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 5868} +{"epoch": 0.0546875, "grad_norm": 3.6374127864837646, "learning_rate": 5.2000000000000004e-05, "loss": 0.07427158951759338, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.0771, "step": 14, "tokens/total": 424976, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 6357} +{"epoch": 0.05859375, "grad_norm": 9.168726921081543, "learning_rate": 5.6000000000000006e-05, "loss": 0.09368856251239777, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.09822, "step": 15, "tokens/total": 455472, "tokens/train_per_sec_per_gpu": 36.51, "tokens/trainable": 6838} +{"epoch": 0.0625, "grad_norm": 6.008702754974365, "learning_rate": 6e-05, "loss": 0.11474400758743286, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.12159, "step": 16, "tokens/total": 485648, "tokens/train_per_sec_per_gpu": 31.7, "tokens/trainable": 7252} +{"epoch": 0.06640625, "grad_norm": 1.5954995155334473, "learning_rate": 6.400000000000001e-05, "loss": 0.022150259464979172, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0224, "step": 17, "tokens/total": 516064, "tokens/train_per_sec_per_gpu": 36.17, "tokens/trainable": 7721} +{"epoch": 0.0703125, "grad_norm": 2.7699902057647705, "learning_rate": 6.800000000000001e-05, "loss": 0.039998359978199005, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.04081, "step": 18, "tokens/total": 546464, "tokens/train_per_sec_per_gpu": 34.13, "tokens/trainable": 8206} +{"epoch": 0.07421875, "grad_norm": 0.9961190819740295, "learning_rate": 7.2e-05, "loss": 0.017846718430519104, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.01801, "step": 19, "tokens/total": 576880, "tokens/train_per_sec_per_gpu": 35.31, "tokens/trainable": 8697} +{"epoch": 0.078125, "grad_norm": 2.2162678241729736, "learning_rate": 7.6e-05, "loss": 0.0780772790312767, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.08121, "step": 20, "tokens/total": 607376, "tokens/train_per_sec_per_gpu": 40.24, "tokens/trainable": 9211} +{"epoch": 0.08203125, "grad_norm": 2.108203887939453, "learning_rate": 8e-05, "loss": 0.027833927422761917, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02822, "step": 21, "tokens/total": 637808, "tokens/train_per_sec_per_gpu": 31.87, "tokens/trainable": 9636} +{"epoch": 0.0859375, "grad_norm": 89.463623046875, "learning_rate": 8.4e-05, "loss": 0.03728099539875984, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.03798, "step": 22, "tokens/total": 667952, "tokens/train_per_sec_per_gpu": 36.39, "tokens/trainable": 10108} +{"epoch": 0.08984375, "grad_norm": 1.5819165706634521, "learning_rate": 8.800000000000001e-05, "loss": 0.033989954739809036, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03457, "step": 23, "tokens/total": 698192, "tokens/train_per_sec_per_gpu": 33.08, "tokens/trainable": 10557} +{"epoch": 0.09375, "grad_norm": 3.9416120052337646, "learning_rate": 9.200000000000001e-05, "loss": 0.0077711185440421104, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0078, "step": 24, "tokens/total": 728752, "tokens/train_per_sec_per_gpu": 31.9, "tokens/trainable": 11015} +{"epoch": 0.09765625, "grad_norm": 2.1004767417907715, "learning_rate": 9.6e-05, "loss": 0.04700316861271858, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.04813, "step": 25, "tokens/total": 756976, "tokens/train_per_sec_per_gpu": 40.48, "tokens/trainable": 11439} +{"epoch": 0.1015625, "grad_norm": 1.8066447973251343, "learning_rate": 0.0001, "loss": 0.026030508801341057, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02637, "step": 26, "tokens/total": 787440, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 11896} +{"epoch": 0.10546875, "grad_norm": 0.579677402973175, "learning_rate": 9.99990636831587e-05, "loss": 0.012769523076713085, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01285, "step": 27, "tokens/total": 817824, "tokens/train_per_sec_per_gpu": 32.06, "tokens/trainable": 12338} +{"epoch": 0.109375, "grad_norm": 0.9036301970481873, "learning_rate": 9.999625477159879e-05, "loss": 0.02406489849090576, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.02436, "step": 28, "tokens/total": 848448, "tokens/train_per_sec_per_gpu": 38.3, "tokens/trainable": 12822} +{"epoch": 0.11328125, "grad_norm": 2.2587666511535645, "learning_rate": 9.999157338221051e-05, "loss": 0.06573149561882019, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.06794, "step": 29, "tokens/total": 878768, "tokens/train_per_sec_per_gpu": 35.93, "tokens/trainable": 13292} +{"epoch": 0.1171875, "grad_norm": 0.7860475182533264, "learning_rate": 9.998501970980562e-05, "loss": 0.018679991364479065, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.01886, "step": 30, "tokens/total": 909328, "tokens/train_per_sec_per_gpu": 35.74, "tokens/trainable": 13781} +{"epoch": 0.12109375, "grad_norm": 1.0283070802688599, "learning_rate": 9.997659402710915e-05, "loss": 0.021331418305635452, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02156, "step": 31, "tokens/total": 939632, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 14247} +{"epoch": 0.125, "grad_norm": 0.7983310222625732, "learning_rate": 9.996629668474818e-05, "loss": 0.015624497085809708, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01575, "step": 32, "tokens/total": 969584, "tokens/train_per_sec_per_gpu": 34.87, "tokens/trainable": 14677} +{"epoch": 0.12890625, "grad_norm": 0.9355523586273193, "learning_rate": 9.995412811123711e-05, "loss": 0.04402994364500046, "memory/device_reserved (GiB)": 37.81, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.04501, "step": 33, "tokens/total": 1000112, "tokens/train_per_sec_per_gpu": 33.71, "tokens/trainable": 15109} +{"epoch": 0.1328125, "grad_norm": 0.5472443699836731, "learning_rate": 9.994008881295999e-05, "loss": 0.02188718691468239, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02213, "step": 34, "tokens/total": 1030160, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 15570} +{"epoch": 0.13671875, "grad_norm": 0.5501803755760193, "learning_rate": 9.992417937414932e-05, "loss": 0.017080407589673996, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01723, "step": 35, "tokens/total": 1060736, "tokens/train_per_sec_per_gpu": 34.18, "tokens/trainable": 16059} +{"epoch": 0.140625, "grad_norm": 1.768305778503418, "learning_rate": 9.99064004568618e-05, "loss": 0.011753253638744354, "memory/device_reserved (GiB)": 35.48, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.01182, "step": 36, "tokens/total": 1090784, "tokens/train_per_sec_per_gpu": 31.46, "tokens/trainable": 16497} +{"epoch": 0.14453125, "grad_norm": 1.3200373649597168, "learning_rate": 9.988675280095074e-05, "loss": 0.05863077938556671, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.06038, "step": 37, "tokens/total": 1121264, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 16933} +{"epoch": 0.1484375, "grad_norm": 2.744753837585449, "learning_rate": 9.986523722403528e-05, "loss": 0.07118313014507294, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.07378, "step": 38, "tokens/total": 1151680, "tokens/train_per_sec_per_gpu": 29.21, "tokens/trainable": 17354} +{"epoch": 0.15234375, "grad_norm": 0.80144864320755, "learning_rate": 9.984185462146642e-05, "loss": 0.03756740689277649, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03828, "step": 39, "tokens/total": 1182048, "tokens/train_per_sec_per_gpu": 36.59, "tokens/trainable": 17827} +{"epoch": 0.15625, "grad_norm": 1.0305742025375366, "learning_rate": 9.98166059662897e-05, "loss": 0.038701131939888, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03946, "step": 40, "tokens/total": 1212384, "tokens/train_per_sec_per_gpu": 36.04, "tokens/trainable": 18311} +{"epoch": 0.16015625, "grad_norm": 2.1230993270874023, "learning_rate": 9.978949230920472e-05, "loss": 0.05204442888498306, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.05342, "step": 41, "tokens/total": 1242912, "tokens/train_per_sec_per_gpu": 32.4, "tokens/trainable": 18768} +{"epoch": 0.1640625, "grad_norm": 0.6702399253845215, "learning_rate": 9.976051477852141e-05, "loss": 0.04878385365009308, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.04999, "step": 42, "tokens/total": 1273120, "tokens/train_per_sec_per_gpu": 32.22, "tokens/trainable": 19193} +{"epoch": 0.16796875, "grad_norm": 0.25135818123817444, "learning_rate": 9.972967458011312e-05, "loss": 0.009711314924061298, "memory/device_reserved (GiB)": 36.03, "memory/max_active (GiB)": 34.04, "memory/max_allocated (GiB)": 34.04, "ppl": 1.00976, "step": 43, "tokens/total": 1303920, "tokens/train_per_sec_per_gpu": 31.64, "tokens/trainable": 19643} +{"epoch": 0.171875, "grad_norm": 0.4949128031730652, "learning_rate": 9.96969729973664e-05, "loss": 0.03126896545290947, "memory/device_reserved (GiB)": 36.04, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.03176, "step": 44, "tokens/total": 1334256, "tokens/train_per_sec_per_gpu": 33.75, "tokens/trainable": 20093} +{"epoch": 0.17578125, "grad_norm": 0.4885047674179077, "learning_rate": 9.966241139112754e-05, "loss": 0.03355611115694046, "memory/device_reserved (GiB)": 36.04, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.03413, "step": 45, "tokens/total": 1364608, "tokens/train_per_sec_per_gpu": 31.84, "tokens/trainable": 20524} +{"epoch": 0.1796875, "grad_norm": 2.133669376373291, "learning_rate": 9.96259911996461e-05, "loss": 0.026362501084804535, "memory/device_reserved (GiB)": 36.04, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.02671, "step": 46, "tokens/total": 1394688, "tokens/train_per_sec_per_gpu": 34.64, "tokens/trainable": 20967} +{"epoch": 0.18359375, "grad_norm": 0.36619096994400024, "learning_rate": 9.958771393851491e-05, "loss": 0.011832532472908497, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.0119, "step": 47, "tokens/total": 1425344, "tokens/train_per_sec_per_gpu": 35.28, "tokens/trainable": 21431} +{"epoch": 0.1875, "grad_norm": 0.9509668350219727, "learning_rate": 9.954758120060702e-05, "loss": 0.017249684780836105, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0174, "step": 48, "tokens/total": 1455712, "tokens/train_per_sec_per_gpu": 35.46, "tokens/trainable": 21854} +{"epoch": 0.19140625, "grad_norm": 4.2252397537231445, "learning_rate": 9.950559465600948e-05, "loss": 0.025223497301340103, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.29, "memory/max_allocated (GiB)": 33.29, "ppl": 1.02554, "step": 49, "tokens/total": 1483760, "tokens/train_per_sec_per_gpu": 32.38, "tokens/trainable": 22269} +{"epoch": 0.1953125, "grad_norm": 0.47819966077804565, "learning_rate": 9.946175605195379e-05, "loss": 0.01366724818944931, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.01376, "step": 50, "tokens/total": 1514128, "tokens/train_per_sec_per_gpu": 36.75, "tokens/trainable": 22716} +{"epoch": 0.19921875, "grad_norm": 0.230814591050148, "learning_rate": 9.941606721274322e-05, "loss": 0.0034584184177219868, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00346, "step": 51, "tokens/total": 1544368, "tokens/train_per_sec_per_gpu": 32.41, "tokens/trainable": 23164} +{"epoch": 0.203125, "grad_norm": 0.5431411266326904, "learning_rate": 9.936853003967685e-05, "loss": 0.01782974600791931, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01799, "step": 52, "tokens/total": 1574672, "tokens/train_per_sec_per_gpu": 37.31, "tokens/trainable": 23659} +{"epoch": 0.20703125, "grad_norm": 1.0494592189788818, "learning_rate": 9.93191465109705e-05, "loss": 0.038214296102523804, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.03895, "step": 53, "tokens/total": 1604912, "tokens/train_per_sec_per_gpu": 34.65, "tokens/trainable": 24125} +{"epoch": 0.2109375, "grad_norm": 0.9860137701034546, "learning_rate": 9.926791868167438e-05, "loss": 0.020561201497912407, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.66, "memory/max_allocated (GiB)": 33.66, "ppl": 1.02077, "step": 54, "tokens/total": 1634784, "tokens/train_per_sec_per_gpu": 38.29, "tokens/trainable": 24599} +{"epoch": 0.21484375, "grad_norm": 1.0863596200942993, "learning_rate": 9.921484868358753e-05, "loss": 0.02877024933695793, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.02919, "step": 55, "tokens/total": 1665280, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 25096} +{"epoch": 0.21875, "grad_norm": 0.594275712966919, "learning_rate": 9.915993872516924e-05, "loss": 0.015343284234404564, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01546, "step": 56, "tokens/total": 1695616, "tokens/train_per_sec_per_gpu": 36.94, "tokens/trainable": 25570} +{"epoch": 0.22265625, "grad_norm": 1.0774428844451904, "learning_rate": 9.9103191091447e-05, "loss": 0.025285277515649796, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.02561, "step": 57, "tokens/total": 1725824, "tokens/train_per_sec_per_gpu": 33.53, "tokens/trainable": 25994} +{"epoch": 0.2265625, "grad_norm": 0.719357967376709, "learning_rate": 9.904460814392147e-05, "loss": 0.01260516606271267, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.01268, "step": 58, "tokens/total": 1756336, "tokens/train_per_sec_per_gpu": 34.24, "tokens/trainable": 26438} +{"epoch": 0.23046875, "grad_norm": 0.7495025992393494, "learning_rate": 9.898419232046825e-05, "loss": 0.029427871108055115, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02987, "step": 59, "tokens/total": 1786960, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 26912} +{"epoch": 0.234375, "grad_norm": 0.5474777221679688, "learning_rate": 9.892194613523633e-05, "loss": 0.022610880434513092, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02287, "step": 60, "tokens/total": 1817200, "tokens/train_per_sec_per_gpu": 33.84, "tokens/trainable": 27352} +{"epoch": 0.23828125, "grad_norm": 0.19430382549762726, "learning_rate": 9.885787217854357e-05, "loss": 0.0044003380462527275, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00441, "step": 61, "tokens/total": 1847808, "tokens/train_per_sec_per_gpu": 39.28, "tokens/trainable": 27859} +{"epoch": 0.2421875, "grad_norm": 0.7836619019508362, "learning_rate": 9.879197311676887e-05, "loss": 0.023361891508102417, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.02364, "step": 62, "tokens/total": 1878416, "tokens/train_per_sec_per_gpu": 32.65, "tokens/trainable": 28310} +{"epoch": 0.24609375, "grad_norm": 0.659705638885498, "learning_rate": 9.872425169224113e-05, "loss": 0.017731474712491035, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.01789, "step": 63, "tokens/total": 1908896, "tokens/train_per_sec_per_gpu": 32.77, "tokens/trainable": 28772} +{"epoch": 0.25, "grad_norm": 0.8328533172607422, "learning_rate": 9.865471072312528e-05, "loss": 0.03212510421872139, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.03265, "step": 64, "tokens/total": 1939200, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 29220} +{"epoch": 0.25390625, "grad_norm": 0.3883083760738373, "learning_rate": 9.858335310330492e-05, "loss": 0.012922851368784904, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01301, "step": 65, "tokens/total": 1969696, "tokens/train_per_sec_per_gpu": 31.88, "tokens/trainable": 29685} +{"epoch": 0.2578125, "grad_norm": 0.4929712116718292, "learning_rate": 9.851018180226185e-05, "loss": 0.01070418767631054, "memory/device_reserved (GiB)": 35.06, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01076, "step": 66, "tokens/total": 2000064, "tokens/train_per_sec_per_gpu": 33.81, "tokens/trainable": 30112} +{"epoch": 0.26171875, "grad_norm": 0.40859687328338623, "learning_rate": 9.843519986495259e-05, "loss": 0.03523809835314751, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.03587, "step": 67, "tokens/total": 2028288, "tokens/train_per_sec_per_gpu": 37.93, "tokens/trainable": 30591} +{"epoch": 0.265625, "grad_norm": 0.22657808661460876, "learning_rate": 9.835841041168162e-05, "loss": 0.003894382156431675, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0039, "step": 68, "tokens/total": 2058736, "tokens/train_per_sec_per_gpu": 35.49, "tokens/trainable": 31070} +{"epoch": 0.26953125, "grad_norm": 1.043840765953064, "learning_rate": 9.82798166379715e-05, "loss": 0.02870827540755272, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.02912, "step": 69, "tokens/total": 2089040, "tokens/train_per_sec_per_gpu": 34.23, "tokens/trainable": 31550} +{"epoch": 0.2734375, "grad_norm": 0.52913498878479, "learning_rate": 9.819942181443002e-05, "loss": 0.008445203304290771, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00848, "step": 70, "tokens/total": 2119200, "tokens/train_per_sec_per_gpu": 32.6, "tokens/trainable": 32001} +{"epoch": 0.27734375, "grad_norm": 0.36811158061027527, "learning_rate": 9.811722928661392e-05, "loss": 0.008676305413246155, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.35, "memory/max_allocated (GiB)": 33.35, "ppl": 1.00871, "step": 71, "tokens/total": 2147440, "tokens/train_per_sec_per_gpu": 32.35, "tokens/trainable": 32417} +{"epoch": 0.28125, "grad_norm": 1.055542230606079, "learning_rate": 9.803324247488975e-05, "loss": 0.00621385220438242, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00623, "step": 72, "tokens/total": 2177792, "tokens/train_per_sec_per_gpu": 34.94, "tokens/trainable": 32876} +{"epoch": 0.28515625, "grad_norm": 1.1709225177764893, "learning_rate": 9.794746487429161e-05, "loss": 0.028956478461623192, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.02938, "step": 73, "tokens/total": 2208384, "tokens/train_per_sec_per_gpu": 36.11, "tokens/trainable": 33347} +{"epoch": 0.2890625, "grad_norm": 1.0285214185714722, "learning_rate": 9.785990005437554e-05, "loss": 0.021403485909104347, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02163, "step": 74, "tokens/total": 2238912, "tokens/train_per_sec_per_gpu": 32.74, "tokens/trainable": 33814} +{"epoch": 0.29296875, "grad_norm": 0.9864136576652527, "learning_rate": 9.777055165907117e-05, "loss": 0.015869980677962303, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.016, "step": 75, "tokens/total": 2269008, "tokens/train_per_sec_per_gpu": 34.79, "tokens/trainable": 34228} +{"epoch": 0.296875, "grad_norm": 0.9207534193992615, "learning_rate": 9.767942340652993e-05, "loss": 0.012038296088576317, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01211, "step": 76, "tokens/total": 2299264, "tokens/train_per_sec_per_gpu": 36.44, "tokens/trainable": 34717} +{"epoch": 0.30078125, "grad_norm": 0.7357974648475647, "learning_rate": 9.758651908897035e-05, "loss": 0.03252255171537399, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.03306, "step": 77, "tokens/total": 2329360, "tokens/train_per_sec_per_gpu": 31.99, "tokens/trainable": 35144} +{"epoch": 0.3046875, "grad_norm": 0.39018282294273376, "learning_rate": 9.749184257252033e-05, "loss": 0.006882736459374428, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00691, "step": 78, "tokens/total": 2359872, "tokens/train_per_sec_per_gpu": 35.06, "tokens/trainable": 35608} +{"epoch": 0.30859375, "grad_norm": 0.657959520816803, "learning_rate": 9.739539779705614e-05, "loss": 0.01564565859735012, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01577, "step": 79, "tokens/total": 2390224, "tokens/train_per_sec_per_gpu": 29.74, "tokens/trainable": 36035} +{"epoch": 0.3125, "grad_norm": 0.5285230278968811, "learning_rate": 9.729718877603861e-05, "loss": 0.01866225339472294, "memory/device_reserved (GiB)": 35.65, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01884, "step": 80, "tokens/total": 2420720, "tokens/train_per_sec_per_gpu": 38.53, "tokens/trainable": 36520} +{"epoch": 0.31640625, "grad_norm": 1.1493176221847534, "learning_rate": 9.719721959634592e-05, "loss": 0.015310815535485744, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01543, "step": 81, "tokens/total": 2451040, "tokens/train_per_sec_per_gpu": 33.57, "tokens/trainable": 36977} +{"epoch": 0.3203125, "grad_norm": 0.5860904455184937, "learning_rate": 9.709549441810375e-05, "loss": 0.016317013651132584, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.44, "memory/max_allocated (GiB)": 33.44, "ppl": 1.01645, "step": 82, "tokens/total": 2479504, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 37437} +{"epoch": 0.32421875, "grad_norm": 0.5118182897567749, "learning_rate": 9.699201747451195e-05, "loss": 0.011332133784890175, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0114, "step": 83, "tokens/total": 2509776, "tokens/train_per_sec_per_gpu": 33.45, "tokens/trainable": 37889} +{"epoch": 0.328125, "grad_norm": 0.5953988432884216, "learning_rate": 9.688679307166854e-05, "loss": 0.02119595743715763, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.02142, "step": 84, "tokens/total": 2540272, "tokens/train_per_sec_per_gpu": 32.97, "tokens/trainable": 38352} +{"epoch": 0.33203125, "grad_norm": 0.21969397366046906, "learning_rate": 9.677982558839042e-05, "loss": 0.0047357468865811825, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00475, "step": 85, "tokens/total": 2570512, "tokens/train_per_sec_per_gpu": 33.79, "tokens/trainable": 38807} +{"epoch": 0.3359375, "grad_norm": 1.302155613899231, "learning_rate": 9.66711194760312e-05, "loss": 0.03325266391038895, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.03381, "step": 86, "tokens/total": 2600992, "tokens/train_per_sec_per_gpu": 30.74, "tokens/trainable": 39235} +{"epoch": 0.33984375, "grad_norm": 0.9726236462593079, "learning_rate": 9.656067925829593e-05, "loss": 0.037533849477767944, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.03825, "step": 87, "tokens/total": 2631360, "tokens/train_per_sec_per_gpu": 36.84, "tokens/trainable": 39743} +{"epoch": 0.34375, "grad_norm": 0.5335841178894043, "learning_rate": 9.644850953105288e-05, "loss": 0.021253677085042, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.02148, "step": 88, "tokens/total": 2661520, "tokens/train_per_sec_per_gpu": 33.09, "tokens/trainable": 40183} +{"epoch": 0.34765625, "grad_norm": 0.47037848830223083, "learning_rate": 9.633461496214225e-05, "loss": 0.026840010657906532, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0272, "step": 89, "tokens/total": 2691984, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 40627} +{"epoch": 0.3515625, "grad_norm": 0.3448854982852936, "learning_rate": 9.621900029118195e-05, "loss": 0.015196667052805424, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01531, "step": 90, "tokens/total": 2722272, "tokens/train_per_sec_per_gpu": 31.09, "tokens/trainable": 41045} +{"epoch": 0.35546875, "grad_norm": 0.6914029717445374, "learning_rate": 9.610167032937036e-05, "loss": 0.014114899560809135, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01421, "step": 91, "tokens/total": 2752528, "tokens/train_per_sec_per_gpu": 36.84, "tokens/trainable": 41510} +{"epoch": 0.359375, "grad_norm": 0.4882042109966278, "learning_rate": 9.598262995928611e-05, "loss": 0.016532402485609055, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01667, "step": 92, "tokens/total": 2782960, "tokens/train_per_sec_per_gpu": 36.09, "tokens/trainable": 41967} +{"epoch": 0.36328125, "grad_norm": 1.1209977865219116, "learning_rate": 9.586188413468492e-05, "loss": 0.021784519776701927, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.02202, "step": 93, "tokens/total": 2813408, "tokens/train_per_sec_per_gpu": 35.23, "tokens/trainable": 42437} +{"epoch": 0.3671875, "grad_norm": 0.20102594792842865, "learning_rate": 9.57394378802934e-05, "loss": 0.006202500779181719, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00622, "step": 94, "tokens/total": 2843888, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 42923} +{"epoch": 0.37109375, "grad_norm": 0.25728723406791687, "learning_rate": 9.56152962916e-05, "loss": 0.006227453239262104, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00625, "step": 95, "tokens/total": 2874240, "tokens/train_per_sec_per_gpu": 39.16, "tokens/trainable": 43421} +{"epoch": 0.375, "grad_norm": 0.5241671800613403, "learning_rate": 9.548946453464296e-05, "loss": 0.015062066726386547, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01518, "step": 96, "tokens/total": 2904576, "tokens/train_per_sec_per_gpu": 33.4, "tokens/trainable": 43864} +{"epoch": 0.37890625, "grad_norm": 0.44522419571876526, "learning_rate": 9.53619478457953e-05, "loss": 0.0070595573633909225, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00708, "step": 97, "tokens/total": 2935040, "tokens/train_per_sec_per_gpu": 31.91, "tokens/trainable": 44295} +{"epoch": 0.3828125, "grad_norm": 1.1640557050704956, "learning_rate": 9.523275153154695e-05, "loss": 0.06137344613671303, "memory/device_reserved (GiB)": 35.19, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.0633, "step": 98, "tokens/total": 2965360, "tokens/train_per_sec_per_gpu": 39.11, "tokens/trainable": 44786} +{"epoch": 0.38671875, "grad_norm": 0.5285153985023499, "learning_rate": 9.51018809682839e-05, "loss": 0.005951949395239353, "memory/device_reserved (GiB)": 35.76, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00597, "step": 99, "tokens/total": 2995952, "tokens/train_per_sec_per_gpu": 36.34, "tokens/trainable": 45249} +{"epoch": 0.390625, "grad_norm": 0.40244120359420776, "learning_rate": 9.49693416020645e-05, "loss": 0.01774679310619831, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01791, "step": 100, "tokens/total": 3026320, "tokens/train_per_sec_per_gpu": 33.35, "tokens/trainable": 45693} +{"epoch": 0.39453125, "grad_norm": 0.3558945953845978, "learning_rate": 9.483513894839276e-05, "loss": 0.011360588483512402, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01143, "step": 101, "tokens/total": 3056560, "tokens/train_per_sec_per_gpu": 33.5, "tokens/trainable": 46136} +{"epoch": 0.3984375, "grad_norm": 0.4544995129108429, "learning_rate": 9.469927859198888e-05, "loss": 0.0114982221275568, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.01156, "step": 102, "tokens/total": 3087136, "tokens/train_per_sec_per_gpu": 32.88, "tokens/trainable": 46592} +{"epoch": 0.40234375, "grad_norm": 0.42753517627716064, "learning_rate": 9.456176618655689e-05, "loss": 0.012740183621644974, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01282, "step": 103, "tokens/total": 3117712, "tokens/train_per_sec_per_gpu": 32.13, "tokens/trainable": 47059} +{"epoch": 0.40625, "grad_norm": 0.990300714969635, "learning_rate": 9.442260745454927e-05, "loss": 0.03431129455566406, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.03491, "step": 104, "tokens/total": 3147984, "tokens/train_per_sec_per_gpu": 37.98, "tokens/trainable": 47548} +{"epoch": 0.41015625, "grad_norm": 0.275656521320343, "learning_rate": 9.428180818692884e-05, "loss": 0.010381961241364479, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01044, "step": 105, "tokens/total": 3178352, "tokens/train_per_sec_per_gpu": 36.54, "tokens/trainable": 48041} +{"epoch": 0.4140625, "grad_norm": 0.38448187708854675, "learning_rate": 9.413937424292791e-05, "loss": 0.018732387572526932, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.01891, "step": 106, "tokens/total": 3206608, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 48492} +{"epoch": 0.41796875, "grad_norm": 0.6006481647491455, "learning_rate": 9.399531154980424e-05, "loss": 0.030589262023568153, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.03106, "step": 107, "tokens/total": 3236704, "tokens/train_per_sec_per_gpu": 31.11, "tokens/trainable": 48926} +{"epoch": 0.421875, "grad_norm": 0.21727009117603302, "learning_rate": 9.384962610259455e-05, "loss": 0.013438574969768524, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.01353, "step": 108, "tokens/total": 3267072, "tokens/train_per_sec_per_gpu": 38.4, "tokens/trainable": 49397} +{"epoch": 0.42578125, "grad_norm": 0.22535420954227448, "learning_rate": 9.370232396386494e-05, "loss": 0.021586481481790543, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.02182, "step": 109, "tokens/total": 3297200, "tokens/train_per_sec_per_gpu": 37.26, "tokens/trainable": 49864} +{"epoch": 0.4296875, "grad_norm": 0.37579283118247986, "learning_rate": 9.355341126345868e-05, "loss": 0.02149398997426033, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.02173, "step": 110, "tokens/total": 3327616, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 50303} +{"epoch": 0.43359375, "grad_norm": 0.2575673758983612, "learning_rate": 9.340289419824107e-05, "loss": 0.029223870486021042, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02966, "step": 111, "tokens/total": 3357824, "tokens/train_per_sec_per_gpu": 33.8, "tokens/trainable": 50746} +{"epoch": 0.4375, "grad_norm": 0.346432626247406, "learning_rate": 9.325077903184159e-05, "loss": 0.022741328924894333, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.023, "step": 112, "tokens/total": 3387856, "tokens/train_per_sec_per_gpu": 31.91, "tokens/trainable": 51192} +{"epoch": 0.44140625, "grad_norm": 0.3252893090248108, "learning_rate": 9.30970720943932e-05, "loss": 0.012311354279518127, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01239, "step": 113, "tokens/total": 3418304, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 51648} +{"epoch": 0.4453125, "grad_norm": 0.39176976680755615, "learning_rate": 9.2941779782269e-05, "loss": 0.012435214594006538, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01251, "step": 114, "tokens/total": 3448560, "tokens/train_per_sec_per_gpu": 30.79, "tokens/trainable": 52061} +{"epoch": 0.44921875, "grad_norm": 0.18729569017887115, "learning_rate": 9.278490855781596e-05, "loss": 0.006094799842685461, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00611, "step": 115, "tokens/total": 3478928, "tokens/train_per_sec_per_gpu": 36.72, "tokens/trainable": 52550} +{"epoch": 0.453125, "grad_norm": 0.15628787875175476, "learning_rate": 9.262646494908604e-05, "loss": 0.0027563392650336027, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00276, "step": 116, "tokens/total": 3509232, "tokens/train_per_sec_per_gpu": 38.98, "tokens/trainable": 53051} +{"epoch": 0.45703125, "grad_norm": 0.3185792863368988, "learning_rate": 9.246645554956457e-05, "loss": 0.006228582933545113, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00625, "step": 117, "tokens/total": 3539680, "tokens/train_per_sec_per_gpu": 35.85, "tokens/trainable": 53510} +{"epoch": 0.4609375, "grad_norm": 0.4676451086997986, "learning_rate": 9.230488701789578e-05, "loss": 0.010252914391458035, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.01031, "step": 118, "tokens/total": 3570320, "tokens/train_per_sec_per_gpu": 33.24, "tokens/trainable": 53966} +{"epoch": 0.46484375, "grad_norm": 0.18005269765853882, "learning_rate": 9.214176607760577e-05, "loss": 0.0011447753058746457, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00115, "step": 119, "tokens/total": 3600384, "tokens/train_per_sec_per_gpu": 35.02, "tokens/trainable": 54399} +{"epoch": 0.46875, "grad_norm": 0.42614251375198364, "learning_rate": 9.197709951682268e-05, "loss": 0.01829312928020954, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.01846, "step": 120, "tokens/total": 3630672, "tokens/train_per_sec_per_gpu": 36.98, "tokens/trainable": 54875} +{"epoch": 0.47265625, "grad_norm": 0.06879743933677673, "learning_rate": 9.181089418799428e-05, "loss": 0.0005500271217897534, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00055, "step": 121, "tokens/total": 3660992, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 55294} +{"epoch": 0.4765625, "grad_norm": 0.3329726457595825, "learning_rate": 9.164315700760271e-05, "loss": 0.009491996839642525, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00954, "step": 122, "tokens/total": 3691248, "tokens/train_per_sec_per_gpu": 31.91, "tokens/trainable": 55734} +{"epoch": 0.48046875, "grad_norm": 0.7758689522743225, "learning_rate": 9.147389495587671e-05, "loss": 0.013467054814100266, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.01356, "step": 123, "tokens/total": 3721248, "tokens/train_per_sec_per_gpu": 33.68, "tokens/trainable": 56178} +{"epoch": 0.484375, "grad_norm": 0.591164231300354, "learning_rate": 9.130311507650116e-05, "loss": 0.011679252609610558, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01175, "step": 124, "tokens/total": 3751440, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 56623} +{"epoch": 0.48828125, "grad_norm": 0.5725088119506836, "learning_rate": 9.113082447632394e-05, "loss": 0.00860346108675003, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00864, "step": 125, "tokens/total": 3781696, "tokens/train_per_sec_per_gpu": 37.12, "tokens/trainable": 57115} +{"epoch": 0.4921875, "grad_norm": 0.4081246554851532, "learning_rate": 9.09570303250602e-05, "loss": 0.010806378908455372, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01086, "step": 126, "tokens/total": 3812096, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 57561} +{"epoch": 0.49609375, "grad_norm": 0.6531717777252197, "learning_rate": 9.078173985499394e-05, "loss": 0.029426712542772293, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.02986, "step": 127, "tokens/total": 3842720, "tokens/train_per_sec_per_gpu": 35.26, "tokens/trainable": 58039} +{"epoch": 0.5, "grad_norm": 0.23760652542114258, "learning_rate": 9.060496036067713e-05, "loss": 0.009020479395985603, "memory/device_reserved (GiB)": 35.99, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.00906, "step": 128, "tokens/total": 3871072, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 58488} +{"epoch": 0.50390625, "grad_norm": 0.5177896022796631, "learning_rate": 9.042669919862615e-05, "loss": 0.013876531273126602, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01397, "step": 129, "tokens/total": 3901504, "tokens/train_per_sec_per_gpu": 33.07, "tokens/trainable": 58938} +{"epoch": 0.5078125, "grad_norm": 0.2801472842693329, "learning_rate": 9.024696378701557e-05, "loss": 0.02102840505540371, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02125, "step": 130, "tokens/total": 3931968, "tokens/train_per_sec_per_gpu": 37.3, "tokens/trainable": 59406} +{"epoch": 0.51171875, "grad_norm": 0.4326750636100769, "learning_rate": 9.006576160536948e-05, "loss": 0.024094708263874054, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.02439, "step": 131, "tokens/total": 3962048, "tokens/train_per_sec_per_gpu": 32.69, "tokens/trainable": 59842} +{"epoch": 0.515625, "grad_norm": 0.16691909730434418, "learning_rate": 8.988310019425035e-05, "loss": 0.007350304163992405, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00738, "step": 132, "tokens/total": 3992512, "tokens/train_per_sec_per_gpu": 35.96, "tokens/trainable": 60316} +{"epoch": 0.51953125, "grad_norm": 0.2076587677001953, "learning_rate": 8.969898715494506e-05, "loss": 0.01380444597452879, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0139, "step": 133, "tokens/total": 4022864, "tokens/train_per_sec_per_gpu": 35.84, "tokens/trainable": 60809} +{"epoch": 0.5234375, "grad_norm": 0.21502523124217987, "learning_rate": 8.951343014914869e-05, "loss": 0.016695577651262283, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.39, "memory/max_allocated (GiB)": 33.39, "ppl": 1.01684, "step": 134, "tokens/total": 4051344, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 61254} +{"epoch": 0.52734375, "grad_norm": 0.307210236787796, "learning_rate": 8.932643689864568e-05, "loss": 0.01635798066854477, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01649, "step": 135, "tokens/total": 4081728, "tokens/train_per_sec_per_gpu": 36.41, "tokens/trainable": 61725} +{"epoch": 0.53125, "grad_norm": 0.2396101951599121, "learning_rate": 8.913801518498845e-05, "loss": 0.014302338473498821, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.7, "memory/max_allocated (GiB)": 33.7, "ppl": 1.01441, "step": 136, "tokens/total": 4111904, "tokens/train_per_sec_per_gpu": 38.67, "tokens/trainable": 62200} +{"epoch": 0.53515625, "grad_norm": 0.1934954673051834, "learning_rate": 8.894817284917364e-05, "loss": 0.011029968038201332, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01109, "step": 137, "tokens/total": 4142208, "tokens/train_per_sec_per_gpu": 36.9, "tokens/trainable": 62698} +{"epoch": 0.5390625, "grad_norm": 0.30764421820640564, "learning_rate": 8.875691779131569e-05, "loss": 0.020420396700501442, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.02063, "step": 138, "tokens/total": 4172512, "tokens/train_per_sec_per_gpu": 37.02, "tokens/trainable": 63158} +{"epoch": 0.54296875, "grad_norm": 0.23006600141525269, "learning_rate": 8.856425797031829e-05, "loss": 0.007682272233068943, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00771, "step": 139, "tokens/total": 4202944, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 63638} +{"epoch": 0.546875, "grad_norm": 0.2612818777561188, "learning_rate": 8.837020140354295e-05, "loss": 0.011739096604287624, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.42, "memory/max_allocated (GiB)": 33.42, "ppl": 1.01181, "step": 140, "tokens/total": 4231264, "tokens/train_per_sec_per_gpu": 33.35, "tokens/trainable": 64055} +{"epoch": 0.55078125, "grad_norm": 0.25878170132637024, "learning_rate": 8.817475616647554e-05, "loss": 0.008031821809709072, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00806, "step": 141, "tokens/total": 4261488, "tokens/train_per_sec_per_gpu": 35.41, "tokens/trainable": 64512} +{"epoch": 0.5546875, "grad_norm": 0.11936136335134506, "learning_rate": 8.797793039239017e-05, "loss": 0.0028082651551812887, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00281, "step": 142, "tokens/total": 4292032, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 64950} +{"epoch": 0.55859375, "grad_norm": 0.17141154408454895, "learning_rate": 8.777973227201069e-05, "loss": 0.004208846017718315, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00422, "step": 143, "tokens/total": 4322032, "tokens/train_per_sec_per_gpu": 38.15, "tokens/trainable": 65429} +{"epoch": 0.5625, "grad_norm": 0.1641281396150589, "learning_rate": 8.758017005316988e-05, "loss": 0.0030169994570314884, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00302, "step": 144, "tokens/total": 4352128, "tokens/train_per_sec_per_gpu": 37.17, "tokens/trainable": 65895} +{"epoch": 0.56640625, "grad_norm": 0.1339448094367981, "learning_rate": 8.737925204046629e-05, "loss": 0.0016135971527546644, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00161, "step": 145, "tokens/total": 4382368, "tokens/train_per_sec_per_gpu": 31.99, "tokens/trainable": 66327} +{"epoch": 0.5703125, "grad_norm": 0.5660213828086853, "learning_rate": 8.717698659491851e-05, "loss": 0.00862264446914196, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00866, "step": 146, "tokens/total": 4412640, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 66778} +{"epoch": 0.57421875, "grad_norm": 0.42477667331695557, "learning_rate": 8.697338213361735e-05, "loss": 0.008355698548257351, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00839, "step": 147, "tokens/total": 4442896, "tokens/train_per_sec_per_gpu": 35.67, "tokens/trainable": 67192} +{"epoch": 0.578125, "grad_norm": 0.40101227164268494, "learning_rate": 8.676844712937552e-05, "loss": 0.01053987629711628, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.0106, "step": 148, "tokens/total": 4473248, "tokens/train_per_sec_per_gpu": 35.79, "tokens/trainable": 67677} +{"epoch": 0.58203125, "grad_norm": 0.1790342479944229, "learning_rate": 8.656219011037509e-05, "loss": 0.0022937909234315157, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0023, "step": 149, "tokens/total": 4503664, "tokens/train_per_sec_per_gpu": 34.25, "tokens/trainable": 68121} +{"epoch": 0.5859375, "grad_norm": 0.49002259969711304, "learning_rate": 8.63546196598125e-05, "loss": 0.0021040919236838818, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00211, "step": 150, "tokens/total": 4533760, "tokens/train_per_sec_per_gpu": 33.44, "tokens/trainable": 68589} +{"epoch": 0.58984375, "grad_norm": 0.635802686214447, "learning_rate": 8.614574441554145e-05, "loss": 0.02125491574406624, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.02148, "step": 151, "tokens/total": 4563872, "tokens/train_per_sec_per_gpu": 30.28, "tokens/trainable": 69055} +{"epoch": 0.59375, "grad_norm": 0.8153690099716187, "learning_rate": 8.593557306971349e-05, "loss": 0.01092493161559105, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01098, "step": 152, "tokens/total": 4594384, "tokens/train_per_sec_per_gpu": 35.77, "tokens/trainable": 69501} +{"epoch": 0.59765625, "grad_norm": 0.5217874646186829, "learning_rate": 8.572411436841618e-05, "loss": 0.009749074466526508, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0098, "step": 153, "tokens/total": 4624816, "tokens/train_per_sec_per_gpu": 34.26, "tokens/trainable": 69964} +{"epoch": 0.6015625, "grad_norm": 0.2303430438041687, "learning_rate": 8.551137711130922e-05, "loss": 0.0031966331880539656, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0032, "step": 154, "tokens/total": 4655136, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 70431} +{"epoch": 0.60546875, "grad_norm": 1.3286867141723633, "learning_rate": 8.529737015125824e-05, "loss": 0.04919695854187012, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.05043, "step": 155, "tokens/total": 4683488, "tokens/train_per_sec_per_gpu": 37.67, "tokens/trainable": 70875} +{"epoch": 0.609375, "grad_norm": 0.6995537877082825, "learning_rate": 8.508210239396639e-05, "loss": 0.02087758481502533, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.0211, "step": 156, "tokens/total": 4714032, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 71333} +{"epoch": 0.61328125, "grad_norm": 0.2431785613298416, "learning_rate": 8.486558279760375e-05, "loss": 0.003255764953792095, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00326, "step": 157, "tokens/total": 4744192, "tokens/train_per_sec_per_gpu": 35.46, "tokens/trainable": 71795} +{"epoch": 0.6171875, "grad_norm": 0.5777634978294373, "learning_rate": 8.464782037243449e-05, "loss": 0.015976078808307648, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0161, "step": 158, "tokens/total": 4774592, "tokens/train_per_sec_per_gpu": 32.59, "tokens/trainable": 72231} +{"epoch": 0.62109375, "grad_norm": 0.3740072548389435, "learning_rate": 8.442882418044202e-05, "loss": 0.015244726091623306, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.01536, "step": 159, "tokens/total": 4804672, "tokens/train_per_sec_per_gpu": 30.89, "tokens/trainable": 72660} +{"epoch": 0.625, "grad_norm": 0.5423697233200073, "learning_rate": 8.420860333495179e-05, "loss": 0.02623123675584793, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.02658, "step": 160, "tokens/total": 4835056, "tokens/train_per_sec_per_gpu": 31.94, "tokens/trainable": 73115} +{"epoch": 0.62890625, "grad_norm": 0.20633140206336975, "learning_rate": 8.398716700025208e-05, "loss": 0.009905166923999786, "memory/device_reserved (GiB)": 37.89, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00995, "step": 161, "tokens/total": 4865296, "tokens/train_per_sec_per_gpu": 31.52, "tokens/trainable": 73536} +{"epoch": 0.6328125, "grad_norm": 0.2795397937297821, "learning_rate": 8.376452439121266e-05, "loss": 0.011865551583468914, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.01194, "step": 162, "tokens/total": 4895872, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 73985} +{"epoch": 0.63671875, "grad_norm": 0.2866630256175995, "learning_rate": 8.354068477290124e-05, "loss": 0.012883953750133514, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01297, "step": 163, "tokens/total": 4926192, "tokens/train_per_sec_per_gpu": 35.7, "tokens/trainable": 74449} +{"epoch": 0.640625, "grad_norm": 0.1895613968372345, "learning_rate": 8.331565746019807e-05, "loss": 0.010326865129172802, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01038, "step": 164, "tokens/total": 4956784, "tokens/train_per_sec_per_gpu": 33.78, "tokens/trainable": 74900} +{"epoch": 0.64453125, "grad_norm": 0.27947714924812317, "learning_rate": 8.308945181740812e-05, "loss": 0.008086128160357475, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00812, "step": 165, "tokens/total": 4987120, "tokens/train_per_sec_per_gpu": 30.91, "tokens/trainable": 75324} +{"epoch": 0.6484375, "grad_norm": 0.20866429805755615, "learning_rate": 8.286207725787153e-05, "loss": 0.0023722797632217407, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00238, "step": 166, "tokens/total": 5017424, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 75765} +{"epoch": 0.65234375, "grad_norm": 0.16256311535835266, "learning_rate": 8.263354324357182e-05, "loss": 0.006408158224076033, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00643, "step": 167, "tokens/total": 5048096, "tokens/train_per_sec_per_gpu": 31.6, "tokens/trainable": 76226} +{"epoch": 0.65625, "grad_norm": 0.24209903180599213, "learning_rate": 8.240385928474219e-05, "loss": 0.015991341322660446, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.01612, "step": 168, "tokens/total": 5078608, "tokens/train_per_sec_per_gpu": 33.52, "tokens/trainable": 76677} +{"epoch": 0.66015625, "grad_norm": 0.09416277706623077, "learning_rate": 8.217303493946967e-05, "loss": 0.0014367344556376338, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00144, "step": 169, "tokens/total": 5108944, "tokens/train_per_sec_per_gpu": 34.52, "tokens/trainable": 77109} +{"epoch": 0.6640625, "grad_norm": 0.35095950961112976, "learning_rate": 8.194107981329746e-05, "loss": 0.018648192286491394, "memory/device_reserved (GiB)": 36.01, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01882, "step": 170, "tokens/total": 5139344, "tokens/train_per_sec_per_gpu": 34.43, "tokens/trainable": 77558} +{"epoch": 0.66796875, "grad_norm": 0.5683367252349854, "learning_rate": 8.170800355882518e-05, "loss": 0.017304519191384315, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01746, "step": 171, "tokens/total": 5169600, "tokens/train_per_sec_per_gpu": 33.58, "tokens/trainable": 78018} +{"epoch": 0.671875, "grad_norm": 0.43827077746391296, "learning_rate": 8.147381587530713e-05, "loss": 0.00554292369633913, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00556, "step": 172, "tokens/total": 5200032, "tokens/train_per_sec_per_gpu": 30.15, "tokens/trainable": 78472} +{"epoch": 0.67578125, "grad_norm": 0.3213059902191162, "learning_rate": 8.123852650824877e-05, "loss": 0.007639693561941385, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00767, "step": 173, "tokens/total": 5230464, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 78919} +{"epoch": 0.6796875, "grad_norm": 0.237381711602211, "learning_rate": 8.100214524900103e-05, "loss": 0.005053661298006773, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.62, "memory/max_allocated (GiB)": 33.62, "ppl": 1.00507, "step": 174, "tokens/total": 5260192, "tokens/train_per_sec_per_gpu": 30.4, "tokens/trainable": 79336} +{"epoch": 0.68359375, "grad_norm": 0.11023443192243576, "learning_rate": 8.076468193435301e-05, "loss": 0.0017536291852593422, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00176, "step": 175, "tokens/total": 5290480, "tokens/train_per_sec_per_gpu": 32.91, "tokens/trainable": 79806} +{"epoch": 0.6875, "grad_norm": 0.5749457478523254, "learning_rate": 8.052614644612253e-05, "loss": 0.009171009063720703, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00921, "step": 176, "tokens/total": 5320992, "tokens/train_per_sec_per_gpu": 36.4, "tokens/trainable": 80284} +{"epoch": 0.69140625, "grad_norm": 0.09884531795978546, "learning_rate": 8.028654871074489e-05, "loss": 0.0016485131345689297, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00165, "step": 177, "tokens/total": 5351328, "tokens/train_per_sec_per_gpu": 32.5, "tokens/trainable": 80731} +{"epoch": 0.6953125, "grad_norm": 0.28975823521614075, "learning_rate": 8.004589869885986e-05, "loss": 0.0046661170199513435, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00468, "step": 178, "tokens/total": 5381536, "tokens/train_per_sec_per_gpu": 34.93, "tokens/trainable": 81199} +{"epoch": 0.69921875, "grad_norm": 0.22287319600582123, "learning_rate": 7.980420642489674e-05, "loss": 0.0017163918819278479, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00172, "step": 179, "tokens/total": 5411536, "tokens/train_per_sec_per_gpu": 33.18, "tokens/trainable": 81646} +{"epoch": 0.703125, "grad_norm": 0.015606923960149288, "learning_rate": 7.95614819466576e-05, "loss": 0.00040557183092460036, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00041, "step": 180, "tokens/total": 5441712, "tokens/train_per_sec_per_gpu": 34.69, "tokens/trainable": 82119} +{"epoch": 0.70703125, "grad_norm": 0.06141373887658119, "learning_rate": 7.931773536489872e-05, "loss": 0.00117710349150002, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00118, "step": 181, "tokens/total": 5471968, "tokens/train_per_sec_per_gpu": 39.13, "tokens/trainable": 82636} +{"epoch": 0.7109375, "grad_norm": 0.348529189825058, "learning_rate": 7.907297682291035e-05, "loss": 0.0045830621384084225, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00459, "step": 182, "tokens/total": 5502320, "tokens/train_per_sec_per_gpu": 34.4, "tokens/trainable": 83116} +{"epoch": 0.71484375, "grad_norm": 0.3752739727497101, "learning_rate": 7.882721650609442e-05, "loss": 0.00952779222279787, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00957, "step": 183, "tokens/total": 5532512, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 83563} +{"epoch": 0.71875, "grad_norm": 0.00500837154686451, "learning_rate": 7.85804646415409e-05, "loss": 0.00013146433047950268, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00013, "step": 184, "tokens/total": 5562992, "tokens/train_per_sec_per_gpu": 35.95, "tokens/trainable": 84069} +{"epoch": 0.72265625, "grad_norm": 0.04002415016293526, "learning_rate": 7.833273149760207e-05, "loss": 0.000704923877492547, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00071, "step": 185, "tokens/total": 5593200, "tokens/train_per_sec_per_gpu": 34.77, "tokens/trainable": 84544} +{"epoch": 0.7265625, "grad_norm": 0.24958014488220215, "learning_rate": 7.808402738346527e-05, "loss": 0.0023764465004205704, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00238, "step": 186, "tokens/total": 5623568, "tokens/train_per_sec_per_gpu": 33.75, "tokens/trainable": 85005} +{"epoch": 0.73046875, "grad_norm": 0.03483692184090614, "learning_rate": 7.783436264872382e-05, "loss": 0.0004840679175686091, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00048, "step": 187, "tokens/total": 5653856, "tokens/train_per_sec_per_gpu": 32.14, "tokens/trainable": 85445} +{"epoch": 0.734375, "grad_norm": 0.8145651817321777, "learning_rate": 7.758374768294647e-05, "loss": 0.024159617722034454, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.02445, "step": 188, "tokens/total": 5684112, "tokens/train_per_sec_per_gpu": 34.38, "tokens/trainable": 85935} +{"epoch": 0.73828125, "grad_norm": 0.5074039697647095, "learning_rate": 7.733219291524489e-05, "loss": 0.010505304671823978, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.01056, "step": 189, "tokens/total": 5714400, "tokens/train_per_sec_per_gpu": 36.63, "tokens/trainable": 86450} +{"epoch": 0.7421875, "grad_norm": 0.4441295862197876, "learning_rate": 7.707970881383977e-05, "loss": 0.012071937322616577, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.01215, "step": 190, "tokens/total": 5744992, "tokens/train_per_sec_per_gpu": 34.21, "tokens/trainable": 86921} +{"epoch": 0.74609375, "grad_norm": 0.25269120931625366, "learning_rate": 7.682630588562518e-05, "loss": 0.005623019300401211, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00564, "step": 191, "tokens/total": 5775568, "tokens/train_per_sec_per_gpu": 36.5, "tokens/trainable": 87376} +{"epoch": 0.75, "grad_norm": 0.40644046664237976, "learning_rate": 7.657199467573129e-05, "loss": 0.01618378609418869, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.01632, "step": 192, "tokens/total": 5806000, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 87806} +{"epoch": 0.75390625, "grad_norm": 0.2630307376384735, "learning_rate": 7.631678576708561e-05, "loss": 0.00848501268774271, "memory/device_reserved (GiB)": 36.02, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00852, "step": 193, "tokens/total": 5836432, "tokens/train_per_sec_per_gpu": 35.5, "tokens/trainable": 88289} +{"epoch": 0.7578125, "grad_norm": 0.10661368817090988, "learning_rate": 7.606068977997255e-05, "loss": 0.00248312926851213, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00249, "step": 194, "tokens/total": 5866752, "tokens/train_per_sec_per_gpu": 35.27, "tokens/trainable": 88760} +{"epoch": 0.76171875, "grad_norm": 0.15434609353542328, "learning_rate": 7.580371737159148e-05, "loss": 0.005539691541343927, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00556, "step": 195, "tokens/total": 5894976, "tokens/train_per_sec_per_gpu": 30.83, "tokens/trainable": 89193} +{"epoch": 0.765625, "grad_norm": 0.15394753217697144, "learning_rate": 7.554587923561324e-05, "loss": 0.0035556911025196314, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00356, "step": 196, "tokens/total": 5925376, "tokens/train_per_sec_per_gpu": 37.72, "tokens/trainable": 89655} +{"epoch": 0.76953125, "grad_norm": 0.2555786371231079, "learning_rate": 7.528718610173511e-05, "loss": 0.0075815883465111256, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.34, "memory/max_allocated (GiB)": 33.34, "ppl": 1.00761, "step": 197, "tokens/total": 5953696, "tokens/train_per_sec_per_gpu": 38.46, "tokens/trainable": 90103} +{"epoch": 0.7734375, "grad_norm": 0.319821834564209, "learning_rate": 7.502764873523431e-05, "loss": 0.013781344518065453, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.01388, "step": 198, "tokens/total": 5983952, "tokens/train_per_sec_per_gpu": 34.59, "tokens/trainable": 90564} +{"epoch": 0.77734375, "grad_norm": 0.25230011343955994, "learning_rate": 7.476727793652011e-05, "loss": 0.009122922085225582, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.00916, "step": 199, "tokens/total": 6014704, "tokens/train_per_sec_per_gpu": 36.87, "tokens/trainable": 91043} +{"epoch": 0.78125, "grad_norm": 0.30380022525787354, "learning_rate": 7.450608454068415e-05, "loss": 0.007986492477357388, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00802, "step": 200, "tokens/total": 6044960, "tokens/train_per_sec_per_gpu": 38.27, "tokens/trainable": 91519} +{"epoch": 0.78515625, "grad_norm": 0.07847452163696289, "learning_rate": 7.424407941704987e-05, "loss": 0.0024348655715584755, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00244, "step": 201, "tokens/total": 6074960, "tokens/train_per_sec_per_gpu": 34.43, "tokens/trainable": 91965} +{"epoch": 0.7890625, "grad_norm": 0.15240712463855743, "learning_rate": 7.398127346871986e-05, "loss": 0.005189534276723862, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0052, "step": 202, "tokens/total": 6105312, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 92430} +{"epoch": 0.79296875, "grad_norm": 0.33844277262687683, "learning_rate": 7.371767763212238e-05, "loss": 0.023467905819416046, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.02375, "step": 203, "tokens/total": 6135392, "tokens/train_per_sec_per_gpu": 37.0, "tokens/trainable": 92877} +{"epoch": 0.796875, "grad_norm": 0.19864974915981293, "learning_rate": 7.345330287655617e-05, "loss": 0.00668431306257844, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00671, "step": 204, "tokens/total": 6165936, "tokens/train_per_sec_per_gpu": 34.46, "tokens/trainable": 93350} +{"epoch": 0.80078125, "grad_norm": 0.20530343055725098, "learning_rate": 7.31881602037339e-05, "loss": 0.007711844518780708, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00774, "step": 205, "tokens/total": 6196128, "tokens/train_per_sec_per_gpu": 33.89, "tokens/trainable": 93774} +{"epoch": 0.8046875, "grad_norm": 0.6116178631782532, "learning_rate": 7.29222606473245e-05, "loss": 0.044379763305187225, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.04538, "step": 206, "tokens/total": 6226288, "tokens/train_per_sec_per_gpu": 38.11, "tokens/trainable": 94253} +{"epoch": 0.80859375, "grad_norm": 0.22179003059864044, "learning_rate": 7.265561527249383e-05, "loss": 0.009884262457489967, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00993, "step": 207, "tokens/total": 6256448, "tokens/train_per_sec_per_gpu": 35.78, "tokens/trainable": 94743} +{"epoch": 0.8125, "grad_norm": 0.13754810392856598, "learning_rate": 7.238823517544436e-05, "loss": 0.004908200353384018, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00492, "step": 208, "tokens/total": 6286736, "tokens/train_per_sec_per_gpu": 34.16, "tokens/trainable": 95228} +{"epoch": 0.81640625, "grad_norm": 0.11986561864614487, "learning_rate": 7.212013148295333e-05, "loss": 0.004731586202979088, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00474, "step": 209, "tokens/total": 6317376, "tokens/train_per_sec_per_gpu": 29.53, "tokens/trainable": 95629} +{"epoch": 0.8203125, "grad_norm": 0.18956328928470612, "learning_rate": 7.185131535190975e-05, "loss": 0.009558689780533314, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.0096, "step": 210, "tokens/total": 6347632, "tokens/train_per_sec_per_gpu": 35.79, "tokens/trainable": 96086} +{"epoch": 0.82421875, "grad_norm": 0.2573530673980713, "learning_rate": 7.158179796885005e-05, "loss": 0.013298135250806808, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.01339, "step": 211, "tokens/total": 6377616, "tokens/train_per_sec_per_gpu": 34.56, "tokens/trainable": 96507} +{"epoch": 0.828125, "grad_norm": 0.24280081689357758, "learning_rate": 7.131159054949273e-05, "loss": 0.00786699540913105, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0079, "step": 212, "tokens/total": 6408128, "tokens/train_per_sec_per_gpu": 32.92, "tokens/trainable": 96931} +{"epoch": 0.83203125, "grad_norm": 0.1858828216791153, "learning_rate": 7.104070433827139e-05, "loss": 0.007124112918972969, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00715, "step": 213, "tokens/total": 6438560, "tokens/train_per_sec_per_gpu": 33.11, "tokens/trainable": 97369} +{"epoch": 0.8359375, "grad_norm": 0.38198429346084595, "learning_rate": 7.076915060786705e-05, "loss": 0.008005295880138874, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00804, "step": 214, "tokens/total": 6468832, "tokens/train_per_sec_per_gpu": 36.86, "tokens/trainable": 97858} +{"epoch": 0.83984375, "grad_norm": 0.16806401312351227, "learning_rate": 7.049694065873882e-05, "loss": 0.004250777885317802, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00426, "step": 215, "tokens/total": 6498992, "tokens/train_per_sec_per_gpu": 37.97, "tokens/trainable": 98308} +{"epoch": 0.84375, "grad_norm": 0.011974122375249863, "learning_rate": 7.022408581865382e-05, "loss": 0.0003424361639190465, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00034, "step": 216, "tokens/total": 6529344, "tokens/train_per_sec_per_gpu": 36.89, "tokens/trainable": 98781} +{"epoch": 0.84765625, "grad_norm": 0.07292452454566956, "learning_rate": 6.99505974422157e-05, "loss": 0.0010470495326444507, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00105, "step": 217, "tokens/total": 6559696, "tokens/train_per_sec_per_gpu": 37.21, "tokens/trainable": 99266} +{"epoch": 0.8515625, "grad_norm": 0.023741189390420914, "learning_rate": 6.967648691039213e-05, "loss": 0.0004652513307519257, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00047, "step": 218, "tokens/total": 6588048, "tokens/train_per_sec_per_gpu": 39.98, "tokens/trainable": 99739} +{"epoch": 0.85546875, "grad_norm": 0.020997121930122375, "learning_rate": 6.940176563004123e-05, "loss": 0.000519782246556133, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00052, "step": 219, "tokens/total": 6618480, "tokens/train_per_sec_per_gpu": 32.08, "tokens/trainable": 100172} +{"epoch": 0.859375, "grad_norm": 0.06872506439685822, "learning_rate": 6.912644503343682e-05, "loss": 0.0009209583513438702, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00092, "step": 220, "tokens/total": 6648880, "tokens/train_per_sec_per_gpu": 37.96, "tokens/trainable": 100646} +{"epoch": 0.86328125, "grad_norm": 0.12261921912431717, "learning_rate": 6.885053657779273e-05, "loss": 0.0011635493719950318, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00116, "step": 221, "tokens/total": 6679040, "tokens/train_per_sec_per_gpu": 32.05, "tokens/trainable": 101097} +{"epoch": 0.8671875, "grad_norm": 0.21586517989635468, "learning_rate": 6.857405174478604e-05, "loss": 0.004545507952570915, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00456, "step": 222, "tokens/total": 6709552, "tokens/train_per_sec_per_gpu": 38.27, "tokens/trainable": 101563} +{"epoch": 0.87109375, "grad_norm": 0.8474878668785095, "learning_rate": 6.82970020400792e-05, "loss": 0.018102096393704414, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.01827, "step": 223, "tokens/total": 6739936, "tokens/train_per_sec_per_gpu": 31.91, "tokens/trainable": 102004} +{"epoch": 0.875, "grad_norm": 0.04830281436443329, "learning_rate": 6.801939899284132e-05, "loss": 0.0005447077564895153, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00054, "step": 224, "tokens/total": 6770144, "tokens/train_per_sec_per_gpu": 35.91, "tokens/trainable": 102452} +{"epoch": 0.87890625, "grad_norm": 0.6089279651641846, "learning_rate": 6.774125415526827e-05, "loss": 0.007109163329005241, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00713, "step": 225, "tokens/total": 6800688, "tokens/train_per_sec_per_gpu": 29.54, "tokens/trainable": 102868} +{"epoch": 0.8828125, "grad_norm": 0.3519023358821869, "learning_rate": 6.746257910210214e-05, "loss": 0.004347245674580336, "memory/device_reserved (GiB)": 35.17, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00436, "step": 226, "tokens/total": 6831376, "tokens/train_per_sec_per_gpu": 37.74, "tokens/trainable": 103375} +{"epoch": 0.88671875, "grad_norm": 0.6944298148155212, "learning_rate": 6.718338543014937e-05, "loss": 0.011605200357735157, "memory/device_reserved (GiB)": 35.17, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.01167, "step": 227, "tokens/total": 6861392, "tokens/train_per_sec_per_gpu": 39.83, "tokens/trainable": 103868} +{"epoch": 0.890625, "grad_norm": 0.061347056180238724, "learning_rate": 6.69036847577983e-05, "loss": 0.0006434498354792595, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00064, "step": 228, "tokens/total": 6891776, "tokens/train_per_sec_per_gpu": 36.72, "tokens/trainable": 104331} +{"epoch": 0.89453125, "grad_norm": 0.1004457101225853, "learning_rate": 6.662348872453553e-05, "loss": 0.0012449490604922175, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.69, "memory/max_allocated (GiB)": 33.69, "ppl": 1.00125, "step": 229, "tokens/total": 6921776, "tokens/train_per_sec_per_gpu": 33.87, "tokens/trainable": 104796} +{"epoch": 0.8984375, "grad_norm": 0.47870421409606934, "learning_rate": 6.63428089904618e-05, "loss": 0.012319323606789112, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.0124, "step": 230, "tokens/total": 6952048, "tokens/train_per_sec_per_gpu": 37.27, "tokens/trainable": 105264} +{"epoch": 0.90234375, "grad_norm": 0.4258827865123749, "learning_rate": 6.60616572358065e-05, "loss": 0.00830936897546053, "memory/device_reserved (GiB)": 35.86, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00834, "step": 231, "tokens/total": 6982384, "tokens/train_per_sec_per_gpu": 34.25, "tokens/trainable": 105727} +{"epoch": 0.90625, "grad_norm": 0.07725202292203903, "learning_rate": 6.578004516044172e-05, "loss": 0.0011023205006495118, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0011, "step": 232, "tokens/total": 7012784, "tokens/train_per_sec_per_gpu": 33.45, "tokens/trainable": 106203} +{"epoch": 0.91015625, "grad_norm": 0.07384540885686874, "learning_rate": 6.549798448339548e-05, "loss": 0.0012245059479027987, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00123, "step": 233, "tokens/total": 7043184, "tokens/train_per_sec_per_gpu": 38.46, "tokens/trainable": 106708} +{"epoch": 0.9140625, "grad_norm": 0.3441248834133148, "learning_rate": 6.521548694236384e-05, "loss": 0.006349029019474983, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00637, "step": 234, "tokens/total": 7073168, "tokens/train_per_sec_per_gpu": 34.19, "tokens/trainable": 107146} +{"epoch": 0.91796875, "grad_norm": 0.019105862826108932, "learning_rate": 6.493256429322259e-05, "loss": 0.0003932910331059247, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00039, "step": 235, "tokens/total": 7103616, "tokens/train_per_sec_per_gpu": 33.1, "tokens/trainable": 107586} +{"epoch": 0.921875, "grad_norm": 0.1013171374797821, "learning_rate": 6.464922830953799e-05, "loss": 0.0018943212926387787, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.0019, "step": 236, "tokens/total": 7133840, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 108051} +{"epoch": 0.92578125, "grad_norm": 0.07892157137393951, "learning_rate": 6.436549078207688e-05, "loss": 0.0005698923487216234, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00057, "step": 237, "tokens/total": 7164016, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 108479} +{"epoch": 0.9296875, "grad_norm": 0.45241886377334595, "learning_rate": 6.408136351831592e-05, "loss": 0.0160800963640213, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.01621, "step": 238, "tokens/total": 7194400, "tokens/train_per_sec_per_gpu": 29.52, "tokens/trainable": 108907} +{"epoch": 0.93359375, "grad_norm": 0.23932066559791565, "learning_rate": 6.379685834195036e-05, "loss": 0.009671966545283794, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00972, "step": 239, "tokens/total": 7224992, "tokens/train_per_sec_per_gpu": 32.81, "tokens/trainable": 109348} +{"epoch": 0.9375, "grad_norm": 0.04965788871049881, "learning_rate": 6.351198709240186e-05, "loss": 0.0012405012967064977, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00124, "step": 240, "tokens/total": 7255424, "tokens/train_per_sec_per_gpu": 32.71, "tokens/trainable": 109812} +{"epoch": 0.94140625, "grad_norm": 0.29825523495674133, "learning_rate": 6.32267616243259e-05, "loss": 0.005660452414304018, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00568, "step": 241, "tokens/total": 7285696, "tokens/train_per_sec_per_gpu": 38.88, "tokens/trainable": 110270} +{"epoch": 0.9453125, "grad_norm": 0.26985254883766174, "learning_rate": 6.294119380711849e-05, "loss": 0.0047082495875656605, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00472, "step": 242, "tokens/total": 7316000, "tokens/train_per_sec_per_gpu": 28.25, "tokens/trainable": 110682} +{"epoch": 0.94921875, "grad_norm": 0.43075069785118103, "learning_rate": 6.265529552442209e-05, "loss": 0.003754058852791786, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00376, "step": 243, "tokens/total": 7346304, "tokens/train_per_sec_per_gpu": 39.54, "tokens/trainable": 111176} +{"epoch": 0.953125, "grad_norm": 0.27702149748802185, "learning_rate": 6.236907867363127e-05, "loss": 0.008292095735669136, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00833, "step": 244, "tokens/total": 7376944, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 111633} +{"epoch": 0.95703125, "grad_norm": 0.07920246571302414, "learning_rate": 6.208255516539749e-05, "loss": 0.0018851464847102761, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00189, "step": 245, "tokens/total": 7407184, "tokens/train_per_sec_per_gpu": 34.65, "tokens/trainable": 112115} +{"epoch": 0.9609375, "grad_norm": 0.306937575340271, "learning_rate": 6.179573692313344e-05, "loss": 0.008207850158214569, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00824, "step": 246, "tokens/total": 7437760, "tokens/train_per_sec_per_gpu": 33.59, "tokens/trainable": 112589} +{"epoch": 0.96484375, "grad_norm": 0.22587862610816956, "learning_rate": 6.150863588251694e-05, "loss": 0.004535369109362364, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00455, "step": 247, "tokens/total": 7466080, "tokens/train_per_sec_per_gpu": 38.2, "tokens/trainable": 113043} +{"epoch": 0.96875, "grad_norm": 0.1514762043952942, "learning_rate": 6.122126399099419e-05, "loss": 0.0027371919713914394, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00274, "step": 248, "tokens/total": 7496672, "tokens/train_per_sec_per_gpu": 35.66, "tokens/trainable": 113484} +{"epoch": 0.97265625, "grad_norm": 0.09306064248085022, "learning_rate": 6.0933633207282615e-05, "loss": 0.002224899362772703, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00223, "step": 249, "tokens/total": 7526816, "tokens/train_per_sec_per_gpu": 33.02, "tokens/trainable": 113928} +{"epoch": 0.9765625, "grad_norm": 0.06742242723703384, "learning_rate": 6.064575550087316e-05, "loss": 0.0009518311126157641, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00095, "step": 250, "tokens/total": 7555120, "tokens/train_per_sec_per_gpu": 34.24, "tokens/trainable": 114360} +{"epoch": 0.98046875, "grad_norm": 0.09547543525695801, "learning_rate": 6.0357642851532245e-05, "loss": 0.0044240644201636314, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00443, "step": 251, "tokens/total": 7585632, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 114850} +{"epoch": 0.984375, "grad_norm": 0.35386043787002563, "learning_rate": 6.0069307248803294e-05, "loss": 0.009106422774493694, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00915, "step": 252, "tokens/total": 7615888, "tokens/train_per_sec_per_gpu": 35.56, "tokens/trainable": 115298} +{"epoch": 0.98828125, "grad_norm": 0.047934144735336304, "learning_rate": 5.9780760691507635e-05, "loss": 0.0007137035136111081, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00071, "step": 253, "tokens/total": 7646464, "tokens/train_per_sec_per_gpu": 32.61, "tokens/trainable": 115719} +{"epoch": 0.9921875, "grad_norm": 0.2060946822166443, "learning_rate": 5.9492015187245334e-05, "loss": 0.0057786800898611546, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0058, "step": 254, "tokens/total": 7676896, "tokens/train_per_sec_per_gpu": 29.56, "tokens/trainable": 116151} +{"epoch": 0.99609375, "grad_norm": 0.02973618358373642, "learning_rate": 5.920308275189541e-05, "loss": 0.000541117915417999, "memory/device_reserved (GiB)": 35.94, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00054, "step": 255, "tokens/total": 7707328, "tokens/train_per_sec_per_gpu": 37.19, "tokens/trainable": 116602} +{"epoch": 1.0, "grad_norm": 0.026812957599759102, "learning_rate": 5.8913975409115874e-05, "loss": 0.0003775915829464793, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00038, "step": 256, "tokens/total": 7737920, "tokens/train_per_sec_per_gpu": 32.45, "tokens/trainable": 117050} +{"epoch": 1.00390625, "grad_norm": 0.13118499517440796, "learning_rate": 5.8624705189843395e-05, "loss": 0.0015422197757288814, "memory/device_reserved (GiB)": 36.21, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00154, "step": 257, "tokens/total": 7768256, "tokens/train_per_sec_per_gpu": 32.54, "tokens/trainable": 117506} +{"epoch": 1.0078125, "grad_norm": 0.007149726152420044, "learning_rate": 5.833528413179249e-05, "loss": 0.0001414301514159888, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00014, "step": 258, "tokens/total": 7798544, "tokens/train_per_sec_per_gpu": 30.82, "tokens/trainable": 117936} +{"epoch": 1.01171875, "grad_norm": 0.25255724787712097, "learning_rate": 5.80457242789548e-05, "loss": 0.0026415749453008175, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00265, "step": 259, "tokens/total": 7828880, "tokens/train_per_sec_per_gpu": 31.96, "tokens/trainable": 118377} +{"epoch": 1.015625, "grad_norm": 0.0023047819267958403, "learning_rate": 5.77560376810977e-05, "loss": 4.6871849917806685e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00005, "step": 260, "tokens/total": 7859312, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 118820} +{"epoch": 1.01953125, "grad_norm": 0.1961309015750885, "learning_rate": 5.7466236393263005e-05, "loss": 0.0034356648102402687, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00344, "step": 261, "tokens/total": 7889776, "tokens/train_per_sec_per_gpu": 34.37, "tokens/trainable": 119267} +{"epoch": 1.0234375, "grad_norm": 0.06521005183458328, "learning_rate": 5.717633247526522e-05, "loss": 0.0006203935481607914, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00062, "step": 262, "tokens/total": 7920256, "tokens/train_per_sec_per_gpu": 35.19, "tokens/trainable": 119776} +{"epoch": 1.02734375, "grad_norm": 0.2380313128232956, "learning_rate": 5.688633799118971e-05, "loss": 0.003634552238509059, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00364, "step": 263, "tokens/total": 7950608, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 120265} +{"epoch": 1.03125, "grad_norm": 0.22213542461395264, "learning_rate": 5.659626500889066e-05, "loss": 0.0028508887626230717, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00285, "step": 264, "tokens/total": 7981168, "tokens/train_per_sec_per_gpu": 37.54, "tokens/trainable": 120742} +{"epoch": 1.03515625, "grad_norm": 0.007840439677238464, "learning_rate": 5.6306125599488905e-05, "loss": 0.00010933953308267519, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00011, "step": 265, "tokens/total": 8011504, "tokens/train_per_sec_per_gpu": 40.03, "tokens/trainable": 121209} +{"epoch": 1.0390625, "grad_norm": 0.06023910269141197, "learning_rate": 5.601593183686955e-05, "loss": 0.0008161423029378057, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00082, "step": 266, "tokens/total": 8042048, "tokens/train_per_sec_per_gpu": 32.67, "tokens/trainable": 121683} +{"epoch": 1.04296875, "grad_norm": 0.014270068146288395, "learning_rate": 5.572569579717961e-05, "loss": 0.00010447163367643952, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0001, "step": 267, "tokens/total": 8072336, "tokens/train_per_sec_per_gpu": 40.53, "tokens/trainable": 122187} +{"epoch": 1.046875, "grad_norm": 0.036489423364400864, "learning_rate": 5.543542955832538e-05, "loss": 0.0003805297310464084, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00038, "step": 268, "tokens/total": 8102592, "tokens/train_per_sec_per_gpu": 35.89, "tokens/trainable": 122649} +{"epoch": 1.05078125, "grad_norm": 0.04869162663817406, "learning_rate": 5.514514519946986e-05, "loss": 0.0005166378105059266, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00052, "step": 269, "tokens/total": 8132752, "tokens/train_per_sec_per_gpu": 33.52, "tokens/trainable": 123109} +{"epoch": 1.0546875, "grad_norm": 0.0886458307504654, "learning_rate": 5.485485480053015e-05, "loss": 0.0009586921078152955, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00096, "step": 270, "tokens/total": 8163232, "tokens/train_per_sec_per_gpu": 36.69, "tokens/trainable": 123599} +{"epoch": 1.05859375, "grad_norm": 0.07190798223018646, "learning_rate": 5.4564570441674645e-05, "loss": 0.0005164776812307537, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00052, "step": 271, "tokens/total": 8193344, "tokens/train_per_sec_per_gpu": 38.15, "tokens/trainable": 124081} +{"epoch": 1.0625, "grad_norm": 0.0680084079504013, "learning_rate": 5.42743042028204e-05, "loss": 0.00043266740976832807, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00043, "step": 272, "tokens/total": 8223632, "tokens/train_per_sec_per_gpu": 35.19, "tokens/trainable": 124551} +{"epoch": 1.06640625, "grad_norm": 0.008596232160925865, "learning_rate": 5.3984068163130464e-05, "loss": 5.5010747018968686e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00006, "step": 273, "tokens/total": 8253680, "tokens/train_per_sec_per_gpu": 34.47, "tokens/trainable": 124981} +{"epoch": 1.0703125, "grad_norm": 0.12593501806259155, "learning_rate": 5.369387440051111e-05, "loss": 0.0006858849665150046, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00069, "step": 274, "tokens/total": 8284256, "tokens/train_per_sec_per_gpu": 35.0, "tokens/trainable": 125442} +{"epoch": 1.07421875, "grad_norm": 0.002708524465560913, "learning_rate": 5.340373499110935e-05, "loss": 2.155889706045855e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00002, "step": 275, "tokens/total": 8314896, "tokens/train_per_sec_per_gpu": 31.81, "tokens/trainable": 125893} +{"epoch": 1.078125, "grad_norm": 0.6500045657157898, "learning_rate": 5.3113662008810304e-05, "loss": 0.0032175867818295956, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00322, "step": 276, "tokens/total": 8345264, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 126367} +{"epoch": 1.08203125, "grad_norm": 0.0035925847478210926, "learning_rate": 5.282366752473479e-05, "loss": 4.369133966974914e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 277, "tokens/total": 8375728, "tokens/train_per_sec_per_gpu": 34.39, "tokens/trainable": 126832} +{"epoch": 1.0859375, "grad_norm": 0.06011431664228439, "learning_rate": 5.2533763606737005e-05, "loss": 0.00032412572181783617, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00032, "step": 278, "tokens/total": 8405952, "tokens/train_per_sec_per_gpu": 38.74, "tokens/trainable": 127335} +{"epoch": 1.08984375, "grad_norm": 0.0041359406895935535, "learning_rate": 5.224396231890232e-05, "loss": 4.0772643842501566e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00004, "step": 279, "tokens/total": 8436096, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 127793} +{"epoch": 1.09375, "grad_norm": 0.0019470915431156754, "learning_rate": 5.195427572104522e-05, "loss": 2.092921204166487e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00002, "step": 280, "tokens/total": 8466464, "tokens/train_per_sec_per_gpu": 33.21, "tokens/trainable": 128242} +{"epoch": 1.09765625, "grad_norm": 0.0017560214037075639, "learning_rate": 5.166471586820751e-05, "loss": 2.1481781004695222e-05, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 281, "tokens/total": 8496736, "tokens/train_per_sec_per_gpu": 36.63, "tokens/trainable": 128724} +{"epoch": 1.1015625, "grad_norm": 0.01728290133178234, "learning_rate": 5.1375294810156615e-05, "loss": 0.00014254249981604517, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00014, "step": 282, "tokens/total": 8526928, "tokens/train_per_sec_per_gpu": 35.5, "tokens/trainable": 129198} +{"epoch": 1.10546875, "grad_norm": 0.014790019951760769, "learning_rate": 5.1086024590884144e-05, "loss": 6.786123412894085e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00007, "step": 283, "tokens/total": 8557296, "tokens/train_per_sec_per_gpu": 33.38, "tokens/trainable": 129648} +{"epoch": 1.109375, "grad_norm": 0.4627615809440613, "learning_rate": 5.079691724810461e-05, "loss": 0.004037827253341675, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00405, "step": 284, "tokens/total": 8587696, "tokens/train_per_sec_per_gpu": 32.72, "tokens/trainable": 130095} +{"epoch": 1.11328125, "grad_norm": 0.3512204587459564, "learning_rate": 5.0507984812754684e-05, "loss": 0.002691472414880991, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0027, "step": 285, "tokens/total": 8618080, "tokens/train_per_sec_per_gpu": 31.74, "tokens/trainable": 130545} +{"epoch": 1.1171875, "grad_norm": 0.009706776589155197, "learning_rate": 5.021923930849237e-05, "loss": 8.887588774086908e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00009, "step": 286, "tokens/total": 8648288, "tokens/train_per_sec_per_gpu": 34.64, "tokens/trainable": 131002} +{"epoch": 1.12109375, "grad_norm": 0.07987845689058304, "learning_rate": 4.99306927511967e-05, "loss": 0.0007880277116782963, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00079, "step": 287, "tokens/total": 8678464, "tokens/train_per_sec_per_gpu": 34.31, "tokens/trainable": 131458} +{"epoch": 1.125, "grad_norm": 0.0013652736088261008, "learning_rate": 4.964235714846775e-05, "loss": 1.4229615771910176e-05, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00001, "step": 288, "tokens/total": 8708816, "tokens/train_per_sec_per_gpu": 36.32, "tokens/trainable": 131932} +{"epoch": 1.12890625, "grad_norm": 0.00022536546748597175, "learning_rate": 4.9354244499126866e-05, "loss": 5.348314971342916e-06, "memory/device_reserved (GiB)": 35.91, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00001, "step": 289, "tokens/total": 8738992, "tokens/train_per_sec_per_gpu": 33.57, "tokens/trainable": 132406} +{"epoch": 1.1328125, "grad_norm": 0.00034249111195094883, "learning_rate": 4.90663667927174e-05, "loss": 6.50786478217924e-06, "memory/device_reserved (GiB)": 35.57, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00001, "step": 290, "tokens/total": 8769424, "tokens/train_per_sec_per_gpu": 36.83, "tokens/trainable": 132855} +{"epoch": 1.13671875, "grad_norm": 0.0011629040818661451, "learning_rate": 4.877873600900581e-05, "loss": 2.100312485708855e-05, "memory/device_reserved (GiB)": 35.57, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00002, "step": 291, "tokens/total": 8799680, "tokens/train_per_sec_per_gpu": 36.18, "tokens/trainable": 133356} +{"epoch": 1.140625, "grad_norm": 0.0002511514176148921, "learning_rate": 4.849136411748306e-05, "loss": 5.5127738960436545e-06, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00001, "step": 292, "tokens/total": 8830000, "tokens/train_per_sec_per_gpu": 34.54, "tokens/trainable": 133788} +{"epoch": 1.14453125, "grad_norm": 0.005342632066458464, "learning_rate": 4.8204263076866574e-05, "loss": 5.693141429219395e-05, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00006, "step": 293, "tokens/total": 8860320, "tokens/train_per_sec_per_gpu": 36.67, "tokens/trainable": 134260} +{"epoch": 1.1484375, "grad_norm": 0.015129564329981804, "learning_rate": 4.791744483460251e-05, "loss": 5.992315709590912e-05, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00006, "step": 294, "tokens/total": 8891072, "tokens/train_per_sec_per_gpu": 34.01, "tokens/trainable": 134684} +{"epoch": 1.15234375, "grad_norm": 0.2627723515033722, "learning_rate": 4.7630921326368736e-05, "loss": 0.0013240529224276543, "memory/device_reserved (GiB)": 35.59, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00132, "step": 295, "tokens/total": 8921504, "tokens/train_per_sec_per_gpu": 36.05, "tokens/trainable": 135166} +{"epoch": 1.15625, "grad_norm": 0.00017540185945108533, "learning_rate": 4.7344704475577916e-05, "loss": 4.621510925062466e-06, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.0, "step": 296, "tokens/total": 8951808, "tokens/train_per_sec_per_gpu": 33.62, "tokens/trainable": 135600} +{"epoch": 1.16015625, "grad_norm": 0.34399038553237915, "learning_rate": 4.705880619288153e-05, "loss": 0.007753564510494471, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00778, "step": 297, "tokens/total": 8982080, "tokens/train_per_sec_per_gpu": 32.99, "tokens/trainable": 136039} +{"epoch": 1.1640625, "grad_norm": 0.010148555971682072, "learning_rate": 4.677323837567412e-05, "loss": 5.8421592257218435e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00006, "step": 298, "tokens/total": 9012560, "tokens/train_per_sec_per_gpu": 33.31, "tokens/trainable": 136496} +{"epoch": 1.16796875, "grad_norm": 0.0038862715009599924, "learning_rate": 4.6488012907598146e-05, "loss": 3.437784835114144e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 299, "tokens/total": 9042944, "tokens/train_per_sec_per_gpu": 35.75, "tokens/trainable": 136954} +{"epoch": 1.171875, "grad_norm": 0.0015254979953169823, "learning_rate": 4.620314165804964e-05, "loss": 2.051112096523866e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00002, "step": 300, "tokens/total": 9073248, "tokens/train_per_sec_per_gpu": 34.64, "tokens/trainable": 137404} +{"epoch": 1.17578125, "grad_norm": 0.4777839481830597, "learning_rate": 4.591863648168407e-05, "loss": 0.006096641067415476, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00612, "step": 301, "tokens/total": 9103728, "tokens/train_per_sec_per_gpu": 36.12, "tokens/trainable": 137899} +{"epoch": 1.1796875, "grad_norm": 0.09574320912361145, "learning_rate": 4.5634509217923135e-05, "loss": 0.00047178880777209997, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00047, "step": 302, "tokens/total": 9134000, "tokens/train_per_sec_per_gpu": 34.52, "tokens/trainable": 138375} +{"epoch": 1.18359375, "grad_norm": 0.003620963543653488, "learning_rate": 4.535077169046201e-05, "loss": 4.1847168176900595e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 303, "tokens/total": 9164400, "tokens/train_per_sec_per_gpu": 36.82, "tokens/trainable": 138834} +{"epoch": 1.1875, "grad_norm": 0.004095857497304678, "learning_rate": 4.506743570677743e-05, "loss": 3.6887995520373806e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00004, "step": 304, "tokens/total": 9194672, "tokens/train_per_sec_per_gpu": 30.99, "tokens/trainable": 139259} +{"epoch": 1.19140625, "grad_norm": 0.004100849851965904, "learning_rate": 4.478451305763618e-05, "loss": 4.464022640604526e-05, "memory/device_reserved (GiB)": 35.67, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00004, "step": 305, "tokens/total": 9224848, "tokens/train_per_sec_per_gpu": 33.42, "tokens/trainable": 139717} +{"epoch": 1.1953125, "grad_norm": 0.22271950542926788, "learning_rate": 4.450201551660454e-05, "loss": 0.002607254544273019, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00261, "step": 306, "tokens/total": 9255440, "tokens/train_per_sec_per_gpu": 37.37, "tokens/trainable": 140207} +{"epoch": 1.19921875, "grad_norm": 0.05722619220614433, "learning_rate": 4.4219954839558276e-05, "loss": 0.00047086336417123675, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.19, "memory/max_allocated (GiB)": 33.19, "ppl": 1.00047, "step": 307, "tokens/total": 9283344, "tokens/train_per_sec_per_gpu": 34.35, "tokens/trainable": 140645} +{"epoch": 1.203125, "grad_norm": 0.6939759850502014, "learning_rate": 4.393834276419352e-05, "loss": 0.004533157218247652, "memory/device_reserved (GiB)": 35.9, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00454, "step": 308, "tokens/total": 9313856, "tokens/train_per_sec_per_gpu": 30.34, "tokens/trainable": 141096} +{"epoch": 1.20703125, "grad_norm": 0.003842511447146535, "learning_rate": 4.36571910095382e-05, "loss": 4.202376658213325e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00004, "step": 309, "tokens/total": 9344064, "tokens/train_per_sec_per_gpu": 29.26, "tokens/trainable": 141518} +{"epoch": 1.2109375, "grad_norm": 0.002201825613155961, "learning_rate": 4.337651127546448e-05, "loss": 4.229221667628735e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00004, "step": 310, "tokens/total": 9374368, "tokens/train_per_sec_per_gpu": 35.84, "tokens/trainable": 141952} +{"epoch": 1.21484375, "grad_norm": 0.0005678636953234673, "learning_rate": 4.3096315242201736e-05, "loss": 1.2881277143605985e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00001, "step": 311, "tokens/total": 9404624, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 142416} +{"epoch": 1.21875, "grad_norm": 0.0005692228442057967, "learning_rate": 4.2816614569850635e-05, "loss": 1.2092226825188845e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00001, "step": 312, "tokens/total": 9434784, "tokens/train_per_sec_per_gpu": 38.63, "tokens/trainable": 142882} +{"epoch": 1.22265625, "grad_norm": 0.00879319105297327, "learning_rate": 4.2537420897897864e-05, "loss": 9.982152550946921e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.0001, "step": 313, "tokens/total": 9465168, "tokens/train_per_sec_per_gpu": 36.84, "tokens/trainable": 143356} +{"epoch": 1.2265625, "grad_norm": 0.008733258582651615, "learning_rate": 4.225874584473174e-05, "loss": 0.00010937166371149942, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00011, "step": 314, "tokens/total": 9495488, "tokens/train_per_sec_per_gpu": 33.31, "tokens/trainable": 143812} +{"epoch": 1.23046875, "grad_norm": 0.016755763441324234, "learning_rate": 4.19806010071587e-05, "loss": 0.00017517709056846797, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00018, "step": 315, "tokens/total": 9525968, "tokens/train_per_sec_per_gpu": 34.95, "tokens/trainable": 144294} +{"epoch": 1.234375, "grad_norm": 0.01962393708527088, "learning_rate": 4.170299795992081e-05, "loss": 0.0001436145685147494, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00014, "step": 316, "tokens/total": 9556368, "tokens/train_per_sec_per_gpu": 35.87, "tokens/trainable": 144751} +{"epoch": 1.23828125, "grad_norm": 0.0018895679386332631, "learning_rate": 4.142594825521398e-05, "loss": 3.0470149795291945e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00003, "step": 317, "tokens/total": 9586800, "tokens/train_per_sec_per_gpu": 37.67, "tokens/trainable": 145251} +{"epoch": 1.2421875, "grad_norm": 0.030299803242087364, "learning_rate": 4.114946342220728e-05, "loss": 0.0003194157325197011, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00032, "step": 318, "tokens/total": 9617152, "tokens/train_per_sec_per_gpu": 37.13, "tokens/trainable": 145711} +{"epoch": 1.24609375, "grad_norm": 0.04471183568239212, "learning_rate": 4.087355496656321e-05, "loss": 0.0005480494000948966, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00055, "step": 319, "tokens/total": 9647424, "tokens/train_per_sec_per_gpu": 37.59, "tokens/trainable": 146205} +{"epoch": 1.25, "grad_norm": 1.604000210762024, "learning_rate": 4.05982343699588e-05, "loss": 0.003160722553730011, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00317, "step": 320, "tokens/total": 9677776, "tokens/train_per_sec_per_gpu": 34.32, "tokens/trainable": 146678} +{"epoch": 1.25390625, "grad_norm": 0.0033273256849497557, "learning_rate": 4.0323513089607876e-05, "loss": 3.962405025959015e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00004, "step": 321, "tokens/total": 9707936, "tokens/train_per_sec_per_gpu": 35.71, "tokens/trainable": 147124} +{"epoch": 1.2578125, "grad_norm": 0.005383767187595367, "learning_rate": 4.004940255778431e-05, "loss": 7.032878784229979e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00007, "step": 322, "tokens/total": 9738448, "tokens/train_per_sec_per_gpu": 35.58, "tokens/trainable": 147574} +{"epoch": 1.26171875, "grad_norm": 0.0028380267322063446, "learning_rate": 3.977591418134619e-05, "loss": 5.103685543872416e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00005, "step": 323, "tokens/total": 9768864, "tokens/train_per_sec_per_gpu": 35.65, "tokens/trainable": 148030} +{"epoch": 1.265625, "grad_norm": 0.025233343243598938, "learning_rate": 3.95030593412612e-05, "loss": 0.00011621347948675975, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00012, "step": 324, "tokens/total": 9799184, "tokens/train_per_sec_per_gpu": 37.22, "tokens/trainable": 148478} +{"epoch": 1.26953125, "grad_norm": 0.04950147494673729, "learning_rate": 3.923084939213296e-05, "loss": 0.0003831804497167468, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00038, "step": 325, "tokens/total": 9829392, "tokens/train_per_sec_per_gpu": 36.95, "tokens/trainable": 148941} +{"epoch": 1.2734375, "grad_norm": 0.0017629251815378666, "learning_rate": 3.895929566172861e-05, "loss": 2.4212906282627955e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 326, "tokens/total": 9859696, "tokens/train_per_sec_per_gpu": 38.19, "tokens/trainable": 149435} +{"epoch": 1.27734375, "grad_norm": 0.025092069059610367, "learning_rate": 3.868840945050728e-05, "loss": 0.0002421572571620345, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00024, "step": 327, "tokens/total": 9889648, "tokens/train_per_sec_per_gpu": 39.82, "tokens/trainable": 149888} +{"epoch": 1.28125, "grad_norm": 0.04499012976884842, "learning_rate": 3.841820203114995e-05, "loss": 0.0005073954816907644, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00051, "step": 328, "tokens/total": 9919968, "tokens/train_per_sec_per_gpu": 34.24, "tokens/trainable": 150362} +{"epoch": 1.28515625, "grad_norm": 0.016609711572527885, "learning_rate": 3.814868464809027e-05, "loss": 0.00013441528426483274, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00013, "step": 329, "tokens/total": 9950368, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 150837} +{"epoch": 1.2890625, "grad_norm": 0.003686394076794386, "learning_rate": 3.787986851704667e-05, "loss": 4.67524841951672e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00005, "step": 330, "tokens/total": 9980688, "tokens/train_per_sec_per_gpu": 34.36, "tokens/trainable": 151273} +{"epoch": 1.29296875, "grad_norm": 0.0016950524877756834, "learning_rate": 3.7611764824555654e-05, "loss": 2.6381265342934057e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00003, "step": 331, "tokens/total": 10011104, "tokens/train_per_sec_per_gpu": 37.63, "tokens/trainable": 151741} +{"epoch": 1.296875, "grad_norm": 0.004727163352072239, "learning_rate": 3.734438472750619e-05, "loss": 4.7812289267312735e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00005, "step": 332, "tokens/total": 10041536, "tokens/train_per_sec_per_gpu": 34.47, "tokens/trainable": 152191} +{"epoch": 1.30078125, "grad_norm": 0.09555032104253769, "learning_rate": 3.707773935267552e-05, "loss": 0.0008173320675268769, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00082, "step": 333, "tokens/total": 10072112, "tokens/train_per_sec_per_gpu": 32.24, "tokens/trainable": 152630} +{"epoch": 1.3046875, "grad_norm": 0.02793118543922901, "learning_rate": 3.68118397962661e-05, "loss": 0.0002140044525731355, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00021, "step": 334, "tokens/total": 10102592, "tokens/train_per_sec_per_gpu": 33.85, "tokens/trainable": 153047} +{"epoch": 1.30859375, "grad_norm": 0.2730892598628998, "learning_rate": 3.654669712344384e-05, "loss": 0.004334757570177317, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00434, "step": 335, "tokens/total": 10132736, "tokens/train_per_sec_per_gpu": 30.89, "tokens/trainable": 153477} +{"epoch": 1.3125, "grad_norm": 0.002235857769846916, "learning_rate": 3.628232236787763e-05, "loss": 9.928971849149093e-06, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00001, "step": 336, "tokens/total": 10162928, "tokens/train_per_sec_per_gpu": 33.72, "tokens/trainable": 153929} +{"epoch": 1.31640625, "grad_norm": 0.0533645860850811, "learning_rate": 3.6018726531280144e-05, "loss": 0.0006771332118660212, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00068, "step": 337, "tokens/total": 10193552, "tokens/train_per_sec_per_gpu": 35.34, "tokens/trainable": 154416} +{"epoch": 1.3203125, "grad_norm": 0.003405389143154025, "learning_rate": 3.575592058295017e-05, "loss": 4.594672645907849e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 338, "tokens/total": 10223808, "tokens/train_per_sec_per_gpu": 34.33, "tokens/trainable": 154865} +{"epoch": 1.32421875, "grad_norm": 0.0007159400847740471, "learning_rate": 3.549391545931585e-05, "loss": 1.3502348338079173e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00001, "step": 339, "tokens/total": 10254272, "tokens/train_per_sec_per_gpu": 30.21, "tokens/trainable": 155290} +{"epoch": 1.328125, "grad_norm": 0.013892000541090965, "learning_rate": 3.5232722063479914e-05, "loss": 7.453892612829804e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00007, "step": 340, "tokens/total": 10284656, "tokens/train_per_sec_per_gpu": 31.6, "tokens/trainable": 155703} +{"epoch": 1.33203125, "grad_norm": 0.005680185277014971, "learning_rate": 3.49723512647657e-05, "loss": 7.792153337504715e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00008, "step": 341, "tokens/total": 10314976, "tokens/train_per_sec_per_gpu": 36.22, "tokens/trainable": 156193} +{"epoch": 1.3359375, "grad_norm": 0.015012380667030811, "learning_rate": 3.471281389826491e-05, "loss": 0.0001705413160379976, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00017, "step": 342, "tokens/total": 10345360, "tokens/train_per_sec_per_gpu": 34.09, "tokens/trainable": 156640} +{"epoch": 1.33984375, "grad_norm": 0.0032200925052165985, "learning_rate": 3.4454120764386764e-05, "loss": 2.8803326131310314e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00003, "step": 343, "tokens/total": 10375648, "tokens/train_per_sec_per_gpu": 34.14, "tokens/trainable": 157102} +{"epoch": 1.34375, "grad_norm": 0.007129024248570204, "learning_rate": 3.4196282628408526e-05, "loss": 5.837350181536749e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00006, "step": 344, "tokens/total": 10406144, "tokens/train_per_sec_per_gpu": 33.51, "tokens/trainable": 157544} +{"epoch": 1.34765625, "grad_norm": 0.30264967679977417, "learning_rate": 3.3939310220027456e-05, "loss": 0.007746525574475527, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00778, "step": 345, "tokens/total": 10436528, "tokens/train_per_sec_per_gpu": 35.6, "tokens/trainable": 157985} +{"epoch": 1.3515625, "grad_norm": 0.014983746223151684, "learning_rate": 3.3683214232914404e-05, "loss": 0.0001249729102710262, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00012, "step": 346, "tokens/total": 10466864, "tokens/train_per_sec_per_gpu": 29.48, "tokens/trainable": 158402} +{"epoch": 1.35546875, "grad_norm": 0.0023245092015713453, "learning_rate": 3.342800532426873e-05, "loss": 2.6811658244696446e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00003, "step": 347, "tokens/total": 10497008, "tokens/train_per_sec_per_gpu": 36.08, "tokens/trainable": 158860} +{"epoch": 1.359375, "grad_norm": 0.01576041616499424, "learning_rate": 3.317369411437484e-05, "loss": 0.00016933181905187666, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.0, "memory/max_allocated (GiB)": 34.0, "ppl": 1.00017, "step": 348, "tokens/total": 10527728, "tokens/train_per_sec_per_gpu": 29.09, "tokens/trainable": 159282} +{"epoch": 1.36328125, "grad_norm": 0.20484071969985962, "learning_rate": 3.292029118616024e-05, "loss": 0.0017917631193995476, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00179, "step": 349, "tokens/total": 10557856, "tokens/train_per_sec_per_gpu": 32.95, "tokens/trainable": 159732} +{"epoch": 1.3671875, "grad_norm": 0.0009161452180705965, "learning_rate": 3.266780708475511e-05, "loss": 1.5678642739658244e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.99, "memory/max_allocated (GiB)": 33.99, "ppl": 1.00002, "step": 350, "tokens/total": 10588368, "tokens/train_per_sec_per_gpu": 35.68, "tokens/trainable": 160199} +{"epoch": 1.37109375, "grad_norm": 0.0552787259221077, "learning_rate": 3.241625231705354e-05, "loss": 0.0003493681433610618, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00035, "step": 351, "tokens/total": 10618608, "tokens/train_per_sec_per_gpu": 33.04, "tokens/trainable": 160657} +{"epoch": 1.375, "grad_norm": 0.01718018762767315, "learning_rate": 3.216563735127618e-05, "loss": 0.00022284504666458815, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00022, "step": 352, "tokens/total": 10648864, "tokens/train_per_sec_per_gpu": 34.98, "tokens/trainable": 161114} +{"epoch": 1.37890625, "grad_norm": 0.015219186432659626, "learning_rate": 3.191597261653475e-05, "loss": 0.00014846389240119606, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00015, "step": 353, "tokens/total": 10679184, "tokens/train_per_sec_per_gpu": 29.27, "tokens/trainable": 161567} +{"epoch": 1.3828125, "grad_norm": 0.002392370719462633, "learning_rate": 3.166726850239794e-05, "loss": 2.664637577254325e-05, "memory/device_reserved (GiB)": 35.41, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00003, "step": 354, "tokens/total": 10709648, "tokens/train_per_sec_per_gpu": 37.27, "tokens/trainable": 162032} +{"epoch": 1.38671875, "grad_norm": 0.012528779916465282, "learning_rate": 3.141953535845912e-05, "loss": 0.00010976240446325392, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00011, "step": 355, "tokens/total": 10740128, "tokens/train_per_sec_per_gpu": 32.56, "tokens/trainable": 162460} +{"epoch": 1.390625, "grad_norm": 0.0019507119432091713, "learning_rate": 3.11727834939056e-05, "loss": 2.361126826144755e-05, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.61, "memory/max_allocated (GiB)": 33.61, "ppl": 1.00002, "step": 356, "tokens/total": 10770128, "tokens/train_per_sec_per_gpu": 28.61, "tokens/trainable": 162875} +{"epoch": 1.39453125, "grad_norm": 0.013965644873678684, "learning_rate": 3.092702317708967e-05, "loss": 0.00010500443022465333, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00011, "step": 357, "tokens/total": 10800432, "tokens/train_per_sec_per_gpu": 35.43, "tokens/trainable": 163329} +{"epoch": 1.3984375, "grad_norm": 0.08449137210845947, "learning_rate": 3.0682264635101276e-05, "loss": 0.00026978645473718643, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00027, "step": 358, "tokens/total": 10830592, "tokens/train_per_sec_per_gpu": 33.54, "tokens/trainable": 163778} +{"epoch": 1.40234375, "grad_norm": 0.07536502927541733, "learning_rate": 3.0438518053342407e-05, "loss": 0.001174358418211341, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00118, "step": 359, "tokens/total": 10861088, "tokens/train_per_sec_per_gpu": 32.17, "tokens/trainable": 164219} +{"epoch": 1.40625, "grad_norm": 0.024741269648075104, "learning_rate": 3.0195793575103266e-05, "loss": 0.00027250737184658647, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00027, "step": 360, "tokens/total": 10891552, "tokens/train_per_sec_per_gpu": 34.66, "tokens/trainable": 164666} +{"epoch": 1.41015625, "grad_norm": 0.6777807474136353, "learning_rate": 2.9954101301140146e-05, "loss": 0.016698025166988373, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.01684, "step": 361, "tokens/total": 10921776, "tokens/train_per_sec_per_gpu": 37.15, "tokens/trainable": 165128} +{"epoch": 1.4140625, "grad_norm": 0.2795168161392212, "learning_rate": 2.9713451289255123e-05, "loss": 0.00555866863578558, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00557, "step": 362, "tokens/total": 10952288, "tokens/train_per_sec_per_gpu": 35.32, "tokens/trainable": 165614} +{"epoch": 1.41796875, "grad_norm": 0.19862814247608185, "learning_rate": 2.9473853553877484e-05, "loss": 0.003639942966401577, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00365, "step": 363, "tokens/total": 10982496, "tokens/train_per_sec_per_gpu": 34.59, "tokens/trainable": 166062} +{"epoch": 1.421875, "grad_norm": 0.020105330273509026, "learning_rate": 2.9235318065647e-05, "loss": 0.0001461556676076725, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00015, "step": 364, "tokens/total": 11012704, "tokens/train_per_sec_per_gpu": 34.55, "tokens/trainable": 166513} +{"epoch": 1.42578125, "grad_norm": 0.09536039084196091, "learning_rate": 2.8997854750998964e-05, "loss": 0.0006832765066064894, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00068, "step": 365, "tokens/total": 11043024, "tokens/train_per_sec_per_gpu": 36.77, "tokens/trainable": 167001} +{"epoch": 1.4296875, "grad_norm": 0.015557884238660336, "learning_rate": 2.8761473491751258e-05, "loss": 0.00021839249529875815, "memory/device_reserved (GiB)": 35.84, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00022, "step": 366, "tokens/total": 11073392, "tokens/train_per_sec_per_gpu": 30.62, "tokens/trainable": 167450} +{"epoch": 1.43359375, "grad_norm": 0.003135454375296831, "learning_rate": 2.8526184124692883e-05, "loss": 4.49724102509208e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00004, "step": 367, "tokens/total": 11103936, "tokens/train_per_sec_per_gpu": 31.79, "tokens/trainable": 167895} +{"epoch": 1.4375, "grad_norm": 0.03941775858402252, "learning_rate": 2.829199644117484e-05, "loss": 0.0003685755073092878, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00037, "step": 368, "tokens/total": 11134256, "tokens/train_per_sec_per_gpu": 36.0, "tokens/trainable": 168338} +{"epoch": 1.44140625, "grad_norm": 0.07836400717496872, "learning_rate": 2.8058920186702553e-05, "loss": 0.0015341172693297267, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00154, "step": 369, "tokens/total": 11164768, "tokens/train_per_sec_per_gpu": 32.61, "tokens/trainable": 168813} +{"epoch": 1.4453125, "grad_norm": 0.47108760476112366, "learning_rate": 2.782696506053033e-05, "loss": 0.009904367849230766, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00995, "step": 370, "tokens/total": 11195136, "tokens/train_per_sec_per_gpu": 37.27, "tokens/trainable": 169315} +{"epoch": 1.44921875, "grad_norm": 0.03302016109228134, "learning_rate": 2.7596140715257824e-05, "loss": 0.0005742749781347811, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00057, "step": 371, "tokens/total": 11225680, "tokens/train_per_sec_per_gpu": 35.47, "tokens/trainable": 169756} +{"epoch": 1.453125, "grad_norm": 0.012702704407274723, "learning_rate": 2.7366456756428184e-05, "loss": 0.0001972567697521299, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.0002, "step": 372, "tokens/total": 11256112, "tokens/train_per_sec_per_gpu": 33.3, "tokens/trainable": 170222} +{"epoch": 1.45703125, "grad_norm": 0.005610617808997631, "learning_rate": 2.7137922742128486e-05, "loss": 8.283223723992705e-05, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00008, "step": 373, "tokens/total": 11286304, "tokens/train_per_sec_per_gpu": 33.6, "tokens/trainable": 170687} +{"epoch": 1.4609375, "grad_norm": 0.13766004145145416, "learning_rate": 2.691054818259188e-05, "loss": 0.00119793601334095, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.0012, "step": 374, "tokens/total": 11316800, "tokens/train_per_sec_per_gpu": 32.43, "tokens/trainable": 171163} +{"epoch": 1.46484375, "grad_norm": 0.4246460795402527, "learning_rate": 2.6684342539801933e-05, "loss": 0.007628333754837513, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00766, "step": 375, "tokens/total": 11346944, "tokens/train_per_sec_per_gpu": 34.48, "tokens/trainable": 171626} +{"epoch": 1.46875, "grad_norm": 0.008275235071778297, "learning_rate": 2.645931522709877e-05, "loss": 0.00011664340127026662, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00012, "step": 376, "tokens/total": 11377376, "tokens/train_per_sec_per_gpu": 28.49, "tokens/trainable": 172043} +{"epoch": 1.47265625, "grad_norm": 0.01685458980500698, "learning_rate": 2.6235475608787365e-05, "loss": 0.00018345039279665798, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00018, "step": 377, "tokens/total": 11407680, "tokens/train_per_sec_per_gpu": 36.55, "tokens/trainable": 172519} +{"epoch": 1.4765625, "grad_norm": 0.33050504326820374, "learning_rate": 2.6012832999747916e-05, "loss": 0.003047082107514143, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00305, "step": 378, "tokens/total": 11437712, "tokens/train_per_sec_per_gpu": 34.88, "tokens/trainable": 172977} +{"epoch": 1.48046875, "grad_norm": 0.1227385401725769, "learning_rate": 2.579139666504821e-05, "loss": 0.0010191744659096003, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00102, "step": 379, "tokens/total": 11468176, "tokens/train_per_sec_per_gpu": 36.22, "tokens/trainable": 173461} +{"epoch": 1.484375, "grad_norm": 0.0017978748073801398, "learning_rate": 2.557117581955798e-05, "loss": 4.36538175563328e-05, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00004, "step": 380, "tokens/total": 11498352, "tokens/train_per_sec_per_gpu": 30.87, "tokens/trainable": 173926} +{"epoch": 1.48828125, "grad_norm": 0.2067205160856247, "learning_rate": 2.5352179627565532e-05, "loss": 0.005216958001255989, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00523, "step": 381, "tokens/total": 11528768, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 174397} +{"epoch": 1.4921875, "grad_norm": 0.0806688517332077, "learning_rate": 2.5134417202396277e-05, "loss": 0.0011719370959326625, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00117, "step": 382, "tokens/total": 11558976, "tokens/train_per_sec_per_gpu": 33.27, "tokens/trainable": 174852} +{"epoch": 1.49609375, "grad_norm": 0.027668939903378487, "learning_rate": 2.491789760603361e-05, "loss": 0.00021959877631161362, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00022, "step": 383, "tokens/total": 11589520, "tokens/train_per_sec_per_gpu": 31.88, "tokens/trainable": 175293} +{"epoch": 1.5, "grad_norm": 0.12873099744319916, "learning_rate": 2.4702629848741764e-05, "loss": 0.0015978836454451084, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0016, "step": 384, "tokens/total": 11619760, "tokens/train_per_sec_per_gpu": 32.87, "tokens/trainable": 175731} +{"epoch": 1.50390625, "grad_norm": 0.037335313856601715, "learning_rate": 2.4488622888690785e-05, "loss": 0.0005178801948204637, "memory/device_reserved (GiB)": 35.89, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00052, "step": 385, "tokens/total": 11650192, "tokens/train_per_sec_per_gpu": 35.62, "tokens/trainable": 176215} +{"epoch": 1.5078125, "grad_norm": 0.06761232763528824, "learning_rate": 2.427588563158384e-05, "loss": 0.0005836813943460584, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00058, "step": 386, "tokens/total": 11680672, "tokens/train_per_sec_per_gpu": 34.8, "tokens/trainable": 176663} +{"epoch": 1.51171875, "grad_norm": 0.0501842126250267, "learning_rate": 2.406442693028651e-05, "loss": 0.0005241117323748767, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00052, "step": 387, "tokens/total": 11710880, "tokens/train_per_sec_per_gpu": 33.11, "tokens/trainable": 177127} +{"epoch": 1.515625, "grad_norm": 0.08328797668218613, "learning_rate": 2.3854255584458547e-05, "loss": 0.00114786671474576, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00115, "step": 388, "tokens/total": 11741408, "tokens/train_per_sec_per_gpu": 36.56, "tokens/trainable": 177613} +{"epoch": 1.51953125, "grad_norm": 0.020661214366555214, "learning_rate": 2.3645380340187508e-05, "loss": 0.00030904496088624, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00031, "step": 389, "tokens/total": 11771712, "tokens/train_per_sec_per_gpu": 33.47, "tokens/trainable": 178062} +{"epoch": 1.5234375, "grad_norm": 0.13835693895816803, "learning_rate": 2.3437809889624914e-05, "loss": 0.0014695629943162203, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00147, "step": 390, "tokens/total": 11801872, "tokens/train_per_sec_per_gpu": 36.96, "tokens/trainable": 178535} +{"epoch": 1.52734375, "grad_norm": 0.061385851353406906, "learning_rate": 2.3231552870624487e-05, "loss": 0.0007466091774404049, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00075, "step": 391, "tokens/total": 11832480, "tokens/train_per_sec_per_gpu": 34.42, "tokens/trainable": 179004} +{"epoch": 1.53125, "grad_norm": 0.003179469145834446, "learning_rate": 2.3026617866382657e-05, "loss": 5.848719956702553e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00006, "step": 392, "tokens/total": 11862672, "tokens/train_per_sec_per_gpu": 33.49, "tokens/trainable": 179461} +{"epoch": 1.53515625, "grad_norm": 0.021778129041194916, "learning_rate": 2.2823013405081507e-05, "loss": 0.00023812934523448348, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00024, "step": 393, "tokens/total": 11893024, "tokens/train_per_sec_per_gpu": 33.22, "tokens/trainable": 179893} +{"epoch": 1.5390625, "grad_norm": 0.10550826787948608, "learning_rate": 2.2620747959533722e-05, "loss": 0.0007996093481779099, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.0008, "step": 394, "tokens/total": 11923328, "tokens/train_per_sec_per_gpu": 34.41, "tokens/trainable": 180344} +{"epoch": 1.54296875, "grad_norm": 0.003161423373967409, "learning_rate": 2.2419829946830123e-05, "loss": 5.3679388656746596e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00005, "step": 395, "tokens/total": 11953792, "tokens/train_per_sec_per_gpu": 32.21, "tokens/trainable": 180757} +{"epoch": 1.546875, "grad_norm": 0.05268798768520355, "learning_rate": 2.2220267727989325e-05, "loss": 0.0009187876130454242, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00092, "step": 396, "tokens/total": 11984208, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 181228} +{"epoch": 1.55078125, "grad_norm": 0.0017847128910943866, "learning_rate": 2.202206960760984e-05, "loss": 3.819329140242189e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00004, "step": 397, "tokens/total": 12014608, "tokens/train_per_sec_per_gpu": 33.66, "tokens/trainable": 181716} +{"epoch": 1.5546875, "grad_norm": 0.1685003638267517, "learning_rate": 2.182524383352446e-05, "loss": 0.0024524193722754717, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00246, "step": 398, "tokens/total": 12045088, "tokens/train_per_sec_per_gpu": 36.01, "tokens/trainable": 182152} +{"epoch": 1.55859375, "grad_norm": 0.11644069850444794, "learning_rate": 2.1629798596457056e-05, "loss": 0.001032137661240995, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00103, "step": 399, "tokens/total": 12075280, "tokens/train_per_sec_per_gpu": 35.37, "tokens/trainable": 182623} +{"epoch": 1.5625, "grad_norm": 0.004440188407897949, "learning_rate": 2.1435742029681725e-05, "loss": 7.53562999307178e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00008, "step": 400, "tokens/total": 12105792, "tokens/train_per_sec_per_gpu": 26.92, "tokens/trainable": 183039} +{"epoch": 1.56640625, "grad_norm": 0.06607124954462051, "learning_rate": 2.124308220868431e-05, "loss": 0.0008609758224338293, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.71, "memory/max_allocated (GiB)": 33.71, "ppl": 1.00086, "step": 401, "tokens/total": 12135680, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 183487} +{"epoch": 1.5703125, "grad_norm": 0.0036999934818595648, "learning_rate": 2.105182715082638e-05, "loss": 6.075722922105342e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00006, "step": 402, "tokens/total": 12165920, "tokens/train_per_sec_per_gpu": 39.03, "tokens/trainable": 183987} +{"epoch": 1.57421875, "grad_norm": 0.010572529397904873, "learning_rate": 2.0861984815011552e-05, "loss": 0.0001372473343508318, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00014, "step": 403, "tokens/total": 12196080, "tokens/train_per_sec_per_gpu": 32.86, "tokens/trainable": 184459} +{"epoch": 1.578125, "grad_norm": 0.3721093237400055, "learning_rate": 2.0673563101354323e-05, "loss": 0.003381735645234585, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00339, "step": 404, "tokens/total": 12226128, "tokens/train_per_sec_per_gpu": 31.95, "tokens/trainable": 184887} +{"epoch": 1.58203125, "grad_norm": 0.08975216001272202, "learning_rate": 2.0486569850851317e-05, "loss": 0.0008267756784334779, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.72, "memory/max_allocated (GiB)": 33.72, "ppl": 1.00083, "step": 405, "tokens/total": 12256176, "tokens/train_per_sec_per_gpu": 29.59, "tokens/trainable": 185320} +{"epoch": 1.5859375, "grad_norm": 0.4965755343437195, "learning_rate": 2.0301012845054956e-05, "loss": 0.012792397290468216, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.01287, "step": 406, "tokens/total": 12286416, "tokens/train_per_sec_per_gpu": 31.77, "tokens/trainable": 185774} +{"epoch": 1.58984375, "grad_norm": 0.0017864442197605968, "learning_rate": 2.011689980574966e-05, "loss": 3.962808841606602e-05, "memory/device_reserved (GiB)": 35.88, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00004, "step": 407, "tokens/total": 12316720, "tokens/train_per_sec_per_gpu": 37.15, "tokens/trainable": 186233} +{"epoch": 1.59375, "grad_norm": 0.006781951989978552, "learning_rate": 1.993423839463052e-05, "loss": 6.552625563926995e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00007, "step": 408, "tokens/total": 12347072, "tokens/train_per_sec_per_gpu": 33.14, "tokens/trainable": 186671} +{"epoch": 1.59765625, "grad_norm": 0.02034713886678219, "learning_rate": 1.975303621298445e-05, "loss": 0.00025357812410220504, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00025, "step": 409, "tokens/total": 12377504, "tokens/train_per_sec_per_gpu": 34.29, "tokens/trainable": 187102} +{"epoch": 1.6015625, "grad_norm": 0.003248689230531454, "learning_rate": 1.957330080137385e-05, "loss": 5.506081652129069e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00006, "step": 410, "tokens/total": 12408048, "tokens/train_per_sec_per_gpu": 35.25, "tokens/trainable": 187549} +{"epoch": 1.60546875, "grad_norm": 0.16686522960662842, "learning_rate": 1.9395039639322864e-05, "loss": 0.0007388527155853808, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00074, "step": 411, "tokens/total": 12438256, "tokens/train_per_sec_per_gpu": 34.61, "tokens/trainable": 187990} +{"epoch": 1.609375, "grad_norm": 0.005520866252481937, "learning_rate": 1.9218260145006073e-05, "loss": 0.00011023284605471417, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00011, "step": 412, "tokens/total": 12466608, "tokens/train_per_sec_per_gpu": 39.21, "tokens/trainable": 188426} +{"epoch": 1.61328125, "grad_norm": 0.1133890226483345, "learning_rate": 1.904296967493982e-05, "loss": 0.0016134778270497918, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00161, "step": 413, "tokens/total": 12497136, "tokens/train_per_sec_per_gpu": 31.96, "tokens/trainable": 188877} +{"epoch": 1.6171875, "grad_norm": 0.09548613429069519, "learning_rate": 1.8869175523676064e-05, "loss": 0.00023090933973435313, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00023, "step": 414, "tokens/total": 12527120, "tokens/train_per_sec_per_gpu": 37.84, "tokens/trainable": 189346} +{"epoch": 1.62109375, "grad_norm": 0.0033725884277373552, "learning_rate": 1.869688492349885e-05, "loss": 4.779225855600089e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00005, "step": 415, "tokens/total": 12557248, "tokens/train_per_sec_per_gpu": 32.34, "tokens/trainable": 189759} +{"epoch": 1.625, "grad_norm": 0.0031356913968920708, "learning_rate": 1.85261050441233e-05, "loss": 4.390432150103152e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00004, "step": 416, "tokens/total": 12587728, "tokens/train_per_sec_per_gpu": 34.5, "tokens/trainable": 190225} +{"epoch": 1.62890625, "grad_norm": 0.0025502180214971304, "learning_rate": 1.8356842992397304e-05, "loss": 5.062075433670543e-05, "memory/device_reserved (GiB)": 35.96, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00005, "step": 417, "tokens/total": 12618160, "tokens/train_per_sec_per_gpu": 35.49, "tokens/trainable": 190717} +{"epoch": 1.6328125, "grad_norm": 0.006292750593274832, "learning_rate": 1.8189105812005714e-05, "loss": 9.369729377795011e-05, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00009, "step": 418, "tokens/total": 12648672, "tokens/train_per_sec_per_gpu": 31.28, "tokens/trainable": 191135} +{"epoch": 1.63671875, "grad_norm": 0.1299043893814087, "learning_rate": 1.802290048317732e-05, "loss": 0.002147798193618655, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00215, "step": 419, "tokens/total": 12678944, "tokens/train_per_sec_per_gpu": 31.86, "tokens/trainable": 191559} +{"epoch": 1.640625, "grad_norm": 0.017163539305329323, "learning_rate": 1.785823392239424e-05, "loss": 0.0001932488230522722, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00019, "step": 420, "tokens/total": 12709408, "tokens/train_per_sec_per_gpu": 36.98, "tokens/trainable": 192033} +{"epoch": 1.64453125, "grad_norm": 0.0045894598588347435, "learning_rate": 1.7695112982104225e-05, "loss": 5.246631917543709e-05, "memory/device_reserved (GiB)": 35.81, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 421, "tokens/total": 12739632, "tokens/train_per_sec_per_gpu": 31.53, "tokens/trainable": 192482} +{"epoch": 1.6484375, "grad_norm": 0.048289310187101364, "learning_rate": 1.7533544450435433e-05, "loss": 0.0002731278946157545, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00027, "step": 422, "tokens/total": 12770224, "tokens/train_per_sec_per_gpu": 33.0, "tokens/trainable": 192937} +{"epoch": 1.65234375, "grad_norm": 0.033917587250471115, "learning_rate": 1.7373535050913946e-05, "loss": 0.0003945066418964416, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00039, "step": 423, "tokens/total": 12800640, "tokens/train_per_sec_per_gpu": 37.81, "tokens/trainable": 193444} +{"epoch": 1.65625, "grad_norm": 0.011636070907115936, "learning_rate": 1.721509144218405e-05, "loss": 0.00016518094344064593, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00017, "step": 424, "tokens/total": 12831104, "tokens/train_per_sec_per_gpu": 36.92, "tokens/trainable": 193925} +{"epoch": 1.66015625, "grad_norm": 0.06534689664840698, "learning_rate": 1.705822021773101e-05, "loss": 0.0008454776252619922, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00085, "step": 425, "tokens/total": 12861296, "tokens/train_per_sec_per_gpu": 29.68, "tokens/trainable": 194342} +{"epoch": 1.6640625, "grad_norm": 0.07199558615684509, "learning_rate": 1.69029279056068e-05, "loss": 0.0007431323174387217, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00074, "step": 426, "tokens/total": 12891504, "tokens/train_per_sec_per_gpu": 35.4, "tokens/trainable": 194830} +{"epoch": 1.66796875, "grad_norm": 0.030936896800994873, "learning_rate": 1.6749220968158415e-05, "loss": 0.00027934886747971177, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00028, "step": 427, "tokens/total": 12921840, "tokens/train_per_sec_per_gpu": 34.47, "tokens/trainable": 195296} +{"epoch": 1.671875, "grad_norm": 0.0030303315725177526, "learning_rate": 1.659710580175893e-05, "loss": 4.713955058832653e-05, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00005, "step": 428, "tokens/total": 12952224, "tokens/train_per_sec_per_gpu": 32.33, "tokens/trainable": 195714} +{"epoch": 1.67578125, "grad_norm": 0.055711835622787476, "learning_rate": 1.644658873654133e-05, "loss": 0.0008513920474797487, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.75, "memory/max_allocated (GiB)": 33.75, "ppl": 1.00085, "step": 429, "tokens/total": 12982416, "tokens/train_per_sec_per_gpu": 33.31, "tokens/trainable": 196179} +{"epoch": 1.6796875, "grad_norm": 0.014902536757290363, "learning_rate": 1.629767603613508e-05, "loss": 0.00024280000070575625, "memory/device_reserved (GiB)": 35.83, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00024, "step": 430, "tokens/total": 13012800, "tokens/train_per_sec_per_gpu": 32.82, "tokens/trainable": 196660} +{"epoch": 1.68359375, "grad_norm": 0.003592823399230838, "learning_rate": 1.615037389740547e-05, "loss": 5.272522321320139e-05, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.00005, "step": 431, "tokens/total": 13043392, "tokens/train_per_sec_per_gpu": 31.51, "tokens/trainable": 197091} +{"epoch": 1.6875, "grad_norm": 0.03395218774676323, "learning_rate": 1.600468845019576e-05, "loss": 0.0003875850234180689, "memory/device_reserved (GiB)": 36.35, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00039, "step": 432, "tokens/total": 13073984, "tokens/train_per_sec_per_gpu": 39.07, "tokens/trainable": 197589} +{"epoch": 1.69140625, "grad_norm": 0.06874402612447739, "learning_rate": 1.5860625757072092e-05, "loss": 0.0007863138453103602, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.95, "memory/max_allocated (GiB)": 33.95, "ppl": 1.00079, "step": 433, "tokens/total": 13104512, "tokens/train_per_sec_per_gpu": 33.15, "tokens/trainable": 198047} +{"epoch": 1.6953125, "grad_norm": 0.04059433192014694, "learning_rate": 1.571819181307116e-05, "loss": 0.00044966390123590827, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00045, "step": 434, "tokens/total": 13134656, "tokens/train_per_sec_per_gpu": 32.89, "tokens/trainable": 198465} +{"epoch": 1.69921875, "grad_norm": 0.1531708687543869, "learning_rate": 1.557739254545075e-05, "loss": 0.0016073889564722776, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00161, "step": 435, "tokens/total": 13165248, "tokens/train_per_sec_per_gpu": 34.75, "tokens/trainable": 198937} +{"epoch": 1.703125, "grad_norm": 0.03812495246529579, "learning_rate": 1.543823381344311e-05, "loss": 0.0005045874277129769, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.0005, "step": 436, "tokens/total": 13195648, "tokens/train_per_sec_per_gpu": 33.7, "tokens/trainable": 199385} +{"epoch": 1.70703125, "grad_norm": 0.3262580633163452, "learning_rate": 1.5300721408011114e-05, "loss": 0.008203462697565556, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00824, "step": 437, "tokens/total": 13226176, "tokens/train_per_sec_per_gpu": 34.78, "tokens/trainable": 199869} +{"epoch": 1.7109375, "grad_norm": 0.0010299552232027054, "learning_rate": 1.5164861051607254e-05, "loss": 1.8928842109744437e-05, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 438, "tokens/total": 13256480, "tokens/train_per_sec_per_gpu": 35.38, "tokens/trainable": 200377} +{"epoch": 1.71484375, "grad_norm": 0.02154897339642048, "learning_rate": 1.5030658397935521e-05, "loss": 0.00018170180555898696, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00018, "step": 439, "tokens/total": 13286528, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 200824} +{"epoch": 1.71875, "grad_norm": 0.02101851999759674, "learning_rate": 1.4898119031716104e-05, "loss": 0.00032509121228940785, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00033, "step": 440, "tokens/total": 13317024, "tokens/train_per_sec_per_gpu": 35.51, "tokens/trainable": 201297} +{"epoch": 1.72265625, "grad_norm": 0.007308815140277147, "learning_rate": 1.476724846845306e-05, "loss": 0.00010265046876156703, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.0001, "step": 441, "tokens/total": 13347440, "tokens/train_per_sec_per_gpu": 34.67, "tokens/trainable": 201764} +{"epoch": 1.7265625, "grad_norm": 0.04583185538649559, "learning_rate": 1.463805215420471e-05, "loss": 0.0005665539065375924, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00057, "step": 442, "tokens/total": 13377664, "tokens/train_per_sec_per_gpu": 30.84, "tokens/trainable": 202204} +{"epoch": 1.73046875, "grad_norm": 0.2436366230249405, "learning_rate": 1.451053546535705e-05, "loss": 0.004352548159658909, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00436, "step": 443, "tokens/total": 13408080, "tokens/train_per_sec_per_gpu": 36.59, "tokens/trainable": 202642} +{"epoch": 1.734375, "grad_norm": 0.015320447273552418, "learning_rate": 1.438470370840001e-05, "loss": 0.0002279828186146915, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.64, "memory/max_allocated (GiB)": 33.64, "ppl": 1.00023, "step": 444, "tokens/total": 13437824, "tokens/train_per_sec_per_gpu": 31.22, "tokens/trainable": 203050} +{"epoch": 1.73828125, "grad_norm": 0.0014951409539207816, "learning_rate": 1.4260562119706606e-05, "loss": 2.6277737561031245e-05, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00003, "step": 445, "tokens/total": 13468032, "tokens/train_per_sec_per_gpu": 34.63, "tokens/trainable": 203509} +{"epoch": 1.7421875, "grad_norm": 0.009211640805006027, "learning_rate": 1.413811586531508e-05, "loss": 6.653230229858309e-05, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00007, "step": 446, "tokens/total": 13498096, "tokens/train_per_sec_per_gpu": 35.78, "tokens/trainable": 203976} +{"epoch": 1.74609375, "grad_norm": 0.15356089174747467, "learning_rate": 1.4017370040713884e-05, "loss": 0.0011966564925387502, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.0012, "step": 447, "tokens/total": 13528288, "tokens/train_per_sec_per_gpu": 37.97, "tokens/trainable": 204412} +{"epoch": 1.75, "grad_norm": 0.008474756963551044, "learning_rate": 1.3898329670629645e-05, "loss": 9.093034896068275e-05, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00009, "step": 448, "tokens/total": 13558752, "tokens/train_per_sec_per_gpu": 36.02, "tokens/trainable": 204868} +{"epoch": 1.75390625, "grad_norm": 0.015406596474349499, "learning_rate": 1.3780999708818058e-05, "loss": 0.0001275191898457706, "memory/device_reserved (GiB)": 36.8, "memory/max_active (GiB)": 33.93, "memory/max_allocated (GiB)": 33.93, "ppl": 1.00013, "step": 449, "tokens/total": 13589136, "tokens/train_per_sec_per_gpu": 34.11, "tokens/trainable": 205338} +{"epoch": 1.7578125, "grad_norm": 0.12964943051338196, "learning_rate": 1.3665385037857758e-05, "loss": 0.0008908645249903202, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00089, "step": 450, "tokens/total": 13619472, "tokens/train_per_sec_per_gpu": 31.08, "tokens/trainable": 205764} +{"epoch": 1.76171875, "grad_norm": 0.010256104171276093, "learning_rate": 1.3551490468947126e-05, "loss": 0.00014732024283148348, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00015, "step": 451, "tokens/total": 13649872, "tokens/train_per_sec_per_gpu": 32.08, "tokens/trainable": 206201} +{"epoch": 1.765625, "grad_norm": 0.006175735965371132, "learning_rate": 1.3439320741704075e-05, "loss": 9.223521192325279e-05, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00009, "step": 452, "tokens/total": 13680064, "tokens/train_per_sec_per_gpu": 35.89, "tokens/trainable": 206657} +{"epoch": 1.76953125, "grad_norm": 0.060654036700725555, "learning_rate": 1.3328880523968808e-05, "loss": 0.0005743277142755687, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00057, "step": 453, "tokens/total": 13710592, "tokens/train_per_sec_per_gpu": 37.59, "tokens/trainable": 207138} +{"epoch": 1.7734375, "grad_norm": 0.003730883589014411, "learning_rate": 1.3220174411609587e-05, "loss": 5.224360211286694e-05, "memory/device_reserved (GiB)": 35.7, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00005, "step": 454, "tokens/total": 13740960, "tokens/train_per_sec_per_gpu": 36.25, "tokens/trainable": 207583} +{"epoch": 1.77734375, "grad_norm": 0.09384409338235855, "learning_rate": 1.3113206928331471e-05, "loss": 0.0008255710126832128, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00083, "step": 455, "tokens/total": 13771296, "tokens/train_per_sec_per_gpu": 33.58, "tokens/trainable": 208025} +{"epoch": 1.78125, "grad_norm": 0.11239906400442123, "learning_rate": 1.300798252548806e-05, "loss": 0.0013365334598347545, "memory/device_reserved (GiB)": 35.78, "memory/max_active (GiB)": 33.86, "memory/max_allocated (GiB)": 33.86, "ppl": 1.00134, "step": 456, "tokens/total": 13801664, "tokens/train_per_sec_per_gpu": 35.32, "tokens/trainable": 208453} +{"epoch": 1.78515625, "grad_norm": 0.007983297109603882, "learning_rate": 1.2904505581896265e-05, "loss": 6.189515988808125e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00006, "step": 457, "tokens/total": 13831936, "tokens/train_per_sec_per_gpu": 35.89, "tokens/trainable": 208947} +{"epoch": 1.7890625, "grad_norm": 0.0063696797005832195, "learning_rate": 1.2802780403654082e-05, "loss": 9.090925595955923e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00009, "step": 458, "tokens/total": 13862160, "tokens/train_per_sec_per_gpu": 35.33, "tokens/trainable": 209435} +{"epoch": 1.79296875, "grad_norm": 0.008154391311109066, "learning_rate": 1.2702811223961408e-05, "loss": 7.762061432003975e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00008, "step": 459, "tokens/total": 13892320, "tokens/train_per_sec_per_gpu": 33.98, "tokens/trainable": 209885} +{"epoch": 1.796875, "grad_norm": 0.015743529424071312, "learning_rate": 1.2604602202943861e-05, "loss": 0.00018653822189662606, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.91, "memory/max_allocated (GiB)": 33.91, "ppl": 1.00019, "step": 460, "tokens/total": 13922752, "tokens/train_per_sec_per_gpu": 34.08, "tokens/trainable": 210344} +{"epoch": 1.80078125, "grad_norm": 0.011715077795088291, "learning_rate": 1.2508157427479686e-05, "loss": 0.00016241846606135368, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.65, "memory/max_allocated (GiB)": 33.65, "ppl": 1.00016, "step": 461, "tokens/total": 13952816, "tokens/train_per_sec_per_gpu": 32.19, "tokens/trainable": 210814} +{"epoch": 1.8046875, "grad_norm": 0.04297679290175438, "learning_rate": 1.2413480911029655e-05, "loss": 0.0005623188917525113, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.38, "memory/max_allocated (GiB)": 33.38, "ppl": 1.00056, "step": 462, "tokens/total": 13981040, "tokens/train_per_sec_per_gpu": 30.87, "tokens/trainable": 211238} +{"epoch": 1.80859375, "grad_norm": 0.004571136087179184, "learning_rate": 1.2320576593470082e-05, "loss": 4.7571604227414355e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.79, "memory/max_allocated (GiB)": 33.79, "ppl": 1.00005, "step": 463, "tokens/total": 14011248, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 211700} +{"epoch": 1.8125, "grad_norm": 0.0022084659431129694, "learning_rate": 1.2229448340928828e-05, "loss": 1.850002445280552e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.74, "memory/max_allocated (GiB)": 33.74, "ppl": 1.00002, "step": 464, "tokens/total": 14041504, "tokens/train_per_sec_per_gpu": 29.47, "tokens/trainable": 212138} +{"epoch": 1.81640625, "grad_norm": 0.08595507591962814, "learning_rate": 1.2140099945624458e-05, "loss": 0.0008459303062409163, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00085, "step": 465, "tokens/total": 14071728, "tokens/train_per_sec_per_gpu": 33.14, "tokens/trainable": 212623} +{"epoch": 1.8203125, "grad_norm": 0.0007654759101569653, "learning_rate": 1.205253512570841e-05, "loss": 1.6132753444253467e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00002, "step": 466, "tokens/total": 14101824, "tokens/train_per_sec_per_gpu": 33.37, "tokens/trainable": 213062} +{"epoch": 1.82421875, "grad_norm": 0.0005283593782223761, "learning_rate": 1.1966757525110255e-05, "loss": 1.0958694474538788e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00001, "step": 467, "tokens/total": 14132080, "tokens/train_per_sec_per_gpu": 32.03, "tokens/trainable": 213501} +{"epoch": 1.828125, "grad_norm": 0.0003768271126318723, "learning_rate": 1.1882770713386095e-05, "loss": 1.1069126230722759e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.73, "memory/max_allocated (GiB)": 33.73, "ppl": 1.00001, "step": 468, "tokens/total": 14162112, "tokens/train_per_sec_per_gpu": 35.13, "tokens/trainable": 213942} +{"epoch": 1.83203125, "grad_norm": 0.010821828618645668, "learning_rate": 1.180057818556998e-05, "loss": 7.607591396663338e-05, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00008, "step": 469, "tokens/total": 14192784, "tokens/train_per_sec_per_gpu": 37.33, "tokens/trainable": 214410} +{"epoch": 1.8359375, "grad_norm": 0.008473122492432594, "learning_rate": 1.1720183362028494e-05, "loss": 0.00010939891217276454, "memory/device_reserved (GiB)": 35.82, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00011, "step": 470, "tokens/total": 14221008, "tokens/train_per_sec_per_gpu": 36.09, "tokens/trainable": 214844} +{"epoch": 1.83984375, "grad_norm": 0.0007591558387503028, "learning_rate": 1.1641589588318387e-05, "loss": 1.7881218809634447e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00002, "step": 471, "tokens/total": 14251536, "tokens/train_per_sec_per_gpu": 35.96, "tokens/trainable": 215336} +{"epoch": 1.84375, "grad_norm": 0.05926339700818062, "learning_rate": 1.1564800135047418e-05, "loss": 0.0005082086427137256, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00051, "step": 472, "tokens/total": 14281664, "tokens/train_per_sec_per_gpu": 33.35, "tokens/trainable": 215777} +{"epoch": 1.84765625, "grad_norm": 0.12773646414279938, "learning_rate": 1.148981819773816e-05, "loss": 0.002721701515838504, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00273, "step": 473, "tokens/total": 14312048, "tokens/train_per_sec_per_gpu": 35.25, "tokens/trainable": 216222} +{"epoch": 1.8515625, "grad_norm": 0.028139308094978333, "learning_rate": 1.1416646896695086e-05, "loss": 0.00013170376769267023, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00013, "step": 474, "tokens/total": 14342112, "tokens/train_per_sec_per_gpu": 33.79, "tokens/trainable": 216662} +{"epoch": 1.85546875, "grad_norm": 0.003854855429381132, "learning_rate": 1.1345289276874717e-05, "loss": 5.479517130879685e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00005, "step": 475, "tokens/total": 14372176, "tokens/train_per_sec_per_gpu": 31.26, "tokens/trainable": 217068} +{"epoch": 1.859375, "grad_norm": 0.0007855606963858008, "learning_rate": 1.1275748307758873e-05, "loss": 1.5755449567222968e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00002, "step": 476, "tokens/total": 14402592, "tokens/train_per_sec_per_gpu": 38.21, "tokens/trainable": 217558} +{"epoch": 1.86328125, "grad_norm": 0.004396429285407066, "learning_rate": 1.1208026883231147e-05, "loss": 5.772991426056251e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00006, "step": 477, "tokens/total": 14432816, "tokens/train_per_sec_per_gpu": 35.98, "tokens/trainable": 218025} +{"epoch": 1.8671875, "grad_norm": 0.03181939572095871, "learning_rate": 1.1142127821456433e-05, "loss": 0.00036515528336167336, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00037, "step": 478, "tokens/total": 14463424, "tokens/train_per_sec_per_gpu": 32.81, "tokens/trainable": 218498} +{"epoch": 1.87109375, "grad_norm": 0.004733108915388584, "learning_rate": 1.1078053864763674e-05, "loss": 4.52871281595435e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.83, "memory/max_allocated (GiB)": 33.83, "ppl": 1.00005, "step": 479, "tokens/total": 14493440, "tokens/train_per_sec_per_gpu": 33.17, "tokens/trainable": 218946} +{"epoch": 1.875, "grad_norm": 0.0026765780057758093, "learning_rate": 1.1015807679531756e-05, "loss": 3.31170012941584e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.98, "memory/max_allocated (GiB)": 33.98, "ppl": 1.00003, "step": 480, "tokens/total": 14524000, "tokens/train_per_sec_per_gpu": 34.52, "tokens/trainable": 219387} +{"epoch": 1.87890625, "grad_norm": 0.006714611314237118, "learning_rate": 1.0955391856078528e-05, "loss": 4.494276799960062e-05, "memory/device_reserved (GiB)": 35.92, "memory/max_active (GiB)": 33.78, "memory/max_allocated (GiB)": 33.78, "ppl": 1.00004, "step": 481, "tokens/total": 14554016, "tokens/train_per_sec_per_gpu": 29.72, "tokens/trainable": 219813} +{"epoch": 1.8828125, "grad_norm": 0.019354507327079773, "learning_rate": 1.0896808908553007e-05, "loss": 0.0001470642746426165, "memory/device_reserved (GiB)": 36.14, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00015, "step": 482, "tokens/total": 14584512, "tokens/train_per_sec_per_gpu": 35.33, "tokens/trainable": 220321} +{"epoch": 1.88671875, "grad_norm": 0.005853369832038879, "learning_rate": 1.0840061274830763e-05, "loss": 3.408601332921535e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.8, "memory/max_allocated (GiB)": 33.8, "ppl": 1.00003, "step": 483, "tokens/total": 14614608, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 220751} +{"epoch": 1.890625, "grad_norm": 0.0007542860112152994, "learning_rate": 1.0785151316412473e-05, "loss": 1.3063606274954509e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00001, "step": 484, "tokens/total": 14644784, "tokens/train_per_sec_per_gpu": 35.32, "tokens/trainable": 221193} +{"epoch": 1.89453125, "grad_norm": 0.0012432060902938247, "learning_rate": 1.0732081318325639e-05, "loss": 2.1572890545940027e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00002, "step": 485, "tokens/total": 14674944, "tokens/train_per_sec_per_gpu": 32.04, "tokens/trainable": 221642} +{"epoch": 1.8984375, "grad_norm": 0.015746230259537697, "learning_rate": 1.0680853489029501e-05, "loss": 0.00010861606278922409, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00011, "step": 486, "tokens/total": 14705456, "tokens/train_per_sec_per_gpu": 31.93, "tokens/trainable": 222086} +{"epoch": 1.90234375, "grad_norm": 0.0022054179571568966, "learning_rate": 1.0631469960323152e-05, "loss": 2.593602766864933e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.67, "memory/max_allocated (GiB)": 33.67, "ppl": 1.00003, "step": 487, "tokens/total": 14735440, "tokens/train_per_sec_per_gpu": 32.44, "tokens/trainable": 222518} +{"epoch": 1.90625, "grad_norm": 0.005776247940957546, "learning_rate": 1.0583932787256783e-05, "loss": 4.026497117592953e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00004, "step": 488, "tokens/total": 14765856, "tokens/train_per_sec_per_gpu": 36.78, "tokens/trainable": 223011} +{"epoch": 1.91015625, "grad_norm": 0.008843912743031979, "learning_rate": 1.0538243948046206e-05, "loss": 6.14341624896042e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.77, "memory/max_allocated (GiB)": 33.77, "ppl": 1.00006, "step": 489, "tokens/total": 14796128, "tokens/train_per_sec_per_gpu": 32.29, "tokens/trainable": 223464} +{"epoch": 1.9140625, "grad_norm": 0.0005093133659102023, "learning_rate": 1.0494405343990523e-05, "loss": 9.642845725466032e-06, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.76, "memory/max_allocated (GiB)": 33.76, "ppl": 1.00001, "step": 490, "tokens/total": 14826240, "tokens/train_per_sec_per_gpu": 29.84, "tokens/trainable": 223908} +{"epoch": 1.91796875, "grad_norm": 0.0012376867234706879, "learning_rate": 1.0452418799392985e-05, "loss": 1.869291918410454e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00002, "step": 491, "tokens/total": 14856512, "tokens/train_per_sec_per_gpu": 33.34, "tokens/trainable": 224365} +{"epoch": 1.921875, "grad_norm": 0.01617741398513317, "learning_rate": 1.0412286061485102e-05, "loss": 0.0001638774701859802, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 34.01, "memory/max_allocated (GiB)": 34.01, "ppl": 1.00016, "step": 492, "tokens/total": 14886848, "tokens/train_per_sec_per_gpu": 35.64, "tokens/trainable": 224847} +{"epoch": 1.92578125, "grad_norm": 0.573005735874176, "learning_rate": 1.03740088003539e-05, "loss": 0.013050932437181473, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.01314, "step": 493, "tokens/total": 14917216, "tokens/train_per_sec_per_gpu": 35.24, "tokens/trainable": 225329} +{"epoch": 1.9296875, "grad_norm": 0.0012499141739681363, "learning_rate": 1.0337588608872463e-05, "loss": 1.8755235942080617e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.84, "memory/max_allocated (GiB)": 33.84, "ppl": 1.00002, "step": 494, "tokens/total": 14947648, "tokens/train_per_sec_per_gpu": 33.23, "tokens/trainable": 225778} +{"epoch": 1.93359375, "grad_norm": 0.00981285609304905, "learning_rate": 1.0303027002633622e-05, "loss": 7.82605420681648e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.81, "memory/max_allocated (GiB)": 33.81, "ppl": 1.00008, "step": 495, "tokens/total": 14977728, "tokens/train_per_sec_per_gpu": 32.99, "tokens/trainable": 226217} +{"epoch": 1.9375, "grad_norm": 0.0030543936882168055, "learning_rate": 1.0270325419886884e-05, "loss": 3.3284988603554666e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.9, "memory/max_allocated (GiB)": 33.9, "ppl": 1.00003, "step": 496, "tokens/total": 15008032, "tokens/train_per_sec_per_gpu": 35.29, "tokens/trainable": 226689} +{"epoch": 1.94140625, "grad_norm": 0.035057004541158676, "learning_rate": 1.0239485221478599e-05, "loss": 0.00040418561547994614, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.0004, "step": 497, "tokens/total": 15038528, "tokens/train_per_sec_per_gpu": 32.2, "tokens/trainable": 227112} +{"epoch": 1.9453125, "grad_norm": 0.0012048225617036223, "learning_rate": 1.0210507690795292e-05, "loss": 2.2646674551651813e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.87, "memory/max_allocated (GiB)": 33.87, "ppl": 1.00002, "step": 498, "tokens/total": 15068832, "tokens/train_per_sec_per_gpu": 38.19, "tokens/trainable": 227577} +{"epoch": 1.94921875, "grad_norm": 0.0027389212045818567, "learning_rate": 1.0183394033710305e-05, "loss": 5.4524585721082985e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00005, "step": 499, "tokens/total": 15099136, "tokens/train_per_sec_per_gpu": 37.11, "tokens/trainable": 228053} +{"epoch": 1.953125, "grad_norm": 0.0021910013165324926, "learning_rate": 1.0158145378533583e-05, "loss": 3.143951471429318e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.82, "memory/max_allocated (GiB)": 33.82, "ppl": 1.00003, "step": 500, "tokens/total": 15129488, "tokens/train_per_sec_per_gpu": 31.33, "tokens/trainable": 228503} +{"epoch": 1.95703125, "grad_norm": 0.11515559256076813, "learning_rate": 1.0134762775964726e-05, "loss": 0.00122018251568079, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.92, "memory/max_allocated (GiB)": 33.92, "ppl": 1.00122, "step": 501, "tokens/total": 15159840, "tokens/train_per_sec_per_gpu": 35.73, "tokens/trainable": 228983} +{"epoch": 1.9609375, "grad_norm": 0.008556320331990719, "learning_rate": 1.0113247199049278e-05, "loss": 4.6531327825505286e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 34.03, "memory/max_allocated (GiB)": 34.03, "ppl": 1.00005, "step": 502, "tokens/total": 15190624, "tokens/train_per_sec_per_gpu": 31.47, "tokens/trainable": 229415} +{"epoch": 1.96484375, "grad_norm": 0.0009091690881177783, "learning_rate": 1.0093599543138205e-05, "loss": 1.879146475403104e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 503, "tokens/total": 15221104, "tokens/train_per_sec_per_gpu": 33.25, "tokens/trainable": 229862} +{"epoch": 1.96875, "grad_norm": 0.000701752258464694, "learning_rate": 1.0075820625850675e-05, "loss": 1.4095108781475574e-05, "memory/device_reserved (GiB)": 36.15, "memory/max_active (GiB)": 33.96, "memory/max_allocated (GiB)": 33.96, "ppl": 1.00001, "step": 504, "tokens/total": 15251520, "tokens/train_per_sec_per_gpu": 39.42, "tokens/trainable": 230366} +{"epoch": 1.97265625, "grad_norm": 0.0043007805943489075, "learning_rate": 1.0059911187040013e-05, "loss": 5.2763592975679785e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.97, "memory/max_allocated (GiB)": 33.97, "ppl": 1.00005, "step": 505, "tokens/total": 15282080, "tokens/train_per_sec_per_gpu": 38.02, "tokens/trainable": 230862} +{"epoch": 1.9765625, "grad_norm": 0.0014679227024316788, "learning_rate": 1.0045871888762893e-05, "loss": 2.3966062144609168e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.85, "memory/max_allocated (GiB)": 33.85, "ppl": 1.00002, "step": 506, "tokens/total": 15312304, "tokens/train_per_sec_per_gpu": 34.73, "tokens/trainable": 231312} +{"epoch": 1.98046875, "grad_norm": 0.2847539484500885, "learning_rate": 1.003370331525184e-05, "loss": 0.0026828020345419645, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00269, "step": 507, "tokens/total": 15342912, "tokens/train_per_sec_per_gpu": 36.68, "tokens/trainable": 231801} +{"epoch": 1.984375, "grad_norm": 0.013710313476622105, "learning_rate": 1.002340597289085e-05, "loss": 0.00012580624024849385, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.33, "memory/max_allocated (GiB)": 33.33, "ppl": 1.00013, "step": 508, "tokens/total": 15371152, "tokens/train_per_sec_per_gpu": 31.96, "tokens/trainable": 232218} +{"epoch": 1.98828125, "grad_norm": 0.47956013679504395, "learning_rate": 1.0014980290194387e-05, "loss": 0.008881042711436749, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 34.02, "memory/max_allocated (GiB)": 34.02, "ppl": 1.00892, "step": 509, "tokens/total": 15401984, "tokens/train_per_sec_per_gpu": 37.67, "tokens/trainable": 232718} +{"epoch": 1.9921875, "grad_norm": 0.021590515971183777, "learning_rate": 1.0008426617789489e-05, "loss": 0.00017190948710776865, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.94, "memory/max_allocated (GiB)": 33.94, "ppl": 1.00017, "step": 510, "tokens/total": 15432352, "tokens/train_per_sec_per_gpu": 36.58, "tokens/trainable": 233207} +{"epoch": 1.99609375, "grad_norm": 0.0010613418417051435, "learning_rate": 1.0003745228401215e-05, "loss": 1.6592484826105647e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.88, "memory/max_allocated (GiB)": 33.88, "ppl": 1.00002, "step": 511, "tokens/total": 15462720, "tokens/train_per_sec_per_gpu": 30.82, "tokens/trainable": 233631} +{"epoch": 2.0, "grad_norm": 0.003317455528303981, "learning_rate": 1.0000936316841296e-05, "loss": 5.1547642215155065e-05, "memory/device_reserved (GiB)": 36.16, "memory/max_active (GiB)": 33.89, "memory/max_allocated (GiB)": 33.89, "ppl": 1.00005, "step": 512, "tokens/total": 15493248, "tokens/train_per_sec_per_gpu": 36.62, "tokens/trainable": 234100}